&EPA
           United States
           Environmental Protection
           Agency
            Great Lakes National
            Program Office
            536 South Clark Street
            Chicago, Illinois 60605
                                  FP/a-905/4-79-016
An Experimental Study
of Lake Loading by
Aerosol Transport and
Dry Deposition
in the Southern Lake
Michigan Basin


-------
                                            EPA-905/4-79-016
                                            July,  1979
     AN  EXPERIMENTAL  STUDY  OF  LAKE  LOADING
   BY AEROSOL TRANSPORT  AND DRY  DEPOSITION  IN
        THE  SOUTHERN  LAKE MICHIGAN  BASIN
                       by

                Herman  Sievering
                   Mehul  Dave'
                Donald  A.  Dolske
               Richard  L.  Hughes
                  Patric  McCoy
             U.  S.  EPA Grant Number
                   R005301  01
                 Project Officer
                J.  Regan,  Chief
            Air Surveillance Branch
      U.S.  Environmental  Protection  Agency
Final  Report for the Period June 1976-July 1979
       GREAT LAKES NATIONAL PROGRAM OFFICE
      U.S.  ENVIRONMENTAL PROTECTION AGENCY
        536 SOUTH CLARK STREET,  ROOM 932
            CHICAGO, ILLINOIS 60605

-------
                                DISCLAIMER
     This report has been reviewed by the Great Lakes National  Program Office,
Region V, U.S. Environmental Protection Agency, and approved for publication,,
Approval  does not signify that the contents necessarily reflect the views and
policies  of the LLS0 Environmental Protection Agency, nor does  mention of trade
names or commercial products constitute endorsement or recommendation for use»
                                       n

-------
                             FOREWORD
This study was supported by a Great Lakes National  Program grant
to Governors State University for investigating the rate of de-
position of atmospheric transported pollutants to southern Lake
Michigan as a result of the industrial  complex located on the southern
shores of Lake Michigan.  The R/V Simons, a laboratory and lake water
quality sampling ship, operated by this office, was used to support
this project.  Data in this report covers a period  from June 1976 to
July 1979.

                                    Madonna F. McGrath
                                    Di rector
                                    Great Lakes National Program Office
                               m

-------
                                   PREFACE
     The enclosed document represents  our best  description  of  a  three year
effort to gather field data,  analyze the collected  data  by  computer  as well
as manually, and interpret the results in the context of Lake  Michigan
climatology.  Data base interpretation is far from  complete at this  writing
for the field program generally met with good and,  often, excellent  weather
conditions.  Coupled with a superior field crew,  conditions were  ripe for a
rich bounty of data.  Even in poor weather the  crew succeeded  upon the seas
of Lake Michigan.  Mehul  Dave, Don Dolske, Mike Eason, Jil  Forst, Vic Jensen,
Pat McCoy, Rich Rupert, Nell  Sutton, Keith Walther  and Ebbe Ward  are all to
be thanked for making possible the great success  achieved.   Their collective
spirit was the equal of this  unknown sailor's:

          A strong nor'wester's blowin, Bill
          Hark! don't ye  hear it roar  now?
          Lord help 'em,  how I pities  them
          Unhappy folks on shore now!

     No less spirited were grant secretaries  Cindy  Overton  and B.J.  Yates,
who created readable English  out of often unintelligible senbblings.
Research Associate Richard Hughes brought order and coherence  to  this docu-
ment as its Editor.  Special  thanks go to all the Research  Associates, Mehul
Dave, Don Dolske, Pat McCoy and Richard Hughes  for  their computer and labora-
tory analysis efforts and help in every other respect of grant activity  over
these past three years.

     Many persons among the Governors  State University support staff were
extremely helpful to the completion of this effort.  Don Douglas  and Ted
Andrews were especially helpful.  Staff members of  the USEPA are  also to be
thanked:  Bob Bowden and Jerry Regan for their  administrative  support;
Capt. McLain and the crew of the R/V Simons for bringing us through  those
storms; and the staff at EPA's Central Regional Laboratory. Aircraft support
of the project was gratefully supplied by the Research Aviation  Facility (RAF)
at the National Center for Atmospheric Research.  To Paul  Spyers-Duran and
Pete Drum of RAF special  thanks.

     As Principal Investigator of this grant I  accept full  responsibility  for
any errors of omission or commission.

July, 1979                                       Herman Sievering
                                                 Park Forest  South, Illinois
                                       IV

-------
                              EXECUTIVE SUMMARY


     A Lake Michigan experimental program to assess the contribution to Great
Lakes loading by atmospheric transport and dry deposition has been completed.
The first midlake trace metal and nutrients data base with associated meteorology
capable of establishing a climatology for both midlake aerosol mass concentrations
and mass transfer to Lake Michigan was collected during May to October 1977U
Trace metal concentrations were determined by atomic emission spectroscopy for
seventeen elements.  Of these, nine elements were significantly above instrumental
detection limits, filter blank concentrations, and laboratory contamination,
viz.. Al, Ca, Cu, Fe, Mg, Mn, Pb, Ti and Zn.  Nutrient species phosphate, nitrate,
and sulfate were determined by standard USEPA automated spectrophotometric tech-
niques.  A mid-Lake Michigan concentration climatology was estimated for each of
these twelve aerosol constituents.  A strong linear dependence of the variability
in all these concentrations upon atmospheric thermal stability and, thus, season
of the year was found.  No linear dependence on wind speed was found, but event
analysis suggests a strong dependence of concentration variability on wind
direction changes.  Source region dependence was found to cause as much as an
order of magnitude change in concentrations.  Source type identification has
been successful to the extent that soil, lake, and anthropogenically derived
components have been specified.  The percentage considered to be anthropogenically
derived for each of the nine trace elements above is:  Al, 0; Ca, 0; Cu,  _> 90%;
Fe, _> 65%; Mg, 0; Mn, _> 80%; Pb, >_ 95%; Ti, 0; and Zn, >_ 75%0  A detailed case
study of aircraft data taken on September 30, 1977 has investigated the change
in aerosol size distribution as a function of traverse over the lake and meteoro-
logical conditions.

     The rate of aerosol deposition from the atmosphere to the lake by dry
processes is taken to be approximately equal to the deposition velocity just
above the lake's surface.  A value for the deposition velocity was estimated
by a product of wind speed and diabatic momentum drag coefficient, accounting
for thermal stability effects.  This parameterized deposition velocity,_vd, is
compared to a small number of directly measured deposition velocities, "v^p,
calculated by a limited profile method.  The ratio of v
-------
      Annual  loadings of  twelve aerosol  constituents to the southern basin, were
 determined  by  a method which  accounts for the dependence of both v. 30



560

20
950

50
90
60
50
100 (5)


Run-Off (3)
13,000
490,000
140 (4)
1450
134,000
250
100
100
180 (4)
~ 3,000(6)


Brackets {} enclosing results indicate caution should be  used  in  interpretation --
see Section 5.
                                       VI

-------
(1)   This  worko
(2)   Gatz, D0FU,  1975,  Water,  Air and  Soil  Pollution,  !5,  ppu  239-251.
(3)   Winchester,  J0W0 and G.D.  Nifong,  1971,  Water, Air and Soil Pollution,
       ]_,  pp.  50-64.
(4)   Robbins,  et  al., 1972,  Proc. 15th Conf.  Great Lakes  Res0  International,
       Assoc.  Great Lakes Res0  pp.  270-2900
(5)   Murphy,  T0JU and P.V.  Doskey,  1976,  J. Great Lakes Res.,  pp. 60-70.
(6)   International  Joint Commission  on  The Great Lakes, T977 Water Quality
       Board  Report, Appendix  B,  Surveillance Subcommittee Reports
     Atmospheric loadings  by dry deposition to the southern  basin  of  Lake  Michigan
are clearly a significant  part of the overall  inputs  of Pb,  Zn,  other metals,  and
sulfate to the lake ecosystem.  This is especially true because  atmospheric  loading
contributes directly to the biologically active surface waters^  Accurate  quantifi-
cation of the atmospheric  input to the Great Lakes is an important continuing
research endeavor.  The identification of source types for Great Lakes atmospheric
inputs will require a still closer scrutiny of existing data.  Only then can manage-
ment and control objectives to meet Great Lakes water quality  goals be realized.

     This report was submitted in fulfillment of Grant No. R005301012 under
sponsorship of the U.S. Environmental Protection Agency, with  additional  support
from the research aviation facility and the computing facility of  the National
Center for Atmospheric Research, Boulder, CO 80301.  This report covers the
period June 1, 1976 to July 31, 1979, and work was completed as  of July 31,  1979.
                                       vn

-------
                             CONTENTS
Foreword ...... „ „ . . . . ...... . .... . „ „ „ 0 „ i-ji
Preface. . . . ,	„ „ . „ „	   .          -jv
Executive Summary. 	 0 . 0 . „ . „ 0 . . . „ 0 . 0 . 0 „ u   v
List of Figures0 „ . . „ . „ 0 . 0 . u . . . „	0 . 0  -jx
List of Tables ........ . . . „ . . . . „ . . „ . „ „ „ . .  xi
List of Symbols. . ... „ . . . . „ . . „ ... 0 ....          xii

   1„  Introduction. „ „ 0 	 ».....»...   1
   2.  Conclusions and Recommendations 	 „ . . „ . 0 0 . 0   4
   3.  Description of Experimental Program „ „	0 „   5
         3ol  General Procedures „ ......„„...„. 0 ..   6
         3.2  Midlake Sampling Procedures. <> „ . . . „ „ „ . „ . 0   9
         3U3  Nearshore Sampling Procedures	<,.„..„.  11

   4.  Data, Analysis and Basic Results,, . . . . „ .... 0 „ „ „  12
         4.1  Results of Chemical Analysis „ . „ . . „ . „ . 0 0 „  12
         4.2  Physical Characteristics of Aerosols 	  17
         4.3  Meteorological and Climatological Considerations „ 0  23
         4.4  Removal of Aerosols at Air/Water Interface . 0 «, 8 0  33

   5.  Interpretation of Results . 0 ... 0 ...... „ 0 „ „ „  45

Figures. .  « . . . „ . „ „ . . „ ... 0 .... „ „ .... „ . „  55
Tables .	 .  100
References » . . . „ . » . . . . 0 0 . . . .	„ . . .  122
Appendices .. o ......„„. 0	„ . . .  128

   A.  Factor Analysis -- A Technique for Interpreting
         Environmental Data	 . „.........„.„.  128
   B.  Micrometeorology Related to Diabatic Drag Coefficient
         Determination ......... ...» ...... u 0 .  151
   C.  Trace Element Loading of Southern Lake Michigan by
         Dry Deposition of Atmospheric Aerosol  	 ....  157
   D.  Aerosol Sample Preparation Procedure for Emission
         Spectroscopy Analysis ..................  171
                                  vi n

-------
                                 FIGURES


Number                                                                       Page
  1   Location and Dates of Sampling Periods and Source Regions  .......  55
  2   Aerosol  Collection Efficiencies for Three-Stage Cascade Impactor.  .  .  .  56
  3   R/V Simons and Sampling Boom	.	„....„...„.  57
  4   68th St. Water Intake Crib and Sampling  Windows 	  »  .  »  .  .  .  .  58
  5   Distribution of Occurrences of Al  Concentrations,, <>	„  .  „  .  59
  6   Distribution of Occurrences of Ca  Concentrations ............  60
  7   Distribution of Occurrences of Fe  Concentrations. .......  ....  61
  8   Distribution of Occurrences of Mg  Concentrations. .  ...  ....  ...  62
  9   Distribution of Occurrences of Mn  Concentrations. .  „  „ „  .  .  .  .  .  „  .  63
 10   Distribution of Occurrences of Pb  Concentrations„ 	  „  .  ....  64
 11   Distribution of Occurrences of Ti  Concentrations. ...........  65
 12   Distribution of Occurrences of Zn  Concentrations. .  „  „ .  .  ......  66
 13   Pb Enrichment Factor vsu  Al Concentration  ............  0  ..  67
 14   Zn Enrichment Factor vs.  Al Concentration  .  .  . . „	  68
 15   Mn Enrichment Factor vs0  Al Concentration  	  ..........  69
 16   Fe Enrichment Factor vs.  Al Concentration  ................  70
 17   Ti Enrichment Factor vs.  Al Concentration  .  .  . .	»  .  71
 18   Mg Enrichment Factor vs.  Al Concentration  .  ........  „  .....  72
 19   AN/A (log r) vs. log r for 17-20 May 77	„  „ ........  73
 20   AN/A (log r) vs. log r for 7-9 June 77.  ...  .............  74
 21   AN/A (log r) vs. log r for 14-19 Aug 77	  75
 22   AN/A (log r) vs. log r for 26-30 Sept  77	  76
 23   AV/A (log r) vs. log r for 17-20 May 77  .  „  „  „ . „  ..........  77
 24   AV/A (log r) vs. log r for 7-9 June 77.  ...........  	  78
 25   AV/A (log r) vs. log r for 14-19 Aug 77  ...  	 ........  79
 26   AV/A (log r) vs. log r for 26-30 Sept  77.  .............    .80
 27   AV/A  log r  vs. log r for 22:15 CDT,  17 May 77 to 02:53 CDT 18  May  77.  81
 28   AV/A (log r) vs. log r for 03:16 CDT,  18 May 77 to 06:18 CDT 18  May  77.  82
 29   AV/A (log r) vs. log r for 08:20 CDT,  18 May 77 to 11:20 CDT 18  May  77.  83
 30   AV/A (log r) vs. log r for 12:00 CDT,  18 May 77 to 17:25 CDT 18  May  77.  84
 31   AV/A (log r) vs. log r for 17:46 CDT,  18 May 77 to 21:10 CDT 18  May  77.  85
 32   AV/A (log r) vs. log r for 22:10 CDT,  18 May 77 to 01:20 CDT 19  May  77.  86
 33   AV/A (log r) vs. lp_g r for 01:44 CDT,  19 May 77 to 06:16 CDT 19  May  77.  87
 34   CD,  CH and G£ vs.  u at 5  m Sampling Height	88
 35   Cumulative Frequency of Occurrence of  Wind Speed	89
 36   Ratio of CDD to CD vs.ls0!	90
 37   Cumulative Frequency of Occurrence of  AT Over  Southern  Lake Michigan.  .  91
 38   Aerosol  Number Concentrations  (cm~3) Measured  in  the Size Range
        0.25 < d < 20 ym for the Midway  Airport  to Midlake Flight Leg
        Collected on 30  Sept 77 from 06:46 CDT to  06:58 CDT	92
                                     ix

-------
Number                                                                       Page

 39   Diagramatic Representation of The Air/Water Interface .  .  e .  . „  „ „ . 93
 40   Normalized Aerosol Mass Fraction for All  1977 Sampling.,  .  . .  „ 0  u „ „ 94
 41   Normalized Aerosol Mass Fraction for Overlake Transport Samples .  „ . . 95
 42   Normalized Aerosol Mass Fraction for West Shore Source Samples„ „  . „ . 96
 43   Normalized Aerosol Mass Fraction for East Shore Source Samples	97
 44   Normalized Aerosol Mass Fraction for Southeast Shore Source Samples „ . 93
 45   Normalized Aerosol Mass Fraction for Chicago/Gary Source Samples0  „ . «, 99

-------
                              LIST OF TABLES


Number                                                                   Page

  1    Listing of Meteorological  Parameters  Measured  and  Type of Sensors
        Used in the Lake Michigan Study 0	„  „	.  .  <>  «   107
  2    Detection Limits  (U),  Filter Blank  Corrections  (3),  and Typical
        Sample Elemental  Concentrations (C) in  the ICAP-AES Liquid
        Sample (all yg-JT1)  „	.	   108
  3    Geometric Mean Concentrations (yg-iH)  of Certain  Elements in
        Lake Michigan Water  at 87°00'W by  42°00'N  	 .....   109
  4    Geometric Mean Concentrations in Air (ng-m~3)  of Certain Elements
        at 87°00'W by 42°00'N  During 1977  Sampling Periods	  .  .   110
  5    Characterization  of Mid-Lake Michigan Aerosol  with Respect to
        Several Chemical  Species. ....................   Ill
  6    Correlation Coefficientsjjatrix Among Trace  Elements, Total
        Aerosol Mass, Uc, and  AT Using 40  filter Sets' Data .  . .  .  .  .  .   112
  7a  Trace Element Analysis  Results for Data Sets Collected at the
        Nearshore Site  in 1978.  .......	......   113
  7b  Results of Passive Aqueous Extraction of  Nearshore Aerosol Samples.   114
  8    50 percent Collection  Efficiency Diameters for Hi-Volume Sampler
        Integrating Nephelometer and Active Scattering Aerosol
        Spectrometer (ASAS)  .  .  	  ... .......   115
  9    Sample Time, Meteorological Parameters, and Aerosol  Data for
        Data Sets 20050 - 20110	 .	   116
 10    Meteorological and Aerosol Data for  the Profile  Data  Sets 	   117
 11    Binned Data Sets, Using  AT as the Defining Parameter. .......   119
 12    Binned Data Sets, Using  U5 as the Defining Parameter .......   120
 13    Binned Data Sets, Using  Source Region as  the Defining Parameter  .  „   121
 14    Annual Loadings of Certain Trace Elements and  Nutrients to the
        Southern Basin  of Lake Michigan (103  kg-yr~')  „  „  . .  . .  .  .  .  .   122
 15    Factor Analysis Results  and Interpretation of  Factors .......   123
 16    Meteorological Influences  Upon the 39 Filter Set Data Base.  .  .  .  .   124
 17    Factor Analysis of 19  Filter Sets' Data with RI'B >. 0.03  ......   125
 18    Factor Analysis of 20  Filter Sets' Data with RI'B < 0.03  . .  „  .  .  „   125
 19    Factor Analysis of 15  Filter Sets' Data with Travel  Time to
        Shore >_ 3.5 hrs	   126
 20    Factor Analysis of 17  Filter Sets' Data with Travel  Time to
        Shore _< 3.0 hrs.   .  .  „  .	  .  .  .  .   127
 21    Percentages of Soil, Lake  and Non-Naturally  Derived  Mid-Lake
        Michigan Atmospheric  Surface Layer  Metal Concentrations .  .  .  .  .   128
                                      XI

-------
                              LIST OF SYMBOLS



SYMBOL	MEANING	UNITS

                                                                2
A               area of southern Lake Michigan basin           m

                                                                 -4 -1
b   t           aerosol light scattering coefficient           10  m


Cry              neutral drag coefficient for momentum


CDQ             diabatic momentum drag coefficient


Crjr             momentum form drag coefficient


Cr              neutral drag coefficient for water vapor


CH              neutral drag coefficient for heat

                                                                   _3
C               elemental concentration in aerosol             ng-m


Cu              concentration in filter blank                  uq-l~
                                                                   ~°
C               total aerosol mass concentration               ng-m

                                                                   _3
C  f            concentration of a reference element           ng-m


D               molecular diffusion coefficient                m-s


d               aerosol diameter                               ym


E.              transfer efficiency


EF              aerosol elemental enrichment factor


F               frequency of occurrence for range of clima-
                tological conditions


FC              fine to coarse particulate fractional ratio


g               acceleration due to gravity                    m-s


H               depth of mixed layer                           m


h               height above nean water level                  m


I               intensity of turbulence

-------
                                                                 2  -1
 KD              eddy diffusion coefficient                      m -s
                                                                 2  -1
 K               momentum diffusion coefficient                  m -s
 Ld              instrumental detection limits                  yg-l~
 N               aerosol number concentration                   m"
 Rig             Bulk Richardson number                         s
 R               resistance to mass transfer                    s-cnf
 Rjl              resistance to mass transfer in turbulent layer s-cm
 r               radi us                                         ym
 S               stop distance                                  m
Sc              molecular Schmidt number, defined as v/D
Sc              turbulent Schmidt number, defined as Km/KD
                                                                   o   o
 S               atmospheric stability parameter                 C-s -m
 Th              air temperature at height h                    °C
 T^              air temperature at 5 m height                   C
 T               water surface temperature                      °C
 T               bulk water temperature                          C
 AT              atmospheric thermal stability                  °C
 t               time                                           s
 u^              windspeed at height h                          m-s~
 Ug              wind speed at 5 meter height                   m-s~
 u               surface water current                          m-s~
 u*              friction velocity                             cm-s~
                                                                 3  -3
 V               aerosol volume concentration                   ym -m
 v               velocity                                       m-s
 v.              deposition velocity                            m-s~
                                 xm

-------
dp
 5
WD
Xn,
XR
XTL
a
3
Y
6
e

n
X
V
n
p
a
T
bulk parameterized deposition velocity
mean bulk deposition velocity
profile method deposition velocity
mean profile deposition velocity
vertical wind at 5 m height
wind direction
mass transfer
remainder mass transfer
turbulent layer mass transfer
empirical exponent
filter blank correction factor
a constant
laminar sublayer thickness
a constant
a constant
a constant
wavelength
kinematic viscosity of air
elemental mass percent of aerosol
density of air
standard deviation
time of year with no precipitation overlake
                                                            cm-s
                                                                 -1
                                                            cm-s
                                                                 -1
                                                            cm-s
                                                            cm-s
                                                                  -1
                                                                  -1
                                                            cm-s
                                                              o
                                                                 -1
                                                               maqnetic
                                                                  -1
                                                              cn-s
                                                              cm-s
                                                                  -1
                                                              cm-s
                                                                  -1
                                                              yg-1
                                                                  -1
                                                              m
                                                              ka-m
                                                                  -3
                                 xiv

-------
                                  SECTION 1

                                INTRODUCTION
     The densely populated and heavily industrialized southwestern shore of
Lake Michigan represents a significant and expanding source of anthropogenic
aerosol to the atmosphere.  The combustion of fossil fuels in residential,
industrial, and transportation activities, as well  as manufacturing processes
such as steel and cement making, are principal sources,,   Pollutant aerosol
emitted in near shore urban/industrial areas provides an input potential for
loading of Lake Michigan via wet and dry removal <,   50 percent or more of the
time, prevailing winds give rise to greater than 80 km long trajectories over
the lake for air masses passing through the Chicago/Gary area (e.g. Sievering
and Williams, 1975).  Not only the well being of the urban population, but
also the well being of the lake ecosystem, may therefore be linked to atmos-
pheric pollution by wet and dry deposition.  The transfer of atmospheric pollu-
tion constituents to land and water surfaces in the Great Lakes region by pre-
cipitation has been extensively documented in the literature (e.g. Murphy and
Doskey, 1976; Torrey, 1976; and Eisenreich, Emmling and Beeton, 1977).  Air
pollution, and particularly dry deposition of atmospheric aerosol, was long
assumed to be a negligible fraction of the total mass flux of most trace
elements to the lakeu  Increasing concern for water quality of the Great Lakes
led to re-examination of that assumption.  Estimates of trace element loadings
to Lake Michigan by aerosol deposition have often been made by using a transfer
efficiency (Et) approach.  Winchester and Nifong (1971)  calculated an emission
inventory for the Milwaukee, Chicago, and northwest Indiana region and assumed
an Et of 0.10 where:

                    F   =   jollutant loaded to lake             r-\\
                     t      pollutant emitted at source  °
Skibin (1973) suggested that 0.10 may be too low a  value for Et; consideration
of mesoscale circulation effects on aerosol trajectories,  as well  as tempera-
ture stability and wind speed effects on aerosol deposition, made a value of
0U25 seem more probable.  Sievering (1976) presented an  overlake aerosol trans-
port and deposition model, and proposed an E^ of 0.20 to 0.40 for total  aerosol
mass transfer to Lake Michigan and 0U01 to 0U15 for trace  elements, dependent
upon elemental aerosol size distribution and season of the year.  Gatz (1975a
and 1975b) calculated an emission inventory based on the chemical  composition
of Chicago area aerosol and showed precipitation and dry deposition to be
about equal contributors to trace element loadings  of the  lake.   Very large
uncertainties in the emission inventory calculations and Et models of these
studies pointed to the need for overlake measurements of aerosol concentration,
composition, and deposition.

                                        1

-------
     Very little has appeared in the literature regarding measurements of
overlake aerosol composition and deposition,,  Eisenreich, Emmling, and Beeton
(1977) and the International Joint Commission review (1978) used primarily
shore-collected bulk precipitation samples, i.e. wet and dry removal combined,
to estimate atmospheric inputs to the lake.  However, the bulk samples give
no better than an order of magnitude idea of the dry deposit!onal component
of total deposition (Cadle, 1975).  Schmidt (1977) gathered overlake aerosol
trace element data and applied wind tunnel measurements of the dry deposition
velocity (Sehmel and Sutter, 1975) to estimate loadings,,  Other workers inves-
tigated nutrient inputs to the Great Lakes from the atmosphere (Delumyea and
Petel, 1977; Murphy and Doskey, 1976) by precipitation and dry deposition.
The level of uncertainty in estimates of trace element deposition to Lake
Michigan remained at nearly an order of magnitudeu  This research has sought
to reduce that uncertainty by observation of aerosol composition and deposi-
tion velocity as part of an intensive midlake data collection experimental
program, described in Section 3.

     The large uncertainty in the aerosol dry deposition velocity at and
above the air/water interface is usually not stated in dry loading estimates
found in the literature.  At a particular sampling height, h,  the deposition
velocity, v,. , can be defined as a function of aerosol  radius, r, by:

                      vdh  -  Fluxh(r)/Ch(r)              (2)


where the upward or downward mass flux of aerosol  through a unit area at h is
divided by the mass concentration C^ (r) of aerosol  at h.  Uncertainty may be
introduced by not considering the variation in v<-|  with  aerosol  radius.  Using
theoretical  arguments, Sehmel (1971) has shown that v^  (0.5 ym)  is approxi-
mately equal to 10 • v, (0.05 urn).  Yet wind tunnel  data by Sehmel and Sutter
(1974) on deposition to a water surface indicated  that v^ (0.5 ym) is approxi-
mately equal to vj  (0.05 ym).  Thus, an order of magnitude uncertainty  con-
cerning  loading estimates for this size range could be introduced by v
-------
climatological variabilityo  Most theoretical models of dry deposition do not
consider this variation with thermal stability,,  There are two direct measure-
ment methods — the eddy correlation and concentration gradient or profile
methods --* which can be used to experimentally measure VQ-,  However, neither
method is well suited to routine use in the field, especially overlake.  The
use of a limited profile method, along with a parameterized measurement which
simplifies the experimental demands, was found to be a good compromise approachu
In particular, a drag coefficient parameterization for mass transfer was used
here, with wind speed and thermal stability as parameters.  Hess and Hicks
(1975) have shown that these two parameters dominate the steady state consid-
eration of meteorological variables for Great Lakes mass transfer.  By this
method a bulk deposition velocity at a measurement height, h, (v^) is
determined by


                       Vdh  =  ("h - V  '  CDD>               (3)

where u^ is the mean wind speed over the sampling time at the height of
measurement, u0 is the water surface speed, and CQD is the diabatic momentum
drag coefficient.  A complete discussion of the use of CQD appears in Section
4.3.  Note that this method does not give VH as a function of both r and hu
The data base obtained by the limited profile method (described in Section 4.4)
was used in conjunction with data from an aerosol size spectrometer to consider
the variation of v^  with aerosol radius.

     Of greater significance to Great Lakes atmospheric loading estimates than
variability in v^ alone is the possibly synergistic interaction of the temporal
variabilities in C^ and v^ As equation (2) indicates, the flux of aerosol to the
lake surface is depending upon the product of two factors:  v^ and C^,u  Both
factors are quite variable across the range of meteorological conditions prevail-
ing at midlake throughout the year.  Wind speed and thermal stability have a
strong influence upon aerosol transport and deposition processes.   In order to
estimate annual total loadings from data collected during a necessarily limited
sampling effort, some means of relating changes in VH and C to meteorological
variation must be included in the sampling and analysis (see Section 4.3).
Moreover, other environmental factors, such as wind direction (WD) changes which
transport aerosol from differing source regions overlake, affect the loading rate
(see Section 4.2).  Aerosol of varying chemical  and physical  characteristics must
also be considered in transport and deposition calculations (see Sections 4.1 and
4.2).  In this research, the approach used was to simultaneously collect meteoro-
logical  data and aerosol samples.  Sampling was performed around the clock during
four several-day-long sojourns at a midlake station, during which  distinct aerosol
samples were collected in many successive 3-to 6-hour periods.  This restricted
length of exposure time for the aerosol collection media resulted  in each sample,
i.e. each value of C, being identified with a value of v^ which was based upon
data collected within a period of well defined and fairly constant meteorological
conditions.  Annual loading estimates were then made, using meteorological
parameters as weighting factors in calculations, and extrapolating the results
by reference to the expected overlake climatology (NOAA, 1975).   These loading
estimates are considered in Section 5.  It is concluded that a 2-to-3-fold
uncertainty in loading estimates remainsu

-------
                                  SECTION 2

                       CONCLUSIONS AND RECOMMENDATIONS
     A major conclusion of this work is that at least 60 percent of the  total
Pb input, 30 percent of the total  Zn input,  and 20  percent of the total  Fe
input to the southern basin of Lake Michigan is by  dry deposition of atmospheric
aerosol.  It was also found that major inputs of sulfate and nitrate are by dry
loading.  Phosphorus input by dry loading is about  equal to precipitation inputs.

     The percentage of anthropogenically derived aerosol at the sampling point
near the middle of the southern basin of Lake Michigan for each of nine  trace
elements was calculated to be:  Al, 0; Ca, 0; Cu, _> 90 percent; Fe, _> 65 percent;
Mg, 0; Mn, ^.80 percent; Pb, >_ 95 percent; Ti, 0; and Zn >_ 75 percent.   Aerosol
trajectory considerations show that more than 75 percent of the anthropogenically
derived aerosol are from the Chicago/Northwest Indiana source region.

     The application of a factor analysis technique (Appendix A) to the  midlake
aerosol chemistry and meteorological data base specifies six uncorrelated factors.
Physical significance was attached to each factor,  resulting in several  general
conclusions:

     1.  The variance in midlake aerosol metal concentrations is linearly
         independent of wind speed, but strongly dependent upon temperature
         stability of the surface layer.

     2.  The variance in midlake mass concentration is more dependent upon
         fine particulate mass (with high Pb, Zn, and Mn content) than upon
         coarse particulate mass.

     3.  Midlake aerosol metal concentration is more dependent upon nearshore
         sources than upon long range transport^

     Determination of the rate of dry deposition to the lake surface by  a bulk
parameterization technique (Appendix B) resulted in values for deposition veloci-
ties which were no worse than a factor of three removed from direct measurement
and generally within a factor of two.  Accuracy of  the bulk parameterization was
identified by simultaneously measuring the deposition velocity by direct aerosol
profile (i.e,,number concentration gradient) measurement.

     Midlake sampling necessitated the use of an anchored ship platform.  As a
result, sampling took place during only 2 percent of the year.  Overlake clima-

-------
tological data were used to estimate annual  loading rates.   Thus,  the results
above are not specific to the year of sampling but rather are extended to  a
climatologically averaged year.

     Elemental analysis of the nearshore samples  has only recently been completed.,
Tentative interpretation shows a three-fold  higher concentration for most  trace
elements than at midlake, with a major portion being in the soluble phase.

IT IS RECOMMENDED THAT:

10  A materials balance analysis be carried out, using the midlake data base
    in conjunction with emission inventory data on mass, trace elements and
    sulfate differentiated by source region and source type in the greater
    Lake Michigan basin region.  This analysis should afford the identification
    of coal and oil burning sites, automotive and other source types, and an
    estimation of their percentage contribution to atmospheric dry loading.

2.  Further interpretation of the combined midlake and nearshore data base
    using factor analysis and aerosol characterization analysis techniques
    should be pursued.  The gross effects of transport and transformation upon
    source region and source type emissions would be significantly clarified
    by these analyses„

3.  Aircraft aerosol data used on a case study basis will further clarify the
    transport and transformation processes.  In addition, aerosol number and
    volume distribution analysis and interpretation can help identify synoptic
    and mesometeorological effects upon aerosol and trace element concentrations
    as a function of location over the lake.

4U  Numerical modeling should be applied to reduce the uncertainty in the exten-
    sion of results to other geographical areas and    time framesu  It is
    expected that modeling will also play an important role in the understanding
    of the process of aerosol transfer near the air/water interface.  A model  of
    turbulent inertial deposition in the buffer layer and laminar sublayer may
    provide a clear picture of aerosol mass transfer at the interface.

5U  Direct measurement of aerosol deposition to the Great Lakes by either the
    profile or eddy flux method under varying meteorological conditions should
    be supported.

6.  Future Great Lakes atmospheric studies should be performed in the context of
    year-round meteorological sampling at several sites on the lakes to establish
    the representativeness of ship-based meteorology, both from a temporal and
    spatial viewpoint.  Temporally, it is necessary to verify that sampling
    aboard ship fits the climatological data for the Great Lakes.  Spatially,  a
    confirmation that single site sampling represents the majority of a Great
    Lake's surface must be further considered.

-------
                                  SECTION 3

                     DESCRIPTION OF EXPERIMENTAL PROGRAM
3ol  GENERAL PROCEDURES

     In order to compile a data base for the estimation of annual  dry deposi-
tion loadings, investigations in two general categories were concurrently
pursued.  These categories were:  1) meteorological  conditions,  both at the
mesoscale (affecting aerosol transport) and microscale (affecting  aerosol
deposition); and 2) the physical (aerosol size distribution and  mass concen-
tration) and chemical (trace element and nutrient content) nature  of overlake
aerosol.  Meteorological data and aerosol samples were gathered  from sampling
sites at midlake (87  00' W by 42° 00'  N) aboard the United States Environ-
mental  Protection Agency (USEPA) R/V Roger R. Simons during 1977,  and near-
shore (87  32' W by 41  47' N) at the City of Chicago 68th Street  water intake
crib during May through November, 1978 (see Figure 1).  Aircraft overflights
were made during June and September, 1977 and May, 1978 by a National Center
for Atmospheric Research (NCAR) Beechcraft Queenaire. for collection of
meteorological and aerosol size spectrometry data from near surface up to
2000 m.  Surface meteorological reports from the National Weather  Service
(NSW),  Argonne National Laboratory, and U.S. Coast Guard were gathered from
various stations around southern Lake Michigan.  NWS low-level  soundings at
Chicago's Midway Airport complemented the soundings  taken by the NCAR Queen-
aire.

     At the sampling sites, a portable array of meteorological  sensors was
operated (Table l)u  Analog output from the five meter height wind speed and
direction (u§ and WD), temperature ^5) and relative humidity (RH) sensors
was scanned at 1 Hz by a signal conditioner/data acquisition system  (Weather
Measure Corp. (WMC), Sacramento, CA, model SC601).  The 1 Hz readings were
stored and fifteen-minute means and standard deviations (a) in all parameters
calculated by a microprocessor  (WMC #M733).  These mean and a values, along
with manually measured water surface temperature (TQ) readings were hardcopy
printed and recorded on magnetic tape (Texas Instruments Inc., Houston, TX,
model ASR/KSR 733).  Surface current (u0) and bulk water temperature (Tw) were
observed and recorded at 4 to 7 hour intervals.

     Data collection was done around the clocku  In order to divide each
sampling day into manageable time segments, data collection was done within
3  to 6 hour'periods, referred to as "data sets."  Each data set, because of
its limited duration, corresponds to a period of fairly constant meteorologi-
cal conditions.  The occurrence of rain or fog events precluded any sampling,

-------
since this study was directed at dry deposition only.  At the beginning of
each data set, aerosol sampling media were replaced and water samples were
taken.  Thus, aerosol and water samples can be classified according to values
of meteorological parameters which prevailed during each data set0  In general,
a minimum of three hours elapsed time was required per data set to provide
sufficient loading of the Hi-Volume filters to be above detection limits of
the chemical analysis procedure used.  Past that minimum time, rapid and
persistent changes in 115, WD, 15, or TO were cause for the termination of a
data set.  Such changes were defined as two successive 15-minute mean values
for a given parameter being more than la removed from the running mean0  A
maximum time of about 8 hours was allowed so that no data set was biased by a
widely differing run time»  Additional constraints upon data set initiation/
termination, unique to each sampling site, will be discussed as part of each
site's description.

     Aerosol samples were taken with standard Hi-Volume samplers (General
Metal Works, Inc., Cleveland, OH #GMW 2000).  Samples for gravimetric analysis
of total aerosol mass concentration were collected on 20 by 25 cm glass fiber
filters (Sierra-Misco, Carmel Valley, CA, #C305); samples for chemical anal-
yses were collected on cellulose fiber substrates (Sierra-Misco #P252A) and
20 by 24 cm filters (Sierra-Misco #810A)0  The three-part cellulose media
were exposed at 1.18 m3 - min"1 (40 SCFM) in a modified cascade impactor
(Sierra  5-stage #230), using the #1, #3, and backup stages only.  Collection
efficiency curves for this configuration were determined by Sievering, et al.
(1978) and were found to resolve well  two aerosol size fractions,,   A coarse
particulate fraction(diameter, d > 1U0 urn, 90% confidence) is collected on
the first impactor stage substrate, while a fine particulate fraction
(d <1.0 urn, 90% confidence) is collected on the backup filter.  The third
stage acts as an intermediate step which improves the separation between
size fractions collected on the other two filter stages (see Figure 2).  Two
such cascades of three-stage cellulose filter media were simultaneously
exposed on separate Hi-Vols during each set,,  One cascade of filters was
prepared for trace element analyses, while the other was analyzed for aerosol
nutrient content,.

     Aerosol samples collected on one of each pair of three-stage cascades
were analyzed for trace elements at the USEPA Central Regional Laboratory in
Chicago.  The filters were placed in acid-washed fused quartz trays and low
temperature ashed in 02 at 75 watts.  The residue was then dissolved in HN03
and distilled deionized water.  Trace element analysis was done by Inductively
Coupled Argon Plasma atomic emission spectroscopy (ICAP-AES) (Jarrell-Ash,
Pittsburgh, PA, Plasma AtomComp 750),   In more than 85 percent of the data
sets, concentrations of Ca, Mg, Cu, Fe, Mn, Pb, Ti, and Zn were above ICAP-AES
instrumental detection limits (L^K  The limited 3 to 8 hour exposure time of
the filters resulted in concentrations of Na, Ba, Cd, Co, Mo,  Ni,  and V which
were below Lj in at least 50 percent of the data sets (Table 2).   Therefore,
loading calculations for these metals  are of limited validity,.  Post-analysis
statistical review of the data revealed contaminant or otherwise anomalous
data for B, so that this element was not used in loading estimates,,   Filter
blanks were subjected to all  handling  and analysis  procedures, except the

-------
collection of ambient aerosol„   The blank correction (3)  applied  to  the
ICAP-AES results for the exposed filters was defined as:

                               3  =  Cb  +  la              (4)

where Ck is the mean concentration of the element  in filter  blanks, and a is the
standard deviation in that mean value.   For those  elements which  were  above
L 95  percent.  The
combined extracts of the 3 hour leaching process were divided  into  separate
aliquots for analyses for P, $64, and NOo/W^.  The P subsamples  were  digested
in (^4)2 $2 Og and CH2S04 to convert all phosphorus to orthophosphate.   The
digestate was then colorimetrically analyzed for orthophosphate by  an  auto-
mated molybdate method.  The $04 subsamples were passed through a Na - form
cation-exchange column to remove multivalent metal  ions.   Sulfate content
was then determined by an automated methyl thymol blue method.   The  N02/N03
subsamples were passed through a Cd/Cu column, reducing all  nitrate to nitrite.
Nitrite was then colorimetrically determined by  an automated diazotization-
coupling methodu  The automated colorimetric analyses were performed following
standard autoanalyzer methods (Jirka, et al.; USEPA, 1974; Technicon Instru-
ments Corpu).

     Several  optical particle size and concentration measuring instruments
were used to collect information on the physical character of  overlake aerosol.
An integrating nephelometer (Meteorology Research, Inc.,  Altadena,  CA, #1550)
continuously monitored the aerosol light-scattering coefficient and thus,
variations in aerosol mass concentration.  Condensation nuclei counters
(Environment One Corp., Schenectady,  NY, models  #E1012E and  Rich  100)  were
also used to routinely observe submicron particle  concentrations.

     The behavior of aerosol in the diameter size  range,  001  < d <1.0 urn is
of particular interest, since trace elements are strongly associated with this
fine particulate fraction*  This is especially true at a  midlake  sampling site
which is substantially removed from sources of large aerosol.   Information
concerning aerosols in this size range was obtained from  an  Active  Scattering
Aerosol Spectrometer (ASAS, Particle Measuring Systems, Boulder,  CO, #PMT300).
This instrument measures in situ aerosol number  by detection of scattering
internal to the cavity of a Helium-Neon laser.  Ambient air  is drawn through
a mini-wind tunnel in which the laser resides.  Number concentration separated
into 60 size channels was recorded on magnetic tape.  Accurate flow rate con-
trol (better than 1%) and low coincidence error  provided  high  reproducibility
and precision,.  If an average density is assumed,  mass information  can be
obtained.

-------
     During sampling on board ship and at the crib,  a 730 cm2 cross sectional
area entry tube facing into the wind at the tip of the boom was tapered to
match the 20 cm2 ASAS sampling cylinder.   A 4.6 m-s"1 entry speed allowed
nearly isokinetic sampling from ~3 to ~6  m-s"1 ambient winds„  The 2 m long
tube was grounded to minimize loss of electrically charged aerosol to the wall.
Results from a wind tunnel test were used to correct for the tube wall loss
that was observed.  The instrument gathered data in the range 0.09 < d <
3.53 ym in size intervals from almost 0.01 ym for the smallest sizes to 0.2
for the 1 ym and larger sizes,,  The ASAS returned measurements of sufficient
precision to provide vertical profiles of surface layer aerosol concentra-
tions in some circumstances.

     Besides being  based at the crib and on board the ship, an ASAS was
mounted on the NCAR Beechcraft Queenaire research aircraftu  Also on board
were sensors for temperature, dewpoint, pressure, height, and other meteoro-
logical parameters,,  Instrumentation for the  analysis of atmospheric turbu-
lence was included  in the September 1977 and  May 1978 flights.  The air
sampled by the ASAS on the aircraft was drawn from the Pennsylvania State
University isokinetic intake  (Pena, Norman and Thomson, 1977) through copper
tubing and flow-controlled to produce a variation in the aircraft monitoring
sample volume flow  rate of about  2 percent.   Corrections for aerosol loss to
tubing walls was again based on wind tunnel data which was gathered using
the actual field instruments and  set-up.  The instrument used in  the 1977
flights provided data in the range 0.23 ym  < d < 2U8 ym in size intervals of
0.01 ym - 1.0 ym.   The aircraft platform provided data which could be used to
investigate vertical and horizontal distributions of meteorological conditions
as well as aerosol  concentration.  Also on board for these 1977 flights was
the Classical Scattering Aerosol  Spectrometer (CSSP), a device very similar to
the ASAS, which gathered data in  the range 1.1 ym  
-------
The 40 km or greater upwind fetch allows a 10 to  30 m thick "surface layer"
of air, just above the water, to approach equilibrium with  respect to surface
effects.  As long as the lake surface appears constant to the  air passing  over
it, the mass flux within the surface layer is constant throughout its vertical
extent (Dyer and Hicks, 1970; Kraft, 1977).   Thus,  under most  conditions,  the
aerosol is well-mixed within the atmospheric surface layer  by  the time it  is
sampled at the midlake site (Gillette and Winchester, 1972).   The site is  also
well away from heavily travelled shipping lanes„

     Aboard the R/V Simons, meteorological sensors  were mounted at the tip of
a 6 m aluminum boom, which extended ahead of the  ship's bow at a mean height
of 5 m above the water (Figure 3).  The results of  bluff-body  turbulence
effects studies by Hunt (1973) and Hunt and  Mulhearn (1973) indicate that  a
boom length of only 3.5 to 4 m would have been sufficient to avoid any turbu-
lence effects due to the ship's presence,,  The undisturbed  ambient wind field
presented to the meteorological sensors is also evidenced by the intensity of
turbulence (IfO, which is calculated as

                                 It  =  a/U5               (5)

where a is the standard deviation about l^,  and "u"5  is the 15-minute mean of
1 Hz wind velocity values.  Overall It at midlake was observed to be
0.10 + 0.05, which is comparable to the 0.02 to 0.14 range  of  It observed  by
SethuRaman and Tichler (1977) from an air-sea interaction tower„  Thus, no
significant increase in It due to the location shipboard of the sensors could
be detected.

     Two Hi-Volume samplers were also located on  the boom,  approximately 305
and 4.5 m ahead of the bow.  The filter holders and heads were carefully
sealed to the pump motor assemblies, and the pump exhausts  were vented through
4 m long, 11 cm diameter hoses (see Figure 3K This was done  to prevent
possible contamination of aerosol samples by the  sampler pump  motors.  Sam-
pling was conducted with the ship bow-anchored on station;  therefore, the
boom pointed upwind and the likelihood of contamination by  the ship's effluent
was reduced (Moyers, Hoffman, and Duce, 1972). Whenever local WD at the ship
shifted quickly or Uc decreased to the point where  the upwind  orientation  of
the boom was uncertain, the Hi-Vols were stopped*  WD data  from the boom-tip
wind vane shows that the boom pointed to within ± 40° of directly upwind
during sampling periods.  The ASAS was located at the proximal end of the
boom, and aerosol number concentration was usually  measured at the 5 m mean
height above water.  When vertical ship motion was  minimal, attempts were  made
to measure vertical gradients in the aerosol number concentration.  The intake
of the ASAS was alternately placed at heights of  3.7 and 6.4 m during succes-
sive counting periods of from 3 to 15 minutes.  In  a certain  number of cases,
the measurement of number concentration differences between the two heights
met an expected logarithmic law criterion and was then used to estimate
aerosol deposition velocities by the profile method (Gillette, 1972).

     The Hi-Vol used with glass-fiber filters for total aerosol mass concen-
tration determination was run on the upper foredeck of the  R/V Simons.  The
                                     10

-------
integrating nephelometer and condensation nuclei  counter also were operated
upon the foredeck.  Observers, using a hand-held  infrared thermometer,
measured TQ by scanning an area of the lake surface directly below the  boom.
Bulk, and when feasible, surface water samples, were taken from a platform
on the upwind side or from a fiberglass launch (when feasible) well  away from
the ship.


3.3  NEARSHORE SAMPLING PROCEDURES

     At the City of Chicago 68th Street Crib, two different configurations of
the sampling equipment were used.  During the last two weeks of May 1978,
when the NCAR aircraft was also collecting data,  the first configuration was
used in collecting 9 data sets.  Later in the summer and fall, the second
configuration was used for an additional 15 data  sets.  Initially, the  meteoro-
logical sensors and Hi-Volume samplers were located on the upper steel-beam
bridge between the two masonry buildings which make up the crib site (Figure  4).
The samplers in that position were located about  13 m above the water surface
and 20 m from either building.  Because of the proximity of these buildings,
one of which is occupied by a crib maintenance crew, WD sampling windows were
defined.  Sampling was only conducted when WD ± 2a was within the directional
sectors 40° < WD < 130° or 210°  <  WD < 330° (see Figure 4).   Sampling  other
than within these windows increases the probability of local  contamination of
aerosol samples and of turbulent disturbance of the observed windfield.
Infrared thermometer T0 observations were made from the lower bridge at
about 5 m above the water.  The integrating nephelometer and vertical wind
sensor were operated on a small boom which extended 2 m towards the upwind
sector from the lower bridge.,  Water samples were taken from a fiberglass
boat, well away (0^75 to 1 km upwind) from the crib, again using the Kemmerer
bulk sampler and surface screen„

     During the June through November period, 15  data sets were collected at
the 68th Street crib site.  The Hi-Volume samplers were relocated to the lower
bridge (5 m above water) level, and the meteorological sensors were installed
on a portable mast.  Sampling was conducted at irregular times, usually one or
two data sets per day in bi-weekly intervals0  The movable mast for the mete-
orological sensors was set up on the windward side of the seawall around the
north (crew quarters) buildingu  Meteorological data could then be collected
at the 5 m height from all WD sectors except for 130  < WD < 210°; WD  windows
for the Hi-Vol samples remained the same as for the May 1978 sampling period.
The integrating nephelometer, condensation nuclei counter, and ASAS were also
made portable and operated at the upwind point of the seawall.  The ASAS
intake was fitted to a 6 m boom, and the vertical wind sensor was mounted at
the tip of this boom.  Measurements of the vertical wind field at distances
from 1 to 6 m upwind of the 1 m high seawall at heights from 0.5 to 4 m above
water were made to verify that the ASAS intake was sampling the ambient aerosol
distribution.  Aerosol number concentration vertical gradients were observed
with the ASAS during the June through November period.  Bulk and surface water
samples were taken for each set.
                                      11

-------
                                  SECTION  4

                      DATA,  ANALYSIS,  AND  BASIC  RESULTS
4.1   RESULTS OF CHEMICAL ANALYSIS

!„   Water Samples


      Bulk water, composited from  samples taken  at 3,  5,  and 7  meter depths,
 was collected once during each data set at the  midlake site.   Table 3 is a
 compilation of the ICAP-AES trace element analyses of these samples.   Nutri-
 ent analyses were not run for any water samples.   Geometric mean elemental
 concentrations for each sampling  period indicate  the  degree of temporal  vari-
 ability in trace metal  levels at  the midlake site. The  data do not provide
 adequate resolution to  infer any  seasonal  pattern; however, the trace element
 levels at that single site are shown to be far  from constant over the summer
 through fall season.   The "overall" mean includes all  of the 40 bulk samples
 collected during the entire May through September, 1977  period.   The overall
 mean concentrations for all elements appear to  fall well  within ranges sum-
 marized from the literature by Torrey (1975).   The water trace element data
 form the basis for assessing the  contribution of  the  lake as a source of
 atmospheric aerosol,  particularly of those with metallic content.

      Surface-water samples were taken concurrently with  the bulk samples dur-
 ing the September, 1977 sampling  period only.   For the September group of 16
 samples, concentrations in the 100 to 200 ytn depth surface sample (Garrett,
 1965) can be compared to the underlying bulk water concentrations.   A con-
 venient method of expressing this comparison is by a  surface-enrichment  value,
 Esb, given by
      -sb
surface
Cbulk
                         1.0
(6)
      Thus, the sign of Esb gives  the direction  of increased  concentration;
 Esb being positive when    Csurfaco > C^ik-   In  Table  3,  then,  it  is  note-
 able that all  the Esb values,  except that  for  B,  are  positive.   The magni-
 tude of Esb indicates the extent  of surface  enrichment.

      The Esb values listed in  Table 3,  however, probably do  not  indicate  the
 actual  magnitude of enrichments  in  the  surface  microlayer.   This is partly
 due to  the sampling technique, which collects water from a surface  layer
 about 102 to 103 times the depth  of the surface organic microlayer  (Andren,

                                    12

-------
et al., 1975).  Also, it is partly due to the fact that the surface samples
were collected during each data set, whether or not a surface organic micro-
layer was present (Elzerman, 1979).  However, the Esb results in Table 3
strongly suggest that the surface water is at least slightly enriched rela-
tive to the underlying bulk water.  Due to the small number of samples rep-
resented in these results, and the fact that the E^ values are not statis-
tically different from zero, any extension or application of the surface
water results must be considered tentative (Owen, et al., 1979).  Current
models  of gas transfer across the air/water interface (Brtko and Kabel,  1976)
do not  adequately apply to the case of aerosol particle transfer.  The ESb
data, however tentative, may prove to be of value for development and evalu-
ation of models of aerosol transfer across the interface.  Bulk water temper-
ature and pH were measured during each data set; the results are included at
the bottom of Table 3.  The pH measurements of the bulk samples are typical
of open-lake waters (Torrey, 1976).  These pH values may be viewed as an indi-
cation  of the buffering capacity of the lake, in light of the inputs of acid
aerosols, i.e., 864 and N03 loadings (see Section 5).  No measurement of
surface water pH was attempted in the 1977 or 1978 sampling.  It is, however,
conceivable that atmospheric inputs of acid aerosol may create a reduced pH
in the  surface microlayer.  This could in turn enhance the solubility of
other aerosol constituents at the time of deposition to the lake's surface.
Such a  reduced pH microenvironment at the interface could enhance the avail-
ability to the ecosystem of toxic aerosol constituents, and should be assessed
in ongoing research.

2.   Aerosol Samples

     Three-stage cascade impactor filter sets were exposed on Hi-Volume sam-
plers during each data set.  The results of ICAP-AES trace element and auto-
analyzer phosphorus, nitrite/nitrate, and sulfate analyses are listed in
Table 4.  The mean concentrations for each element show a wide range of vari-
ability over the summer through fall season.  This is primarily due to the
atmospheric thermal stability.  A detailed discussion regarding this vari-
hility  will appear in Section 5.  For the trace element results, a general
filter  blank correction from equation (4) was applied after the analysis of
all 48  data sets and eight filter blanks was complete.  The P, N02/N03,  and
S04 blank corrections were generated for each sampling period's group of fil-
ters as they were analyzed by USEPA-CRL.  The B values were, in general, small
compared to typical sample values for the majority of analyses (see Table 2).
The mean values of chemical parameters reported are given as geometric means.
Rahn (1976) has suggested that geometric means tend to better represent dis-
tributions of environmental parameters, as compared to simple arithmetic
means.   Figures 5, 6, 7, 8, 9, 10, 11, and 12 also strongly suggest this.
Each of these figures is a histogram plot of elemental concentration distri-
bution  for those data sets in which C > 0.  The entire range of C measured
for each element was divided into 60 increments; the number of data sets for
which C values within a particular increment were observed was then plotted.
For the elements Al, Ca, Fe, Mg, Mn, Pb, Ti, and Zn, shown in these figures,
the concentration distributions are seen to tend toward log-normal frequency

                                   13

-------
distributions, for which a geometric mean is appropriate.   There are
data in the literature with which to compare the midlake aerosol trace ele-
ment and nutrient mean concentration data.   Schmidt (1977)  presented trace
element concentration data for samples collected over southern Lake Michigan
which generally correspond with the ranges  of values reported here.  The
results of Table 4 represent a midlake aerosol  chemistry data base which has
been collected at a single site under a wide range of meteorological condi-
tions.

     Over the entire May through September, 1977 period, a  total of 48 aero-
sol samples were collected.  In the various data set aggregation schemes used
in this work, it has often been necessary to exclude some data sets from some
calculations.  For example, due to handling and analysis problems, such as
internal reference element (Yttrium) spiking inaccuracy in  the ICAP-AES pro-
cedure, or some obvious contamination, nine of the trace metal samples have
been eliminated from further consideration.  Similarly, seven samples for P,
N02/N03, and $04 are unusable.  Wherever mean concentrations or other mean
values are reported in this work, those results are based upon only those
data sets which appear to be consistent with the entire data base for that
parameter.

     A chemical characterization of atmospheric aerosol over southern mid-
Lake Michigan can be made using the C data  of Table 4.  The geometric mean
total aerosol mass concentration, Cm, observed at the midlake site was 32
ug - m~3; the range of values measured was  10 < Cm < 94 yg  - m~3.  Trace
element composition is thus expressed in Table 5 by three ratios of concen-
trations:

                    n =  ?- • 100                                 (7)
                         r
                   FC = C (d < 1.0 um)
                        C (d > 1.0 ym)

                        c/c
                   r-r _    ref (in aerosol )
                   Lh "   C/Cref (in soil)

     The first of these ratios, n, from (7), represents the percentage of
total aerosol mass contributed by the concentration of a particular element.
Gatz1 (1975a) composite model of Chicago/Northwest Indiana aerosol corres-
ponds closely to the n values observed at midlake.  This suggests that trace
element composition of the aerosol is not drastically altered during over-
lake transport (see Appendix C).  The second ratio, FC, from (8), gives an
indication of the degree to which the fine particulate (d < 1.0 ym) contrib-
utes to the total concentration of a particular element.  In this calculation,
fractional concentrations from the first cascade impactor stage (d > 1.0 ym)
and backup filter (d < 1.0 ym) are used (Figure 2).  The intermediate impac-
tor stage serves to improve resolution between the fine and coarse particu-

                                   14

-------
late fractions.  The intermediate stage concentration is included, however,
in all total concentration calculations.  The third ratio in Table 5, aero-
sol Enrichment Factor, EF, from (9), relates the concentration of an element
to the concentration which might be attributed to a reference natural source
material.  In this work, the composition of midwestern soil  described by
Bowen (1966) is used, with Al concentration defined as the soil source ref-
erence, Cref.  EF values close to unity for a given element indicate that the
element is present in aerosol in soil-derived proportions; large EF values
indicate other major sources for that element (Rahn, 1976).   From Table 5, it
appears that at least Cu, Zn, and especially Pb are greatly enriched in mid-
lake aerosol, i.e., come from other than midwestern soil sources.

     Disaggregation of the grand average EF values by filter set is facili-
tated by the use of scatter diagrams (Rahn, 1976).  Metal enrichment factors
(ordinate) are plotted versus Al concentration (abscissa) by filter set on a
log-log scale.  The scatter diagram shows how the enrichment factor of a
given element varies with the concentration of soil material in sampled aero-
sol.   Points at the high-Al end of the diagram are usually from nearshore,
polluted urban areas; points at the low-Al  end are from long over lake tra-
jectory areas (or are otherwise low in Al).  Scatter diagrams for Pb, Zn, Mn,
Fe, Ti, and Mg are shown in Figures 13, 14, 15, 16, 17, and 18 respectively.
The dashed straight lines on each scatter diagram are lines of constant con-
centration in the metal  plotted.

     Overall comparison of the scatter diagrams with those in a compilation
of over 100 data sources (Rahn, 1976) shows a higher enrichment factor trend
(except for Fe) for the Lake Michigan site  than many urban sites.  Disaggre-
gation by filter set also shows greater variability in the enrichment factors
for Fe, Mg, Ti, and Mn,  than most other field sites.  This suggests a vari-
able depletion mechanism not only for Al but also for these four additional
metals.

     The scatter diagram for Pb (Figure 13) falls within the range of data
reported (Rahn, 1976).   However, there are  two regimes of data points which
are somewhat unique.   The first is in the upper left where enrichment fac-
tors of 3 x 10-3 to 1  x 10^ are associated with Al concentrations less than
35 ng/m3.  These data points are generally  related to sampling of air with
north winds and long over water trajectories.  The second unique regime of
data points is that with enrichment factors less than 1  x 10^.  These data
are all related to northeast to east winds  with intermediate to long over
water trajectories.

     The Zn scatter diagram (Figure 14) shows no peculiarities.  It is quite
consistent with the reported literature (Rahn, 1976), although greater vari-
ability in the enrichment factor is present.  The Mn scatter diagram (Figure
15) presents a different picture.   The data not only convey a larger enrich-
ment factor than other data sources (Rahn,  1976) but there is a rather wide
scatter in their magnitude.  This may be due to a combination of variability
in source region concentrations and variable over water depletion in Mn.


                                  15

-------
The Fe scatter diagram (Figure 16) shows variability in enrichment factors
similar to that of Mn but without a large average enrichment factor.   The
same pattern is present in the Ti and Mg scatter diagrams (Figures 17 and 18)
although the average Mg enrichment factor is very likely high due to  lake
source contributions of this metal.  The Ti  scatter diagram has a strong
trend toward constant concentration.   It appears that contamination from
shipboard and laboratory handling probably contributes to the Ti midlake
concentration data.  It appears that shipboard and laboratory handling prob-
ably contributed to the Ti midlake concentration data.  The enrichment factor
scatter diagrams support the notion that differences in source region is not
the only major factor influencing the variability in enrichment factors.  The
unique over water trajectory of an air mass  must also be considered to explain
this variability.

     Results of correlation coefficient determinations are shown in Table 6.
Non-chemical parameters included in this Table are total aerosol mass, Cm, as
well as u^ and AT.  Also shown are the squared multiple correlation coeffi-
cients (underlined) which give the percentage linear variation that can be
explained for the specified variable by all  the others combined.  Clearly,
little of the Mo and Ti concentrations and wind speed variability are linearly
dependent upon the other variables.

     The individual element dependencies of Table 6 also suggest that Ti is
anomalous, since it has only moderate correlation with two or three other
metals.  Al, is strongly correlated with other soil derived metals: Ca, Mg,
Fe, and Mn.  Ca and Mg have similar correlations to that of Al.  Fe is
strongly correlated with all three of these metals but is also highly corre-
lated with Mn, Pb, and Zn.  In fact, Fe seems well correlated with more vari-
ables than any other metal.  Mn, Pb, and Zn are also strongly correlated with
several other metals.  Mn shows the highest correlation with mass.  Mo cor-
relates only moderately with Pb,

     The trace element concentrations in aerosol observed at the City of __
Chicago 68th Street Crib are summarized in Table 7a.  For most elements, C
is usually increased three-fold in the nearshore area.  However, a close scru-
tiny of the n and FC results suggests that the trace element composition of
nearshore aerosol (predominantly Chicago/Northwest Indiana source) is not
significantly altered (see Table 4) during transport to the midlake region.
That is to say, although the gross concentration of aerosol is reduced by
dispersion and deposition during overlake transport, the composition and
proportion of fine and coarse particulates do not appear to change.  This
is evidence supportive of the contention that vj overlake is not a strong
function of aerosol size.  Use of the nearshore data in the loading estimate
calculations of this work is precluded by the state of development of the
loading model.  The priority application of these data will lie in the devel-
opment of loading models which integrate nearshore and midlake data,  and in
the source area/source type resolution calculations, discussed in succeeding
sections of this report.


                                   16

-------
     A passive aqueous extraction procedure (see Appendix D) was used on two
9 cm by 9 cm sections of crib backup filters0  The filter sections rested
on the surface of deionized water for 24 hours,,  The extraction liquid (i0e.,
soluble aerosol fraction) along with the samples generated from the extracted
section (ic,e., insoluble aerosol fraction) and from the unextracted remaining
portion of the backup filter were analyzed for trace metal.  The ambient air
concentrations of 17 trace metals were calculated and compared for each of the
three fractions,,  The percent soluble and the sum of the soluble and insoluble
fractions (the total recovery from the extraction procedure) are listed in
Table 7b for five important trace metals,,  With the exception of copper, the
extraction procedure accounted for, on average, the trace metal concentration
determined from the unextracted portion,,  The mean value of the "percent
soluble" column indicates the range of solubility of the trace metal.  Iron
and lead were found to be very soluble, 80-85 percent, whereas copper, manga-
nese, and zinc were found to be only partially soluble, approximately 50 percentc

     Filter sets 70740 through 70810 present anomalous results.  The analysis
of the extracted portion of these filters indicated a lack of trace metal„
The results for the five metals in Table 7b are typical of the rest of the
elements analyzed.  All or a large percentage of the trace metal recovered in
the extraction procedure was found in the soluble portion.  The absence of
material in the insoluble fraction could not be traced to a procedural or
operator error*  The source region of the aerosol appears as a first explanation
of the anomalous results„  The wind directions for these filter sets were west
to southwest — from the Chicago urban area.  The filter sets with measurable,
insoluble, small aerosol had wind directions from the east-northeast.  Secondary
particle generation by acid gas dissolution into and reaction within liquid
aerosol  could possibly account for the preponderance of soluble aerosol found
near the Chicago urban source.

4.2  PHYSICAL CHARACTERISTICS OF AEROSOLS

     The chemical  composition data were not presented in the context of mass
concentration by data set since the cellulose filter medium is quite hygro-
scopic and, therefore, not conducive to gravimetric analysis.  A standard Hi-
Volume sampler with type A glass fiber filter was run for 8 to 24 hours to
collect sufficient aerosol on the glass fiber filter.  Eighteen mass concentra-
tion samples were obtained this way.  In addition, integrating nephelometer
scattering coefficient da_ta (b§cat) was linearly regressed against the Hi-Vol
mass concentration data (C ) with the result:

                        bscat  =  0.298+ 0.0138 Cm                  (10)

An extremely good correlation of 0.97 was found.  The importance of the correla-
tion is  that bsc^ values can be averaged across each cellulose filter sampling
period to accurately reflect the associated aerosol  mass concentration by
using equation (10).  The 49 mass concentrations determined in this way have
a normal  distribution mean of 34 yg/m^, but are a better fit to the log-normal
distribution with a geometric mean of 31  ± 19 yg/nA   The mean of 31  is clearly

                                   17

-------
below typical urban and even rural concentrations but above the 15 to 25 yg/m
found at remote, land-based sites (Rahn, 1976) and 10 to 15 yg/m3 found at
remote oceanic sites (Windom and Duce, 1976)0

     The size distribution of this midlake mass concentration was monitored
by the ASAS -- at least that mass fraction in the Ool < d < 3<,5 ym aerosol
size range.  The ASAS directly monitors the number of aerosol in each of 60
size fractions.  Table 8 shows the 50 percent collection efficiency diameters
for the ASAS with a comparison to those for the Hi-Volume sampler and integrat-
ing nephelometer.  The ASAS counts aerosol over the entire range of sizes
observed by the nephelometer, but only a portion of the sizes collected on the
Hi-Volume sampler,,  The aerosol particles which are detected are electronically
sized and the number of particles falling within each of the 60 size ranges
is accumulated.  Aerosol distributions are normally displayed on plots having
d or log d as the abscissa, and AN/A (log r) as the ordinate, where AN is the
number of particles counted in a size range and A (log r) is the difference
between the logs of the upper and lower limits of the range.

     Figures 19, 20, 21, and 22 display this plotting of the average number
concentration for each of the four midlake sampling periods:  17-20 May 77;
7-9 June 77; 14-19 August 77; and 26-30 September 77.  The average slope of
the data points is -2.3, -3.4, -3.3, and -3.8 for May, June, August, and
September, respectively^,  September data are representative of unstable air
over the lake.  As a result, large aerosol are indeed depleted at the midlake
sampling point0  In fact, the -3.8 slope is rarely encountered in surface layer
air unless the air resides several hundred kilometers from pollution or natural
primary aerosol sources.  The August data is representative of a neutral surface
layer.  The slope of -3.3 suggests less depletion of large aerosol than in
September.  During the June sampling period, though slightly stable air pre-
vailed, heavy rains washed out a significant number of aerosol  in all size
ranges.  Possibly due to the greater washout efficiency of primary aerosol,
the slope of -3A is slightly more than that in August.  During the May
sampling, a very stable surface layer prevailed.  The stable layer clearly
"held" more large aerosol during the May sampling period^  Overall, the
slope of -2.3 indicates that the d > 1 ym aerosol are enhanced when compared
to urban and even many rural concentrations of these primary aerosols0

     Similar distributions of particle volume can be obtained by multiplying
the number counted in each size bin by 4/3itr3. The ordinate for those plots,
then, is AV/A (log r)«   These plots are shown in figures 23, 24, 25, and 26.
Each figure also shows  "typical" urban and continental background volume
distributions for comparative purposes.  The May period data (Figure 23) show
aerosol volume distributed approximately as found in urban environments
except for the relatively large midlake contribution by the r = 0.5 to 100 ym
aerosol sizesu  This may be attributed to a relatively poor depositional loss
of this size fraction.   The close proximity of the remainder of midlake aerosol
volume to that of the typical urban air mass may be caused by the slow rate of
depositional loss during transport from the predominantly Chicago source of
May period aerosol „  During June sampling, the heavy rains just prior to the

                                    18

-------
period are the major cause for the aerosol volumes encountered (Figure 24).
The August and September periods both show a pronounced volume peak in the
r = 0.1 to 0.2 ym size range (Figures 25 and 26).  Since the r = 0.1  ym
aerosol number concentration was ~3 x 10'°-nT3, an order of magnitude
greater than it was during May, coagulative transfer of aerosols from the
r  <0.1 ym size to the r > 0.1 ym size could occur within a 4-hour period as
against a 25-hour period in May.  The coagulative aerosol transfer from any
one smaller size range to the next larger size range in a unit time increment
is proportional to the square of the number of aerosols present in the smaller
size range.  To illustrate, the time in which an initial concentration of
3 x 10'°-m~3 aerosols falls to one-half that value due to coagulative loss
is less than 4 hours, whereas a concentration of 4 x 109-nr3 takes more than
25 hours to reach one-half its initial value.  If an aerosol population is
sampled after the r  <  0.1 ym aerosol concentration falls to half its initial
value, it may be referred to as an "aged" aerosol population.  The aerosol
populations sampled in August and September can then be described as  par-
tially aged, and some enhancement of certain trace metal concentrations
sampled on a Hi-Vol filter may have occurred.  Gillette (1972b) found a
10 percent or more enhancement of Pb concentrations in aerosol populations
which underwent aging during transit over Lake Michigan.  The August  volume
distribution carries more volume overall than September, with substantially
more in the 0.1 < r  <1.0 ym size range.  Wind speeds were not too dissimilar:
August, uij = 4.9 ± 1.3 m-s"1; September, ~u$ = 509 ± 1.6 m-s~'u  In both periods,
air downwind from the Chicago/Gary source area was sampled for less than 10  per-
cent of the total sampling timeu  Neutral to slightly unstable temperature
conditions prevailed more often in September.  A data set by data set analysis
showed that a thermally stable surface layer prevailed in August for  the tine
when Chicago/Gary source aerosols were traversing to the midlake sampling site.
Order of magnitude aerosol volume in the 0.1 <  r < 1.0 ym size was present for
only two data sets thereafter, but this was sufficient to cause the substantially
greater volume in August sampling compared to September.  This last point indi-
cates volume distributions averaged over data sets are valuable analysis tools.

     Several  sequences  of data set average ASAS volume distributions  were
considered.  The most significant outcome to date from this sequencing analysis
is the often controlling influence which wind direction shifts seem to have
upon midlake mass concentration,,  That is, wind direction shifts have a strong
influence upon aerosol  deposition to the lake in addition to bringing differing
source region aerosols  to bear on the total  mass flux of aerosol.

     Figures 27-33 show the sequence including  the data sets numbered 20050-
20110 (elsewhere referred to as sets 5-11) and  covering the period 17 May 77,
22:15 CDT to 19 May 77, 06:15 CDT0   The wind speed was relatively constant on
the evening of the 17th, but a wind direction shift of 60° to 70° occurred
between 18:00 and 21:00 CDT.  An extremely stable surface layer prevailed
throughout the entire sequence of 7 data sets with the difference between air
and surface temperatures (AT)  being a minimum of 5«9°C, and a maximum of 13.3°o
                                    19

-------
Data set averaged wind speeds were as low as 1.7 m-s~^  and as high as 406 m-s"1.
Table 9 shows the data set start and stop times, the meteorological  data
averaged across each set's sampling time, and the aerosol  parameters also
averaged across the set sampling times.

     The value for Cm averaged across the 7 sets, as determined by integrating
nephelometry (equation 10) is 129 ± 54 yg-m~3, whereas  the standard Hi-Volume
sampler, run across all 7 data sets, gave a value of 132 yg-m~3a   The nephelo-
meter primarily responds to the d < 1 ym aerosol and should give a value some-
what below the Hi-Volume sampler.  The bscat values suggest variations in
aerosol mass concentrations from set to set.  This is substantiated by the ASAS
volume distributions.  The area under each volume distribution curve in
Figures 27-33 has been producted by a size-independent  aerosol  density of
2 g-cm~3 to give the values in the last column of Table 9.  The mean of these
equivalent mass concentrations is 120 yg/m3 with a large standard deviation of
113 yg-m~3o  The second to last column gives the equivalent mass  concentration
measured by the ASAS in d <  1 ym aerosol sizes.  The mean of 84 ± 64 yfj-nr3
shows somewhat less fluctuation from set to set than does  the total  ASAS mass,
but the extreme values of 19 and 174 yg-m~3 show that very wide fluctuations
in midlake fine particulate concentration are possible.  A consideration of the
meteorology related to each set may help explain why such  Targe variability
is possible.

       The  first set  in this  sequence,  20050,  had a moderate 51 yg-m~3 ASAS
mass concentration with nearly all  of  that mass  in the d  < 1 ym size  range.
Strong thermal stability, moderate  wind  speeds  and winds  from the Chicago
source area with each meteorological variable  rather steady all lead  one to
expect the  concentration  to  be substantially  higher.   However, a wind  direc-
tion shift  of 60° to  70°  had  occurred  only two  hours prior to sampling.  Since
the Chicago source aerosol would be expected  to  reach midlake during  the sub-
sequent data set (20060), it  is  not surprising  that the mass concentration
determined  from ASAS  data more than tripled at  that time.  The smaller AT
due to cooler nighttime air  still maintains a  very stable surface layer while
wind speed  and direction  do  not  change  at all.   During set 20070, wind speeds
drop off a  bit, AT increases  again, but  an unchanged wind direction allows
midlake air to become thoroughly populated by  Chicago  source area aerosol.
During set  20080, wind speed  continues  to decrease to  a value 1.9 m-s'1,
whereas the thermal  stability continues  to increase with  daytime heating.
However, a  60° wind  direction shift to  southerly flow  probably accounts for
the six-fold fall~off in  the ASAS mass  concentration.  The southern shore
area is a weaker aerosol  source  region  than the  Chicago area, but far  from
six-fold weaker.  Since stronger thermal stability and lower wind speeds
prevail, aerosol deposition  to the  lake  should  be reduced during the  20080
set when compared to  the  20070 set.  However,  the wind direction shift before
and the relatively large  a of 46° during set  20080 sampling appear to  be
strongly increasing  depositional loss.   Note,too, that a on wind speed was 30% of
                                    20

-------
D5 during 20080 sampling, but only 8 percent of D5 during 20070 sampling.
Another 35° wind shift with a 40° deviation during the 20090 set, along with
a doubling of wind speed contributes to the maintenance of the ASAS-derived
mass at about 30 yg-m~3.  By set 20100 the Chicago/Gary source is again
upwind but too little time has elapsed for Chicago-derived aerosol to reach
midlake.  If anything, the wind shift, along with the 25 percent a on wind
speed, appears to have caused still more depositional loss to the lake, for
the lowest ASAS and bscat mass concentration for this seven set sequence was
observed during set 20100.  By the 20110 set the 212° wind direction has
brought aerosol from the Gary area to the midlake site.  The source types in
Gary produce a substantial amount of coarse particulate mass (d > 1  ym).
Combined with a shorter fetch (40 km, as compared to 55 km for the Chicago
source) this resulted in a dramatic increase of total and especially of
d > 1  pm aerosol mass.

       This sequence of seven May sets indicates that wind direction change
is an important Great Lakes loading factor in its own right, i.e., direction
shifts can result in strong depositional loss of aerosols to the lake inde-
pendent of the differing aerosol source regions presented to the lake by
these direction changes.  Donelan (1977) contends that the size and state
of development of wind generated waves are important and occasionally con-
trolling factors for momentum transfer at the air/water interface of the
Great Lakes.  In particular, the ratio of wind speed to the speed of the wave
edges and the angle between their respective directions are critical para-
meters.  When the waves are "young" they will not be in equilibrium with the
wind field.  As the waves "age" they approach an equilibrium with the wind
field, the difference in speed reaches a minimum, and the angle between the
wind direction and wave propogation direction usually approaches a minimum.
Donelan (1977) has measured momentum drag coefficients (CD) 5-to 10-times
those observed under steady state conditions (see Figure 34).  It is quite
reasonable to expect aerosol transfer to increase several fold, if not by an
order of magnitude during these same conditions.  In fact, the  large loss  of
aerosol mass between  set  20070 and  20080 and again between sets 20090 and
20100 seems  to exemplify the large mass  transfer  increase associated with
young waves.  The fact that the ASAS mass was at a minimum during set 20080
despite Q5  of only 1.9 m-s'1  is another indicator of substantial mass trans-
fer due to  wind diviation shifts and associated young waves.   Since the major-
ity of  filter/data sets on  Lake Michigan were collected  during  steady state
conditions  and not soon after  large wind direction shifts, the  majority of
sampling occurred with aged waves  present.   As a result, the aerosol con-
centration  measured  may be  high while  the deposition velocities used_do not
represent young wave conditions.   The  resultant  loadings found  in this  study
do not  reflect these non-steady state  conditions.  Though  the  consideration
of non-steady  state  conditions would  likely  increase lake  loadings, this is
not obvious or even  certain.

        One  sampling  day during which  non-steady  conditions were followed by
unusually  high loadings was 30  September 77.  Such occurrences  are  best inves-
tigated  by  a case-by-case approach, integrating  all  available  data  sources.


                                   2.1

-------
On 29 Sept 77, three weakly coupled low pressure  areas  were  located  in  the
Texas panhandle, over Lake Superior, and over New England.   The Lake Superior
low pressure system moved Southwest and eventually formed an elongated  trouah
By 2 October 77, there was a cold front stretching from Louisiana to Maine."
While it was in its early stages of formation,  the front passed through the
sampling site.  Pre-frontal rain was observed at  the ship the night  of  the
29th, and moderate to heavy rain followed the front, though  no further  precipi-
tation was observed at the sampling site.  Prior  to frontal  passage, there
were broken clouds at 4000 m, while after the passage,  the  broken clouds
were at 3000 m and the sky was generally overcast.  The relative humidity on
30 September 77 was 70 to 90 percent.  Winds at 300-600 m (1000-2000 ft.)
were from the South to South-Southwest ahead of the front,  at 1900 CDT  and
by 0700 CDT were from the West to West-Southwest,,  To analyze the effect of
these occurrences, it is necessary to estimate the time of  frontal passage
through the sampling site.  There was a 40° windshift at the ship between
0500 CDT and 0530 CDT.  The pressure reached a local minimum at the  same time.
The pressure minimum was reached at Midway Airport between  0300 and  0500 CDT0
However, wind shifts were observed at Midway, the 68th Street water  intake
crib, Argonne National Laboratory  and Governors  State University in the
period 0900-1300 CDT0  Because of reduced surface roughness, it is not
unusual for an over-water system to be retarded upon land fall.  This con-
jecture is borne out by consideration of the subjectively drawn streamlines
for this period which show that at 0700 Milwaukee, WI, Muskegon, MI, and
the midlake Sampling Site had undergone windshifts while Chicago, Michigan
City, MI, and Benton Harbor, MI were still reporting southerly flow.  The
front had apparently formed a bulge over the laJ
-------
upon consideration of the back trajectories  calculated  for  this  period,,   Since
the surface wind speed was near 4 m-s-1  during  the  sampling period,  the  air
samples left shore 3-5 hours earlier.   During this  period,  the wind  shifted
in such a manner that the volume, which was  eventually  sampled,  swept  through
the heavily industrialized area along  the south shore0   It  is these  types  of
occurrences which make the gathering of data which  is representative of  a
large area very difficult,,

4.3  METEOROLOGICAL AND CLIMATOLOGICAL CONSIDERATIONS

       The  bulk deposition  velocity was  shown  in equation  3 to be determined
by  the product of wind speed  and the  diabatic  drag coefficient.   The mean
wind speed  at the nominal sampling  height of 5 meters was  quite steady over
a filter  sampling set.   Values ranged from  a low of 1.7 +  0.6 m-s"1 to a max-
imum of 8.3 i 0.6 m-s-1.  These  values are  not entirely representative of
conditions  over Lake  Michigan  since the  local  climatology  indicates that wind
speeds higher than 8.5 m-s'1  occur  one-fourth  of the year  (see Figure 35).
However,  the observed range of wind speeds  results in an underestimate of
annual average loading,  provided, of  course, that the low  wind speed regime
is  carefully considered.  It  is  well  known  that the condition of smooth flow
over a water surface  is  established at wind  speeds below 2 m-s"1 and possibly
below  3 m-s-1.  The  smooth  flow  condition is one in which  laminar (non-
turbulent)  flow prevails just above the  air/water interface.  When  this lam-
inar sublayer  is  spatially  continuous, the  transfer of aerosols to  the water
is  impeded.  With higher wind speeds,  the roughness elements at the surface
begin  to  protrude beyond the  sublayer.   Turbulent transport pathways may  then
strongly  enhance  aerosol  transfer across the interface.  Data of Kondo, et a1.
(1973) show that  high frequency  wave  components cause 30 percent of the mea-
sured  surface  roughness  elements to extend  outside the laminar sublayer at
wind speeds of 2 m-s"1.   Although these protrusions into the turbulent^zone
may give  rise  to  substantial  aerosol  transfer, wind speeds of 2.0 m-s l or
less are  assumed  here to allow a continuous laminar sublayer to be  estab-
lished.   This  essentially blocks all  aerosol transfer to the water  surface,
i.e.,  vd  =  0.   It  is  felt that this may  again  be an assumption which  leads to
underestimates of loadings.  Further,  it obviates the determination of C0 ,
since  C0  for almost  any  constituent will  be a  significant  fraction  of C5  only
when the  continuous  laminar sublayer  is  present.

        For wind  speeds  in the 2.0 to  8.5 m-s'1  range, a neutral drag  coeffic-
ient is  graphically  determined from Figure  34.  The drag coefficients for
neutral  momentum  (CD),  heat (CH) and  water  vapor  (CE) are  presented for  wind
speed  measurement at the 5  m height.   An analogy  between water  vapor  transfer
and mass  transfer is sometimes suggested to support the use of  CE.  However,
the momentum drag coefficient was used because of a possibly still  better  anal-
ogy between momentum and aerosol transfer to be described  later  in  this  sec-
tion.   The results  of several experiments (Sheppard, 1963; Hicks,  1972;  and
Kondo, 1975) were used  to identify  the CD curve of Figure  34.   Note the  weak
dependence of  CQ on  wind speed.   Indeed, some  micrometeorologists  claim  CD
to be  independent of wind speed  (Wu,  1969).

                                     23

-------
       Having obtained a value for CD, a correction for thermal  stability
can be incorporated by determining the diabatic drag coefficient,  CDD.   This
coefficient is usually determined experimentally by monitoring both  the wind
speed and temperature profiles in the lowest ten meters above the  surface of
interest.  This regime is referred to as the surface layer.   A Monin-Obukhov
length is then calculated (Montieth, 1973).   In the case of sampling on board
an anchored ship, the temperature profile can be measured with reasonable
accuracy, but the wind speed profile cannot  be monitored (Donelan, 1977).  A
simplification of the Monin-Obukhov length determination must then be used if
    is still to be calculated.
       The surface layer can be considered as a constant flux layer (here, the
aerosol mass flux).  Hicks (1972) and Kraft (1977)  have shown the constant
flux assumption to be valid to at least one meter of height for every several
hundred meters of constant fetch.  Fetch is here defined as that distance
across which an invariant surface condition is presented to the wind field.
As long as no sharp wind direction changes occur within five kilometers of
the ship sampling site for the duration of a data set,  a sufficient fetch and
thus constant flux surface layer prevails.  The determination of sufficient
fetch is simply based on the absence of fronts within 5 km upwind of the ship.
If this condition was not met, the data set was not considered for loading
estimates.

       In essence, then, the use of the drag coefficient method as stated here
assumes a steady state condition during sampling.  This assumption results in
an underestimate of the annual average bulk deposition  velocity,  v^o   Donelan
(1977) has shown that the drag coefficient may at times be two to five times
the neutral steady state CD-  This may be primarily due to wind direction
shifts which then cause the wind flow lines to cut  across the wave edges.  Of
course, a rapid 180° wind direction change gives the most dramatic increase  in
drag coefficient—at least for momentum transfer.

       Whether such large deposition velocities are appropriate for aerosol
mass transfer when compared to momentum transfer as determined by CQQ directly
is uncertain.  This point will be further considered later in this section.
Under the condition of adequate fetch, then, an analysis suggested by Kondo
(1975) can be performed, as in Appendix B, to give  the  ratio of the diabatic
drag coefficient to the neutral drag coefficient, CQQ/CQ.  This ratio is
expressed as a function of a stability parameter SQ, which for h = 5 is given
by

                                                                    (ii)
            1.3u|
     The dependence of CDD/CD on temperature stability is developed more fully
 in Appendix B.  For thermally stable conditions, i .e. , AT = T,- - T  > 0
                                                             0    0

                 CDD/ CD * 0.1 + 0.03 SQ + 0.9 exp (4U8/SQ)          (12)


 and for thermally unstable conditions, i.e., AT = Tr - T   <  0
                                                   3    0
                    CDD/CD   , KO + 0.47^'                         (13)


                                   24

-------
     These equations are plotted in Figure 35 as a function of S0.   The
reduction in CDp/CD with increasing thermal stability is quite evident
For U5 = 5 m-s-i CQQ/CD equals 0.22 when AT = 8°C0  These conditions prevail
over southern Lake Michigan 12 percent and 3 percent of the average year,
respectively.  Unstable air prevails more often.  AT of -4°C and -8°C occur on
the average 30 percent and 13 percent, respectively„  Yet the increased  CDp/Cp
does not compare to the reduced CQQ/CQ under stable air conditions,,  This  ratio
is 1.07 and 1.15 for AT of -4°C and -80Q, respectively. Thus, for conditions over
southern Lake Michigan, thermal stability may cause a 5-fold or greater  varia-
tion in the deposition velocity, but the frequency of occurrence of this wide
variability is not very high.  Figure 37 is a cumulative frequency of occur-
rence plot  for AT between -16°C and +16°C over southern Lake Michigan.
These climatological data were derived from the National Oceanic and Atmos-
pheric Administration (NOAA) Summary of Synoptic Meteorological Observations
for the Great Lakes Areas, Vol. 3: Lake Michigan (1975)!!These data were
obtained from edited observations reported in the period 1961-1973 by ships
in passage.  It should be noted that since such ships attempt to avoid bad
weather, the data are thereby biased.

     Having estimated the deposition velocity as a function of micrometeorology,
it is important to consider synoptic and mesometeorological effects (especially
on V(j) in conjunction with micrometeorology.  An additional factor in this
discussion is the fact that source region and source type may contribute
substantially to aerosol concentration variability, even at a midlake sampling
point.  Trajectory analyses -- especially back trajectory analyses -- based on
meso- and synoptic meteorology    can be used to establish source regions  for
each data set.

     A simple back trajectory calculation followed aerosol sampled at the
ship back toward shore horizontally and upward through first an assumed 50 m
deep surface layer and then into the mixed layeru  Upon reaching shore the
trajectory was stopped and the time to reach shore was obtained to within
± 20 percent.  The back trajectory procedure during traverse within the surface
layer used the at-ship horizontal wind speed and direction averaged across
the time of sampling for each data set and a vertical velocity estimated by vju
Although not a real velocity, VH represents the time-averaged vertical motion
of aerosols within the surface layer, so long as it is truly a constant flux
layeru  Within the mixed layer, vertical velocities are estimated by the mass
continuity equation, whereas horizontal velocities are estimated by a weighted
average of the sampling ship's horizontal velocity and a triangulation
average of National Weather Service upper level sounding data.  The horizon-
tal spread in the completed trajectory was assumed to be determined by trajec-
tories using two standard deviations about the mean wind direction.  The
vertical spread was determined through use of stability category dispersion
coefficients determined as a function of the bulk Richardson number (Rig)
(Nagib, 1978).

     Although this simple trajectory procedure may be criticized on several
counts -- especially regarding its accuracy — the results are used simply to
identify five horizontal source regions,.   In addition, the predominance of

                                    25

-------
nearshore mesoscale and long range macroscale sources can usually be identi-
fied.  The five source regions are shown in Figure 1  and are defined as:

       1)  An overlake source with a long fetch across the northern
           and upper part of southern basin of Lake Michigan,
           referred to as L.  It encompasses the ship to trajectory-
           at-shore angles 325° to 30°.

       2)  A west shore source of moderate to light industrial and
           residential development (nearshore) and largely forested
           land (longer range).  Encompassing the ship to trajectory-
           at-shore angles 270° to 325°, it will be referred to as WS.

       3)  A southwest shore source of heavy industry and dense popu-
           lation.  This source takes in Gary, Indiana with its heavy
           steel industry by beginning at 190° and the entire Chicago
           source except its northern suburbs by ending at 270°.  Near-
           shore contributions from the source region will in most instan-
           ces mask any long range contributions.  It will be referred  to
           as C/G for the Chicago-Gary source.

       4)  A southeast shore source of moderate industrial and residen-
           tial development (nearshore) and agriculture (longer range).
           Encompassing the ship to trajectory-at-shore angles 90° to
           190°, it will be referred to as SES.

       5)  An east shore source encompassing the ship to trajectory-at-
           shore angles 30° to 90Q.  This source is largely agricultural
           (nearshore) and industrial (longer range).  It will be referred
           to as ES.

As to nearshore mesoscale vs.  long range macroscale source identification,
whenever the mean trajectory remains within the 50 m deep surface layer and
the dispersion maximum trajectory within a 200 m deep layer all the way to
shore, the nearshore source contributions are assumed to be predominant.
Whenever the mean trajectory projects above a 100 m height before reaching
shore, both nearshore and long range sources are assumed to contribute  to
midlake aerosol concentrations.

       Some of the data sets cannot be analyzed by this trajectory approach
because synoptic scale fronts were present on southern Lake Michigan or
because the variation in wind direction across the data set was too great to
identify any one source region.

       Certain days during which interesting or anomalous conditions may have
prevailed were subjected to a case study analysis.  In the case of data set
50560, this has resulted in a strong suspicion that data taken were not truly
representative of the indicated source region.  These data were excised from
source region consideration.

                                   26

-------
     The possibility that coagulation of d < 0.1 jam aerosols may be an
indirect source of sampled aerosol is addressed by the tine the aerosol
takes to reach shore.  Gillette and Uinchester  (1972) found that coagulative
aging of 6 hours or more does contribute to Hi-Vol filtered Pb concentration;
however, for less than 4-hour aging no coagulative source for Pb is present,,
Given the fact that only five data sets had times-to-reach-shore of more than
6 hours, it is not very likely that a coagulative source contributes to
sampling except for a few sets from the overlake source sector.

_    The_determination of a Richardson number (Rig) can be calculated from
u5 and AT, measured ahead of the ship's bowu   The smaller this number, the
more turbulent is the surface layer air0  Several of the May and one each
of the June and September data sets have large Rig.  These sets were obtained,
then, during conditions of reduced turbulence.  One indication of synoptic
effects is obtained by identification of the air mass type in which the
sampling took place.  Rossby diagrams were used to identify air mass types,,
Only about half of the sampling periods could be clearly associated with an
air mass type.  Continental polar (cP) air masses prevailed during 15 of the
sampling periods, and maritime tropical (nT) air masses prevailed during 7 of
the sampling periods,,

       Though the processes  that  directly  determine  the  interfacial  aerosol
removal rate are microscale  in nature,  these  processes are  to  some  degree
energetically parasitic on larger scales of motion.  A complete  understanding
of microscale events requires some knowledge  of the  effect  of  measured  synop-
tic  scale phenomena.  Because of  the  complex  and  non-linear nature  of the
interaction of  synoptic scale events  with  the meso and micro-scales  and with
the  climatology, a  quantifiable relationship  has  not been sought.   Neverthe-
less, a general consideration of  the  synoptic situation  during sampling peri-
ods  is necessary for several reasons.

        It is  not to be expected that events of a single year will parallel a
 regional  climatology exactly,  but significant anomalous events may indicate
 certain periods or types  of data  for which scepticism is appropriate when
 climatological  extrapolations  are made.  Precipitation  events  were  excluded
 since this  study focused  on dry deposition removal of atmospheric constitu-
 ents to the lake.   Frontal effects were also avoided as much as possible so
 that a relative constancy in meteorological  conditions prevailed during each
 filter sampling data set.   From a climatological viewpoint this does intro-
 duce anomalies since the  greater  mass transfer to the lake associated with
 frontal  turbulence is not considered.  Conservative estimates of atmospheric
 loading should result.

        In the absence of frontal  or precipitation events, the process most
 directly influenced by the synoptic scale which is of interest in this study
 is aerosol  transport.   The advective history of an air parcel  will  largely
 determine the character of the background aerosol it contains.  The inter-
 action of synoptic and mesoscale  influences determines the short-to-mid-
 range vertical  and horizontal  transport of aerosol.


                                    27

-------
       The direction of the near-surface winds reflected in the backward-in-
time trajectory manual calculation revealed significant differences when
aggregated by outing.  If a mean is taken of all  May, 1977 sampling-period-
averaged wind directions, a value of 202°, with a a of 40° is obtained.   The
climatology shows the two major wind directions in May are northerly or south-
erly.  Thus, this outing is not particularly anomalous, but is lacking in rep-
resentation of a major (and presumably "cleaner") source region characteristic
of this period.  The 300-600 m (1000-2000 ft.) winds were observed to be gen-
erally from the southwest during this period, as  revealed by streamlines based
on NWS data.  A comparison of NWS upper air soundings taken at Midway Airport
with those of standard air-mass types indicates that May, 1977 air in this
region was most similar to maritime tropical (mT) air masses., consistent
with the conclusion that this outing sampled air  which arrived from the south.

       A synoptic-scale vertical velocity was crudely calculated from one-
dimensional divergences based on several NWS surface observations.  The magni-
tude of the vertical velocities is not considered reliable, but these calcula-
tions do provide some indications of whether the  synoptic situation tended to
support rising motion or subsidence during the outings.  The May, 1977 outing
was characterized by a tendency toward subsidence, probably related to the
not uncharacteristic residence of a high pressure area over the Great Lakes.

       The June, 1977 outing mean wind direction  was 203° (o = 24°), which is
quite consistent with the southern Lake Michigan  climatology.  However, the
300-600 m streamlines indicate a North-Northwest source for this level.   This
is substantiated by the persistent presence of continental polar (cP) air mass
types. This air mass type typically has experienced a trajectory over rela-
tively non-industrualized land.  One would expect the "background" values of
aerosol for this outing to be somewhat depleted compared to, say, the May out-
ing.

       The divergence-derived vertical velocities for this outing were posi-
tive  (upward) on the average, indicative of the influence of two low pressure
system passages during this time.  As a general expectation, this should effec-
tively increase the mixing depth and lower surface concentrations.  The occur-
rence of substantial precipitation just prior to the outing resulted in the
lowest concentrations experienced throughout the entire field program.

       The back-trajectory source regions for the August, 1977 outing seems to
be divided on the morning of August 16 by the passage of a low pressure sys-
tem and associated developing front.  Before this point, the winds were vari-
able, but generally from the East (91°, a = 44°).  After frontal passage, the
winds were from the northwest  (327°, a = 20°).  The 300-600 m streamlines were
from  the northwest throughout the period, except during the frontal passage.
As might be expected, the air mass type was cP after the evening of the 16th
and mT before that.  These occurrences are sharply in contrast with the clima-
tology for the month, which indicates a strong tendency for south-southwest
winds with a rather sharp "roll-off" of probability outside of these wind
directions.  Because of the strong and proximate aerosol sources that an air

                                    23

-------
mass with a traverse from the South would encounter, one would expect that the
air sampled during August 1977 was atypically "clean."  The calculated vertical
velocities before the frontal passage were ambiguous, but afterwards the indi-
cations were for general subsidence0

     The September outing is also seemingly divided into two regimes.  Weather
of the period before the evening of September 28 is dominated by a low pressure
area which was to the East of Lake Michigan at the beginning of the outing and
moved eastward.  The combination of this system and a high pressure area to the
West induced generally northerly winds (308°, a = 19°) during this time.  On
the evening of the 28th, the low had moved far enough off shore to allow the
high to establish itself southeast of Lake Michigan.  This induced southerly
winds for the remainder of the outing (181°, o = 37°)„  The 300-600 m winds
were westerly to northwesterly for the first portion of the outing and southerly
thereafter,,  The air mass type was cP throughout.  The climatology shows a
general tendency for southerly winds in September, so that once again the
outing-averaged concentrations for September might be looked upon as conserva-
tive estimates, with respect to the climatology.  The calculated vertical
velocity was quite variable, largely due to the synoptic disturbance, but the
overall tendency was for subsidence^

    A common mesoscale event is the lake breeze.  Because of the large perturba-
tion a lake breeze event has on transport, it is fruitful to attempt to assess
the frequency of lake breezes during the sample periods.  A major complicating
factor is the uncertainty in whether to call certain events lake breezes.  Past
concern with lake breezes has stemmed from their effect on populated shorelines,
so that occurrence criteria normally include a substantial degree of penetration
inland.  Objective indices are usually calibrated to those types of observations
which are not necessarily of prime interest to this study, but still should pro-
vide at least a general guideline.  Lyons and Olsson (1973) have documented the
southern Lake Michigan lake breeze climatology.  Lake breezes are observed
overland approximately 35 percent of all days during May through August, and
occasionally in March, April and September.  Of the 15 days on which midlake
samples were taken, one would expect sliqhtlv over five observations of lake
breezes.   In fact, only one day, 19 flay 77, developed a lake breeze strong
enough to be observed at Midway Airport (approximately 12 km inland).  The
68th Street Crib observations show two additional days, 7 June 77 and 18 Aug 77
which had significant lake breezes.

    A recent index was discussed by Lyons (1972) which is based on water-land
temperature differences and pressure gradients.  He reported a hindcast accuracy
of 90-95 percent, if cloudiness and gradient flow influences were considered.
An index which is perhaps more easily calculated from the data routinely
gathered during the field program was developed by Hall (1954).  This index is
principally based on pressure differences^  Hall found this index, in conjunction
with other criteria, produced a hindcast accuracy of 92 percent but an overpredic-
tion of 65 percent.  Using the Hall criterion, 18 May 77, 19 May 77, 7 June 77,
18 Aug 77, and 28 Sept 77 were identified as days with high lake breeze poten-
tial.  The occurrence of lake breezes was confirmed for 7 June77 and 28 Sept 77.


                                      29

-------
Considering the paucity of over-lake data,  lack  of confirmation  does  not  neces-
sarily indicate non-existence.

     The lack of roughness over the lake's  surface,  especially in  comparison
to that present over the Chicago/Gary source  region,  will  tend to  reduce  the
frictional turning effect on aerosol trajectories.  Pilot-balloon  (pibal)
releases from the midlake sampling  point tend to confirm this expectation for
the August and September outings.   Of 12 pibal  releases  during the August
outing, only 2 were not straight line trajectories0   Of  11  pibal  releases
during the September outing, only  1 was not a straight line trajectory.   This
would suggest that the August 18 and September 28 field  days were  not affected
by the lake wind or gustfront phenomena.  In  contrast, the May 19  lake effect
appears to have met classical lake  breeze criteria,  having a return flow
layer aloft.  Lyons (1975) has  shown this return flow to bring more large
aerosol (d > 1 urn) to the midlake  surface layer than would otherwise  be
expected.   This is exactly what the ASAS data shown in Table 9  indicates.
The mass concentration of total aerosol increases from 23 to 333 yg/m3 between
the morning and afternoon of May 19 according to the ASAS data and from 47 to
141 yg/m3  according to the bscat of the IN.  The increase is greater  for  the
ASAS since it responds to d > 1.0  yro aerosol.

     SethuRaman (1976) has considered the air mass modification  caused by the
change in  surface characteristics  from the  rough, warm overland  surface to the
smooth, cool overlake surface.   In  addition to the mesoscale wind  effects
stated above, a pronounced thermal  internal boundary layer (TIBL)  often devel-
ops.  This TIBL reduces the vertical dispersion of aerosol  overlake and may cap
off the vertical mixing.  The height to which mixing occurs (H)  may have
considerable horizontal variation  and is difficult to deduce from  a single
measurement.  One semi-empirical calculation of H is given by Raynor, et  al.  as
                                                                (14)

                                         -AT/Az

where 14 is the friction velocity, F is the fetch in meters, AT/A? is the lapse
rate in oC-nT1, Ty is the upwind (land-based) surface temperature, and TQ is
the downwind (lake) surface temperature^  A calculation of this TIBL height, H,
gives 235 ± 100 m for the May outing as a whole, with August and September
having too small a percentage of cases with TQ
-------
following a wind shift can also be expected to affect the upward flux of
aerosolsc  This is assumed to be of minor importance for the elements
analyzed except Ca and Mgu

     Windshifts are often accompanied by rapid changes in air temperatures.
Besides having the obvious effect of changing stability conditions, this
temperature change and any associated moisture change will  affect the evap-
oration rate.  It will always be somewhat simplistic to view these various
atmospheric changes as independent effects„  The amount of moisture in the
air will affect the air buoyancy and, more importantly, can affect both
the surface reactivity and the chemical  characteristics of aerosol particles.
Since changes in cloud cover are often associated with air mass changes, the
effect of solar radiation on many atmospheric chemical reactions, particularly
those involving nitric oxide, ozone and various organic materials, may provide
another link between synoptic events and aerosol characteristics.

     The air mass changes resulting from synoptic flow have effects similar  to
mesoscale air mass changes.  They are generally on a longer time scale and are
more easily identifiable.  This may not be the case, however, when the front
is weak, or undergoing formation or dissolution,,  Indeed an instance of the
first effect was only discerned for the case of September 30, 1977 by post-
analysis.

     Besides estimates of uncertainty based on temporal variability or
meteorological parameters, spatial variability must be considered.  The
midlake v^ values obtained can be considered locally valid within a factor
of 2 or 3.  The process of extrapolation of these results to the whole of
the southern lake or to an entire year may introduce even more uncertainty.
At several points in this analysis, an effort is made to determine what is
the minimum reasonable annual loading.  Since these values  are rather uncertain,
an investigation of the sources of uncertainty and methods  of increasing the
degree to which the results may be trusted is in order.

     The deposition of aerosols over the surface of the lake is considered
to be uniform and to vary principally with distance from the shore — probably
in a non-linear fashion.  It is likely,  however, that most of the gradient
occurs near-shore.  The principal factors to consider are the short but
increasing fetch as an air parcel moves away from shore, as well as the change
in wind field and its effect on surface roughness and on AT with increasing
distance from shore.  The surface layer of constant mass flux will be nonexistent
at the sharp discontinuity that is the shore (Vugts and Businger, 1977).  In
the first few kilometers from shore the reduced friction over water often causes
a wind direction and speed change which, along with the temperature instability
nearshore, causes a slow growth in the surface layer depth.  However, between
5 and 10 kilometers from shore the surface layer can be expected fco have reached
50 to 90 percent of its midlake depth (Hess and Hicks, 1975).  The temperature
difference between air and water surface stabilizes within a few kilometers
distance from shore.  In over 20 cases of NCAR aircraft overflights, the air
temperature at 15 to 35 meters above the lake's surface always came to equil-
ibrium with lake effects within 5 to 10 kilometers of shore.  The water surface

                                    31

-------
temperature itself decreased in the first few kilometers,  but also came to an
equilibrium temperature, usually within the first 5 kilometers from shore0
Water surface temperature anomalies did appear occasionally even  at midlake —
very likely due to upwelling of warm,  or cold, water from  below the surface„
It is safe to say, however, that the AT observed at the midlake sampling point
is generally representative of 70 to 90 percent of the  southern basin's lake
surface -- unless synoptic scale or strong mesoscale fronts are presento  The
subsequent spatial variation of wind speed and direction could interact with
aerosol concentrations in such a way as to make a plume either more or less
diffuse, or far-reaching (Lyons, Keen, and Northhouse,  1974).  Indeed the
spatial variations in concentration found in plumes or  even more  diverse
gradients bring into question the representativeness of sampling.   Nearshore
wind speed changes can be expected but should come to equilibrium with the
relatively constant roughness of the lake's surface at  least as rapidly as
the air temperature,,  Thus again, unless synoptic conditions are  dominant, the
wind speed measured at the midlake sampling site should be representative of
70 to 90 percent of the southern basin's lake surface.

     Changes in wind direction and in the variability in wind direction can
be expected to affect not only the transport of aerosol, but the  drag coef-
ficient, since it takes some time for a wave regime to  come into  equilibrium
with a new wind direction (Donelan, 1977K  Though there is insufficient data
to verify that directions measured at midlake were representative, the lack of
large roughness elements on the lake and the care taken to restrict sampling
periods to a constant wind regime suggests that the ± a envelopes used were
sufficient for the purposes to which the back trajectory method was applied.

     Since the steady state 3-to 6-hour averaged bulk deposition  may be
considered reasonably constant over most of the lake's  surface, the nidlake
concentration of any one constituent is a fraction of the  source  region concen-
tration given by
                     Cmidlake    Cshore P  " vd  *   TTJ          ^15^

where Cmidlake varies only as a function of source  concentration,  shore to
midlake travel time (t), AT, and the mean depth (H) to which mixing of that
constituent occurs.  When considering long  range transport,  H is usually
chosen as the average mixed layer height of approximately 1000 m.   However,
transport of warm air from Chicago over the cold lake surface in spring and
early summer creates the TIBL discussed previously,which sharply reduces mixing
in the lowest few hundred meters above the  surface  and allows almost no mixing
above the TIBL (Lyons, et al., 1974).  If Equation  (14) is used to estimate the
TIBL height and u* is taken to be 10 cm-s'1, H is 150 m for the period of
May 18-20, 1977.  The wind direction was 225° ± 15°, and as a result CSj1ore
was the Chicago source concentration.  Using H = 150 rc, the previously esti-
mated AT and v^ by data set and knowing Cro-j^lake' an estimate of Csnore
can be obtained by equation (15).  Si even1 ng, et al. (1979)  then used
Chicago's 22 sampling station average sulfate, Pb,  Fe, and Mn concentrations
to confirm the validity of H - 150 r, and of the use of the bulk deposition
velocity v,j as being representative of the  actual deposition velocity pre-
vailing.
                                    32

-------
     This confirmation assumes the aerosol  size distribution remains  essentially
invariant with distance from shore,.  In an  effort to assess  the effects  of  short-
range overlake transport, the aircraft data taken on 30 Sept 77 for the  period
06:45-06:57 CDT were examined.  During this time, the aircraft  was  headed
directly from Midway Airport toward the rcidlake sampling site at an altitude
of 310 m prior to 06:48 CDT (shore) and 30m after 06:49 CDT.   There  was a steady,
persistent tailwind.  Although these circumstances augered well, there were
some complicating circumstances.  A front was just forming and  passing  through
the area, as described above, but exact time of passage could not be determined.
Additionally, the flight path passed over a tanker about 20 km  from shore  at
6:50 CDT.  A large increase in aerosol concentration was observed (Figure  38)
throughout the size range measured (0.1 
-------
of interest to be a small  fraction of the sampling  height  concentration.
Only if this is true will  the mass transfer coefficient, Xm,  to  be  equal
to v^.  It is felt that as long as the surface  microlayer  (Andren,  1975)
is not present, this assumption is likely to be valicL  The  validity of
this assumption is integrally tied to a fourth  assumption; that  a spatially
continuous laminar sublayer prevails only with  wind speeds of less  than
2.5 m-s""1 and probably only for 2 m-s"1 or less.   It is very  likely that
high concentration surface microlayers occur primarily when  the  laminar
sublayer disallows transfer to the lake surface.  Wind speed  measurements
in conjunction with surface microlayer studies  are  not entirely  consistent
for Lake Michigan samples  (Andren, 1975; Eisenreich, 1978), but certainly  the
vast majority of cases with surface microlayer  enrichment  of  trace  elements
or nutrients occur in very light winds.  If the ratio of surface to bulk
water trace element concentration is near one,  the  microlayer is probably
not present and the laminar sublayer is also likely not to be present.
Thus, the concentration of trace elements or nutrients at  the air/water
interface will be much less than their respective concentrations at the 5  m
sampling height.

     Support for the assumption that Xm - vjc (i.e., that  CQ  a 0) in higher
winds — when the laminar  sublayer is discontinuous -- is  given  by  Stulov,
et al. (1978),who considered the collision efficiency of aerosols at water
surfaceSo  At a clean water surface all collisions  are effective and lead  to
aerosol transfer into the  liquid.  When aerosols  collide with other aerosols
previously deposited on that water surface, they  usually rebound.   Aerosol -
aerosol collisions might occur when the surface microlayer is present,  but the
probability is otherwise extremely low.  Further, Maszaros  (1977) has found
that 90 percent of summer  continental aerosols  not  unlike  those  found over Lake
Michigan are water-soluble, and in winter months 50  percent or more  are water-
soluble.  As long as the laminar sublayer and surface microlayer are not
present, it appears that the air/water interface  constituent concentrations  are
indeed a small fraction of 5 m height concentrations„ By  assuming  v^ to  be
zero for wind speeds of <_2.0 m-s"1, the strong reduction  in  aerosol mass
transfer in the presence of either of these layers  should  yield  conservative
estimates of total loading.

     One final assumption  -- a most important one — must  be considered.   The
use of the momentum drag coefficient to estimate  v^ carries  with it the implicit
assumption that the momentum drag coefficient,  Km,  is equal  to the  eddy diffu-
sion coefficient, KQU  This equality is usually referred to  as the  Reynolds
analogy0  Since much more  is known about momentum transfer than  mass transfer
in the atmosphere, correlations of mass transfer based on  momentum  transfer
have been developed by Reynolds and others.  The  Reynolds  analogy simply
states that in a turbulent regime mass and momentum are  transferred in an
analogous way and, thus, KQ = 1C.  This analogy can be  restated  for the experi-
mental conditions over Lake Michigan as:

                     Momentum Flux    _  Aerosol  Flux           (13)
                     Momentum at 5 m          C5

                                    34

-------
The deposition velocity given by the latter ratio can then be equated to the
former ratio which is identically Cpp • 115,  Does the Reynolds analogy,
Kp - Kfj,, hold for sampling conditions over Lake Michigan?  This question
cannot be simply answered; it requires a significant digression by careful
consideration of mass and momentum transfer near the air/water interface.
The reader to whom this is of little concern is advised to omit this discus-
sion and resume at the beginning of Section 5.

     A diagramatic representation of conditions near the interface is shown
in Figure 39.  Two quite different conditions can be expected to prevail
within the surface layer above and at the lake's surface0  A smooth flow
regime is shown on the left.  Here, the surface layer is seen as three
separate ones:  a turbulent layer, a buffer layer, and a laminar sublayer.
The turbulent layer constitutes the largest portion of the surface layer
depth of several meters to several tens of meters.  A continuous laminar
sublayer enveloping all or nearly all of the surface roughness elements
follows the gross observable contour of the lake's surface.,

     Mass transfer within the laminar sublayer is dominated by molecular
transport, i.e. Brownian motion.  The buffer layer is a transition zone  in
which turbulence is reduced due to the close proximity of the surface.  Both
turbulence-induced eddies and Brownian motion contribute to mass transfer in
the buffer layer.  Turbulent transfer may be characterized by the eddy transfer
coefficient (Kp) (cm2-sec~1)> whereas molecular transport due to Brownian
motion is characterized by the molecular diffusion coefficient (D) (cm2-sec~1)

     The molecular Schmidt number, Sc , is the ratio of the kinematic viscosity
of air, v(cm2-sec~1) to d and gives a measure of the rate of mass transfer  to
be expected due to molecular diffusion within the laminar sublayer.  When
ScL < 1 or D > v,a condition of relatively good transfer will prevail, whereas
ScL > 1 or D < v is a condition of relatively poor transfer,,  v for air  at
20° C and 1 atm pressure is 0015 cm2-sec~1, while D for spherical aerosols  of
diameter 0.1 and 1,0 \m is 6U75 X 10"6 and 2.77 X 10~7 cm2-sec~1 respectively.
The molecular Schmidt number is clearly much greater than one for the OJ  to
1.0 ym and larger aerosol size range.  Thus, the continuous laminar sublayer of
the smooth flow regime blocks the transfer of aerosols and associated trace
elements and nutrients.

     As higher wind speeds and less stable air prevail within the surface layer,
a rough flow regime is eventually met (shown on the right of Figure 39)  in  which
many of the surface roughness elements protrude outside of a now discontinuous
laminar sublayer resulting in a more efficient path for mass transfer.  The wind
speed and stability at which this rough flow regime is actualized is quite
uncertain.  Yet it is certain that turbulent transfer will  increase the  overall
mass transfer substantially.  Turbulent transfer dominates the mass transfer of
aerosols in both smooth and rough flow regimes for the majority of the surface
layer which constitutes the turbulent layer.  A turbulent Schmidt number, ScT,
is defined as K^/Kp and is applicable to the turbulent layer where Km and Kp are
the momentum and eddy transfer coefficients as before.  Of course, when  ScT = 1

                                    35

-------
or Km = KQ the Reynolds analogy holdsu   Businger,  et_alo  (1971)  found ScT - 0.74
in neutral conditions within the turbulent portion ofHEhe surface layer»
Blackader (1975) estimated ScT to be 0.85 in  neutral  conditions.   Unstable
conditions within the turbulent portion of the  surface  layer could result in
Sc"f values of 0U75 to no less  than 0.67 times the  neutral  Sc"'" or 0.50 (Kraft,
1977).  Stable conditions could result  in Sc' values  of no more  than 1.17 times
the maximum neutral  value, or 0.99 (Kraft, 1977).   Because of sedimentation
effects, mass transfer is usually more  efficient than momentum transfer in the
turbulent layer.  The Reynolds analogy  holds, and  the momentum drag coefficient
is an accurate parameter for use in mass transfer  calculations.   The buffer layer
is a layer of transition between turbulent transfer and.molecular transfer.  As
such, mass transfer may be reduced relative to  momentum transfer.  However, this
reduction is very much dependent upon whether the  smooth  or rough flow regime,
i.e. a continuous or discontinuous laminar sublayer,  prevails.

     Given that the turbulent layer may be characterized  in the  same way
regardless of whether smooth or rough flow prevails,  the  total surface layer
mass transfer, Xm, can be split into two parts:   a turbulent layer mass transfer,
Xj|  and a remainder mass transfer, XR.   Since the  resistance to  mass transfer,
   i), is additive in the surface layer

                  R(h)  =  RTI(h) ^  RR(h)  = y— + i-              (16)
                            I L       l\        ATI    An
where h is the height of measurement,  Ry|_(h)  is  the  resistance  to  mass  transfer
in the turbulent layer, and R^(h)  is the remainder nass  transfer resistance.
     It can be shown (SI inn, 1978)  that

                                  1
                             -  	  _  _    =   c
                      TL     "          "   *        D
     Kraft (1977) has shown that a more accurate statement must account for the
combined thickness of the buffer layer and laminar sublayer and that Sc1  may not
be assumed equal to one.  As a result
                RTL
                           *TL
where Y  is a constant between 2 and 10.

      In  most cases  (and certainly over the lake)Y'U* is much less than un so
that  XT,  =  [TTh  • CD • Sc1]'1.  Since XJL « C'1, the mass concentration (or that

-------
of an aerosol constituent) is proportional to un- Cg • Sc1.  The resistance
to mass transfer for the turbulent layer is then approximately proportional
to uft since, as Figure 34 shows, CQ is a weak function of wind speecL   This
suggests that the midlake mass and trace element concentrations should vary
with wind speed in a nearly linear fashion.  The correlation coefficient
between wind speed and mass was -0.12, and the percentage of linear variation
in all chemical constituents explained by wind speed was 9 percent.  However,
a strong linear dependence of mass and trace elment concentrations on  AT was
found.  The correlation coefficient between AT and mass was 0.52.  Since the
dispersion of aerosols in the vertical as well as the intensity of vertical
velocity fluctuations is inversely proporational to AT in neutral to nearly
all conditions of thermally stable surface layer air (Izumi and Caughey, 1976),
the aerosol concentration may be expected to be directly proportional  to ATU
All but two or three concentration data sets were collected in neutral to stable
air, and the correlation coefficient shows the direct proportionality  expected.
Thermally controlled turbulence within the turbulent and buffer layers must be
contributing significantly to aerosol  mass transfer in the overlake surface
layer as these aerosol traverse from shore to midlake.  The manner in  which
thermally generated turbulence is reduced in the buffer layer may then be an
important research question.  Kp in the buffer layer will be somewhat  reduced
relative to KM as a result, but it is  unknown by how much.  Binkowski  (1979)
has shown that Kg « X • aw in the surface layer where A is the wavelength of
the peak in the vertical velocity spectrum and aw is the standard deviation of
the fluctuating vertical velocity component.  As theory or experiments determine
the difference between A and aw in the turbulent surface layer from that in the
buffer layer, a proportionality constant e in KQ = eKm may be estimated and the
degree and conditions under which the  Reynolds analogy hblds approximated.

     The lack of any linear dependence of the measured midlake concentration
upon wind speed strongly suggests mass transfer may be largely independent of
sublayer presence.  Fewer than 10 percent of the data sets were obtained in
wind speeds of 2 m-s"1 or less.  Kondo, et al. (1973) found that 30 percent of
the roughness elements related to high-frequency components of ocean waves
protrude outside the laminar sublayer  at 2 m-s'1 wind speeds.   At 8 m-s'1 over
99 percent of the high-frequency roughness elements protrude outside the remain-
ing discontinuous sublayer.  Schlichting (1968) observes that "from a  physical
point of view it must be concluded that the ratio of the height of the protru-
sions to the laminar sublayer thickness should be the determining factor" for
smooth or rough flow.  Since 30 percent of the high frequency ocean wave
components protrude outside the sublayer at 2 m-s-1 and the mean sea-surface
roughness height equals the laminar sublayer thickness at 3 m-s-1, smooth flow
conditions will probably not prevail above 2 m-s-1 and certainly not above  3 m-s-1.
Thus, a continuous sublayer was likely present for no more than 10 percent  of the
data sets collected,,

     Fully rough flow for momentum transfer  is usually assumed to prevail at
UIQ  =   7  to 8 m-s'1 and above  (Wu, 1972), whereas one may expect a transition
flow between ~ 3 m-s'1 and 7-8 m-s"1,  and smooth  flow below-3 m-s'1.   It has
been stated that momentum transfer or Km can  be expected to be larger  (possibly
much larger) than KD in the buffer layer during transition and smooth  flow
due  to what is known as the bluff body effect  (Chamberlain, 1968).  In the

                                    37

-------
immediate vicinity of the surface,  momentum is  transferred  to  the surface  by
skin friction and pressure forces,  which are due to fluid  impacting on  the
roughness elements of the surface.   Momentum transferred by pressure forces
at the surface is known as form drag,  and since there  is no analogue in mass
transfer a possibly severe limitation  of the Reynolds  analogy  is  encountered
within the buffer layer.  As a result, it is uncertain at what wind speeds
the smooth, transition, and rough flow regimes  apply to mass transfer.

     The form drag force depends on the shape and orientation  of  the bluff
body — here, the roughness elements at the air/water  interface.   Maximum
form drag is experienced by surfaces at right angles to fluid  flow and  the
force can be estimated by assuming  pu^ to be the momentum  transferred per
unit volume of air at a point on the surface where fluid is brought to  rest
after being decelerated from a velocity un.  Since uh/2 is  the mean velocity
of the bulk fluid, the rate at which momentum is lost  is puu • u^/2. In fact,
fluid slip around the sides of the  bluff body results  in a  force.  <  pu^/20
It can be expected that there is sufficient eddy formation  in  the vicinity of
roughness elements to produce a form drag less  than the theoretical  maximum
per unit area of Cpp- ufj • p/2, where  CDF 1S the form  dra9  coefficient.

     Even though the form drag force is not at  its maximum, it still contrib-
utes to the  momentum transfer coefficient, Km, with no  analogous contribution
to KQ.  The wind speed regime for which smooth, transition, and rough flow
prevails may therefore differ in addition to Km - KQ.   There is,  however,  the
aerosol mass transfer mechanism of  turbulent inertial  deposition  in the buffer
layer.  Suppose aerosols in an idealized circular eddy are  deflected through
90° around the circular arc at the  radius of this eddy, r.   The centrifugal
acceleration involved is r (de/dt)2, where de/dt is the angular velocity.
This produces a radial "skid" of the aerosols relative to  the  accelerating air.
This implies that aerosols immediately above the laminar sublayer will  travel
a finite distance before coming to  rest,,  This  finite  distance is called the
stop-distance, S, and is given by (Friedlander, 1977):
                                   p
                             S  =
(19)
                                      18 v
     where:   p  is the aerosol  density

              v  is the kinematic viscosity of air

              p  is the air density


              vo is the initial  velocity imparted to the aerosol,  and

              d  is the aerosol  diameter.
                                    38

-------
Note the strong d2 dependence of S on aerosol diameter.  Clearly, when S
is greater than or equal to the laminar sublayer thickness, 6, the combined
mechanisms of turbulence and inertia of the aerosol or turbulent inertia!
deposition will cause an aerosol to traverse the sublayer.   It can be shown
(Twomey, 1977) that the collection efficiency of the water surface beneath the
sublayer is in fact S/5 = S/r, or that the ratio of aerosol stop-distance to
the mean radius of eddies determines their collection efficiency.  As before,
a knowledge of the wavelength spectrum of buffer layer turbulence would add
greatly to understanding of mass transfer at the lake's surface.

     Owen and Thomson (1963) have shown that horseshoe eddies wrap themselves
around the individual roughness elements for those that are closely spaced.
Such eddies can be expected to result from the high frequency wave components
referred to earlier.  Numerical models (Batchelor, 1967) and wind tunnel
experiments (Taneda, 1956) of the flow past a circular cylinder  (somewhat
like an individual high frequency wave component when protruding outside the
laminar sublayer) for neutral conditions and at 2  < Ui0 <  5 m/s show the
formation of closed streamlines downstream of the cylinder from 0.2 to 2.0
times the cylinder diameter.  This eddy formation may well  occur in the space
between the closely packed roughness elements at the lake's air/water interface..
Since the eddies are smaller than the roughness elements themselves and the
roughness element heights measured by Kondo, et al. (1973) at Uio  = 2 to 3 m-s"1
were about 1000 ym, the mean eddy radius may be 500 ym or less.  Turbulent
inertial deposition could then be a contributor (due to roughness element
created eddies) to deposition at the lake's surface despite the low wind speeds
of 2 to 3 m-s"1 for the half-micron and larger aerosol.

     If we now reconsider the data of Kondo, et a1. (1973), it is likely that
turbulent inertial deposition in the 2 to 5 or 6 m-s"1 range and turbulent
impaction to the roughness elements directly in the ~4 m-s"1 range or greater
wind speed ranges induce a low resistance path for aerosol  transfer when
compared to molecular diffusion transfer through the laminar sublayer.  At
2 m-s"1, 30 percent of the roughness elements protrude outside the sublayer so
that direct impaction due to turbulence-induced aerosol velocities and turbulent
inertial deposition across the sublayer combine to provide a "short-circuiting
path around" the laminar sublayer.  This path may be available to half or fewer
of the aerosols in the buffer layer but will be an efficient transfer path
despite the 2 m-s"1 wind speed.

     Thus, the transition from smooth to rough flow for aerosol mass transfer may
have already begun at 2 m-s"1.  At 3 m-s"1 and more the mean height of the rough-
ness elements  is equal to or greater than the sublayer thickness, and the aero-
sols may follow the low resistance turbulent inpaction or inertial  deposition
pathway to the lake's surface.  At 8 n-s"1, the mean roughness element height
is probably five times the mean laminar sublayer thickness  of <_  0.08 cm and
over 99 percent of the roughness elements protrude through  a veTy discontinuous
sublayer.   At  this wind speed the buffer layer itself -- usually  estimated to be
five times the sublayer thickness--is  only about equal  to  the mean roughness
element height.   Fully rough flow in the sense of turbulent eddies  controlling


                                    39

-------
mass and momentum transfer prevails.   In this  case  the data  of Businger,
et al.  (1971) and Blackader (1975)  can certainly  be applied  and KQ  *  K^
Though it cannot be stated with certainty,  Km  will  likely  become no more  than
2-to 3-fold KQ as smooth flow conditions become established,,   It is only  below
2 m-s~^ (and possibly 3 ni-s-^  under thermally  stable surface  layer  conditions)
that large discrepencies between Kp, and KQ  are realized.   The Reynolds  analogy
certainly holds for aerosol mass transfer in rough  flow conditions  and  may  well
be valid within a factor of two or  three at wind  speeds above 2-ms~^u

     All of the above digression into buffer layer  and laminar sublayer
dynamics argue for a rather small  resistance to aerosol mass  transfer for
those layers.  In particular, Xn becomes equal to Xy[_ as the  rough  flow
condition is approached, leaving only the mass transfer resistance in tne
turbulent layer, RjL(h), as the total resistance  to transfer.  As smooth flow
conditions are approached, Xp passes through a  point of equality with Xj|_ and.
may  become an order of magnitude less than Xj|_ for  wind speeds of  <  2  m-s"1.

     If the vertical gradient in aerosol number  concentration can be measured,
a field experiment to test the above theoretical  discussion is feasible.  It
was  stated in Section 3 that this was found possible using the ASAS when
vertical ship motion was minimal.  The vertical  velocity measured at a  height
of 5 m  (W$) as detected by a Gill anemometer,  was used to monitor the ship's
vertical motion.  Wind speeds above 6 to 7 m-s"1  and/or wind direction changes
causing ship instability usually forced aerosol  vertical gradients to be
disregarded -- simply due to the large vertical  motions encountered.  There
were two notable exceptions in the 7 to 8 m-s'1  wind speed range.  Wind speeds
above 6 m-s-1 also usually generated sufficient lake spray that the logarithmic
criterion for the aerosol vertical concentration gradient was not met to within
the  90  percent confidence expected.  That is,  the aerosol concentration profile
with height should be logarithmic -- if no lake  source contributes.  Only those
concentration profiles that were logarithmic  at the 90 percent confidence level
or better were considered for analysis.  Heights of 3.7 ± 0U2 m and 6U4 ± 0.2 m
at the ship and 4, 2, and  1 m to within ± 0.05 m at  the crib site were used here
to estimate deposition velocities.

     The profile  (or gradient) method has been used by Gillette  (1972a) to
estimate aerosol deposition velocities above  a soil surface.  Garland  (1974)  and
Whelpdale  (1974) have used the method to estimate sulfur dioxide gas deposition
over land and water surfaces.  When possible, wind  speed and  concentration
profiles are simultaneously measured  in the lowest  10  to 20 meters above the
surface of interest.  The  profile method deposition velocity, vdp  at h is then
given  by:
                    ,      -   p  • CDD  • "h  •  LCh1 - ChJ           (20)
                    dPh         n.    _   ,,  1.  c
                                              uh
      where:    h2- <  h   <   hi


                                    40

-------
      Since a  certain  time must pass before sufficient aerosol are sampled by the
 ASAS  at  each  of  the two heights, the concentrations, wind speeds and profile
 method deposition velocities, v^p, are, in actuality, average values.  Sequen-
 tial  rather than simultaneous monitoring is essential for the few percent
 intercalibration accuracy required of two separate aerosol sampling instru-
 ments is  not  possible.  Precision and reproducibi1ity of ASAS aerosol number
 concentration counting was found to be better than 2 percent for the relatively
 low concentrations at midlake.  Thus, equation 20 becomes for the 5 meter mean
 height of measurement, with h-, = 6.4 m and h0 = 3.7 n and air density,
 p - 1 kg-nr3                  '


                              CDD ' U5 r6.4 " ^3.7j             (21)
                     Vdp5
         ["6.4 - U3U7J'
Because the ship sampling platform is not sufficiently stable to allow wind
speed profile measurement, a power law, un = 115 • (h/5)a, can be used as an
estimator of the 6.4 and 3.7 m wind speeds in the surface layer.  The power law
can be used in equation 21 to estimate the difference in 6.4 and 3.7 m wind
speeds as U5 [(6.4/5)a - (3.7/5)aJ.  a was determined by first calculating
the bulk Richardson number, RiR, as:
            Ri,
     g [AT + r (hi - h2)j

(T5 + 273)  |I352.  (h!  -  h2)
                                                                 (22)
     The Richardson number has been related to the Monin-Obukhov length for
Great Lakes conditions by Donelan, Birch and Beesley (1974), and finally Irwin
(1979) has related the Monin-Obukhov length to values of a.  For the conditions
encountered on Lake Michigan, OJ4 < a  <  0.65 values were obtained.  The overall
uncertainty introduced by not directly measuring wind speed was found_ to be
20 to 40 percent of the calculated wind speed difference.  Finally, v^ps can be
stated in terms of actually measured parameters at the ship by
                         (6.4/5)a - (3.7/5)
                                           a
                                                                 (23)
where the uncertainty introduced by Tg 4 - ^3 7 has been somewhat reduced by
sequentially sampling several times at'the 6,4 at the 3.7 m heights.  However,
an additional uncertainty of 10 to 30 percent was found.  Overall uncertainty
in vcjp5 estimates are then 30 to 70 percent.   Profile measurements at the crib
site were more accurate due to the fixed platform.  Aerosol number concentration
measurements at 4, 2, and 1 m heights gave an estimated 2 m reference deposition
velocity v    with 10 to 40 percent uncertainty.
     Since the bulk momentum drag coefficient-estimated deposition velocity is
Vh = CDD ' %, a correction factor for the difference between momentum transfer
and aerosol mass transfer can be stated as

                                   41

-------
                                "  Ch2   .         ]                 (24)
                   V           Q         / "1 \     / "2
                                        ._.     , ^
                    d          "h

This calculation affords a way to  use  the  experimental  data  to  test  the
theoretical arguments made earlier.   Four  field  profile data  sets will be
considered here.  These profile data  sets  were selected to demonstrate
variation of v 1.5 ym were observed to  be statistically significant.
For the half micron aerosol, the v"d  /vd ratio of  0.2 was  one  of the  smallest val-
ues of the entire profile data sets.   At the other extreme,  the 16  Aug  77 ship
profile set with a wind speed of  7.9 ± 0.3 m/s has the  extreme high vdp/vd
value of the errtire profile data  sets.  For all  sizes of aerosol OJ  <  d
<  2U0 ym the v^n "is larger than  vd, despite the calculated  1.3 cm-s"1  bulk
deposition velocity for momentum  transfer.  Note also the  relative  independence
from aerosol size as compared to  the case of 23  June  78 which  had lower wind
speeds, yet had similar temperature  stability.

     The 28 Sept 77 profile set had one of the lowest average  wind  speeds,
1.3 ± 0U3 m-s-1, but still about  the same temperature stability as  the  23 June  78
and 16 Aug 77 cases.  This low wind speed most certainly has brought  on the
smooth flow case since the ratio  of high to low  position counts is  one  or less
than one throughout the OJ  < d  <  2.6 ym size range.

     The most interesting profile  set may be the__sequence  on 19 May 77, when
unseasonably warm air over land brought about  a  AT =  8.5°C or  more  at midlake.
This extremely stable surface layer was observed at 13:30  to 16:30  CDT  to be  in
a state of near microscale stagnation since the  wind  speed was only 1.6 ±
0.5 m-s"1 and wind direction was  extremely steady  (105° ±  5°)  during  the  four
hour sampling period.  As a result, the lowest ratios of high  to low  position
counts for any profile set were recorded.  This, despite a 180° wind  direction
shift in the 12:00 to 13:00 CDT period.  The laminar  sublayer  and buffer  layer
apparently combine to totally block aerosol mass transfer.   Since the wind
speed is below 2 m-s"1, this is also the result  expected for momentum transfer.
Since Rig is 0,22, turbulence within the surface layer  as  a  whole can be  expected

                                    42

-------
to be substantially suppressed.  The average ratio of aerosol  counts at the
6.4 m height to that at the 3.7 m height is less than 009, with a range from
0.75 to 0.95.  Thus, aerosol were actually being "stacked up"  in the lowest
few meters, with gravitational settling possibly having an effect on the ratio
for 1 ym and larger aerosols.

     Between 16:30 and 17:00 CDT, a rather unusual sharp increase in wind speed
unaccompanied by any noticeable wind direction or AT change was observed.
Thereafter, for over one hour the wind speed was steady at 4.2 ± 0.3 m/s.  The
surface layer during the 17:00 to 18:30 CDT period was clearly one in which
mass (and momentum) transfer took place.  This is exemplified  by the count
ratio for 0.6  <  d  <  2.6 ym aerosol being 1.34,_as compared to 0.77 before
jthe wind speed increase.  However, the ratio of vdD/vd is quite sma_ll with
vdp (0.1 < d < 2.2 ym) =[o.05 ± 0.02jvd.  Since vd = 0.25 cm-s'1, vdp * 0.01
cm-s'1. The gravitational settling contribution to 0.6 < d < 2»6 ym aerosol
still appears to be present._ A comparison with the 23 June 78 profile set
shows an eight-fold smaller vdp for this 19 May 77 17:00 to 18:30 period for
the 0.1 < d < 0.2 ym aerosols.  The much stronger temperature  stability on
19 May 77 probably reduces the actual deposition velocity.  Given the short-
lived micrometeorological change from 17:00 to 18:30 CDT on 19 May 78, the
laminar sublayer so well established during the 13:30 to 16:30 time period
was very likely a much more significant factor during the 17:00 to 18:30 time
period than on 23 June 78, when the wind speed was approximately 4 m/s that
entire dayu  Although the data cannot conclusively argue the point, it is
likely that turbulent impaction and turbulent inertial deposition are both
substantially reduced for the 17:00 to 18:30 period of 19 May  77, as compared
to the 23 June 78 case.

     By 20:00 CDT on 19 May 77, the strong thermal stability of the surface
layer appears to have blocked most aerosol transfer again.,  Despite the 40°
wind direction shift which has occurred^ the strong stability, along with
somewhat lower wind speeds, results in vdn - 0 for d >. 0.3 ym aerosol, as well
as a small value for the ratio vdp/vd of 0.1 to 0.2.  This does not seem
unusual except that the small aerosol now are depositing, whereas during the
17:00 to 18:30 period the large aerosol had five-fold larger vdp than did the
small aerosol„  The large aerosol concentration was depleted relative to small
aerosol concentration by 20:00 CDT.  Note that the lake source (WD = 105°) did
not contain a high concentration of large aerosol,.and that the wind direction
change did not bring replacement large aerosol to the sampling site.  A possible
explanation for the deposition of small aerosol during the 20:00 to 21:15 CDT
period, despite the lack of large aerosol, is that there had been little previous
deposition of small aerosol.  Their high concentrations, in conjunction with
the wind direction change, finally brought about deposition of small aerosol
even though wind speeds diminished.  Finally, by the 22:45 to  24:00 CDT period,
vdp is zero, and steady state conditions were obtained with aerosol mass transfer
completely blocked.  Even though the wind speed is essentially the same as
during the 23 June 78 profile set, no deposition occurs because of the extreme
temperature stability,,


                                   43

-------
     Consideration__of nine moderate wind speed (2.4 < U5 < 8.2 m/s)  profile
sets results in a v^/v^ _ratio of 0U48 ± 0.19.  The mean_ vd for the  nine
sets was 0U7 cm-s"1 and AT = 2U6°C.  The dependence of v^p upon aerosol
size is somewhat variable from profile set to set.   However, no more than a
three-fold, and only occasionally more than a two-fold difference, in v^n
across 0.1 < d  <2.5 urn was observed.  The 23 June  78 data set was the only
exception among steady state profile sets.  Both these results support the
suggestion that turbulent impaction and turbulent inertial deposition contribute
to aerosol mass transfer in the transition from smooth flow to rough flowu
Indeed, since the v^p/v^ ratio for this transition  wind speed range  was
found to be about one-half, the Reynolds analogy is only mildly violated with
KD - K^/2 for these nine profile setsu  A hypothesis might then be set forth:
turbulent inertial deposition plus impaction may be for aerosol mass transfer
what form drag is for momentum transfer at the air/water interface of large
lakes.  This can only be stated as a hypothesis at  this point for too few data
have yet been obtained by the profile method.  Further, other transfer mechanisms
such as phoretic effects may also contribute to aerosol mass transfer.

     On a clinatological basis, the Reynolds analogy may hold to better  than
the factor of two suggested by the nine steady state profile sets.   As the
19 May 77 sequence pointed out, non-steady state conditions will  tend to
increase mass transfer.  Of course, momentum transfer will also be increased.
Donelan (1977) has shown that CQ approaches a minimum as tne wave field  approaches
maturity.  In contrast, "during a rising wind or near an upwind shore higher CQ
are encountered."  There is every reason to believe aerosol transfer may also
be increased by five-to ten-fold, for the laminar sublayer could not have estab-
lished itself in the early stages of wave development and turbulent eddies will
then cause aerosol impaction upon the developing waves„  Perhaps more significant
for mass transfer to the Great Lakes is that by having sampled aerosol concentra-
tions only under steady state conditions, the use of only steady state drag
coefficients in v^ estimation errs on the conservative (low estimate) side and
thus contributes to Great Lakes loading estimates also on the conservative or low
side of the probable meanu
                                    44

-------
                                  SECTION 5

                          INTERPRETATION OF RESULTS


     Two approaches were followed in the calculation of estimated annual  dry
loadings to the southern basin of Lake Michigan.  The most basic approach is
the overall-average method:

                   dry loading  =  Vd  •  C  •  A  •  T                 (25)

where "v^ is the mean bulk deposition velocity for all data sets (m-s""1),
C is the mean concentration of a given element (kg-nr3), A is the total water sur-
face area of thesouthern basin of Lake Michigan (2.9 X 1010 m2), T is that
part of the year when no precipitation occurs overlake (s-yr'1), and loading
is in kg-yr-.1  Equation (25), although straightforward, entails an underlying
assumption that deposition at the midlake point represents deposition over the
entire southern basin0   For single-site sampling, this is an unavoidable
assumption.  However, the surface layer conditions at midlake are certainly
more similar to the majority of the lake surface than are conditions on shore
or near to shore (see Section 4U3).  Moreover (25)  assumes that the data sets
which made up the mean  Vjj and C values were representative of the entire year's
variability in vj and C.  The values of v
-------
over a yearu  Even so, (26) is a refinement over (25) in that the variations in
Vg( and C due to surface layer meteorology and source region are taken into con-
sideration.

     Two factors in the loading estimate calculations, (25) and (26), were
quantified through field observations and related to overlake meteorological
variability:  vj and C0  Because the results here are based upon_sampling
done at a single site, temporal variations within data-set-mean vj and C are
important factors in viewing the resultant loading estimates.  The time-scale
resolution afforded by the 3 to 6 hour data set duration eliminates short-term
"patchiness" in the observed parameters, yet provides definition of the pre-
vailing surface-layer meteorological regime.  The data set initiation/termina-
tion criteria give priority consideration to the constancy of meteorological
observations.  As a consequence, the sets used in loading estimations generally
had standard deviations about the data set means of:  aWD  <  15°, au5 <
0.6 m-s-1, and aAT <  1°C.  However, this degree of steadiness does not con-
clusively point to a similar aerosol concentration steadiness.  Analysis of
the ASAS data indicates that variation about the set-averaged number concentra-
tion was generally less than twenty percent (see Section 4.2).  The IN-monitored
aerosol mass concentration also showed a similar degree of steadiness.   Each
data set thus represents a period of fairly homogeneous meteorological  and
aerosol concentration conditions.  The data set mean v
-------
presented, but caution in interpretation is suggested.  For the remaining 7
elements  (Al, Ca, Fe, Mg, Fin, Pb, and Zn), the uncertainty introduced to the
loading calculation by the concentration, C, factor is expected to be the
ICAP-AES  instrumental reproducibility confidence of less than 20 percent.

    _Jhe  binned data sets of Tables 1]  and 12 demonstrate clearly that both v 7.0 m-s"1) extrapolate
the data  to more unstable-air and high wind speed periods which were not
sampled in the summer season.  If winter season data had been available, the
values of v 7.0 m-s"1.  The highest
case sampled was "u"5 = 8.3 m-s"1.  Again, the available data sets  miss the cases
where higher V(j values might have been determined.  The F0 values used in the
extreme-case bins thus apply the available data to periods of the year when[
loading rates would be greater.  Because of this, and despite the reduced C in
the winter season, the loadings calculated here very likely are underestimates.
Given this consideration and the 2-to 3-fold uncertainty in the Cnp-based VH
determination, the lowest loading estimate from the overall-mean  (25) and bin-
method (26) calculations is divided by 2 and reported as a "minimum" loading.
                                     47

-------
     The binned data £f Tables 11,  12, and 13 reveal  several  kinds of general
trends in the v,j and C results which have direct bearing  upon interpretation
of the loading results.  In Table 11, C for almost all  elements  generally
increases as AT becomes more stable  (positive);  at the same_time, v,j decreases
by a factor of 5.  That is, for the stable-air cases  when C is relatively
higher, vd tends to be reduced^  Because loading depends  upon the product of
these two factors, a summertime "self-protection" mechanism of the lake surface
from stable-air (increased C) periods is implied.  For the late  falj_ and winter
unstable-air season, this mechanism is reversed; even though  lower C values
might be expected, increased v'd would still result in significant cold season
loadings.  In Table 12, similar seasonal trends appear.  Although less clearly
apparent in the C data, v^ varies strongly with increased wind speed.  This is
due not only to the use of the Coo-based vj, which increases  somewhat with us
directly (Figure 34), but also the  incidence of more  unstable air (AT <  0)
during higlher wind speed regimes.  Thus, the winter season overlake being the
increased U5 season (NOAA, 1975), the__summer_" self -protection" of the lake is
reversed in winter.  These trends in AT and us, then, appear to  indicate that
while winter season, i.e. unstable  air and high wind, conditions may lead to
Jnwer C expectations overlake, a partially compensatory trend toward increased
v,j during that season tends to continue loadings at a significant rate.   It
should be stressed here that these  data were collected during the May through
September 1977 period and therefore do__not include any winter season data.
The actual frequency of observance  of AT and "115 ranges during sampling does,
however, correspond to the climatologically expected  values of F0.  Trends in
these data are thus probably quite  valid.  The  source region bin data in
Table 13 point to  another aspect of dry deposHion loading.   Although clima-
tological expectations of AT and U5, and thus,  v^ are only weakly variable with
source sector, a C maximum is clearly indicated from  the  Southeast Shore and
Chicago/Gary areas (Figure  1 ); not an unexpected result.  The sum of F0 values
show that 53 £ercent of the year these two high-air pollution regions are sources
of increased C at the midlake site.  As a consequence, the Chicago/Gary and
Southeast Shore areas contribute from 60 to 85 percent of the total dry deposi-
tion loading to the southern basin.

     The significance of atmospheric inputs of trace  elements to the Lake
Michigan ecosystem is not well known (IJC, 1978).  Dry deposition appears to
be a major contributing source of at least four elements  on a percentage of
total loading basis (Table 14).  10 percent of Mn, 20 percent of Fe, 30 percent
of Zn, and at least 60 percent of Pb inputs to the lake are by dry deposition
of atmospheric aerosol.  Only for those elements which have very large natural
land run-off inputs are atmospheric loadings seen to  be negligible.  The trace
element  loading estimates are discussed in detail in  Appendix C.

     The quantitative minimum estimates for annual dry deposition of nutrients,
such as P, N02/N03, and $04 loading of the southern basin pose the immediate
qualitative questions of relative significance of dry deposition and ecological
impact.  Unfortunately, there is to date no data regarding atmospheric nitrate
inputs to the lake.  Gatz and Changnon (1976) estimate atmospheric aerosol N
input to be 20 to 45 times that of  P, which agrees well with the result here of
input N/P = 26U  Eisenreich, Emmling, and Beeton (1977) summarized current data

                                     48

-------
from the literature on the relative importance  of  various  P  loading  routes.
Based on their analyses of monthly bulk samples collected  around  the shore of
Lake Michigan in 1976, the atmospheric route accounts  for  15.7  percent  of all
P inputs.   They estimate P loading to the southern basin to  be  1.05  X 106 kg/yr.
Shoreline  data may, however,  overestimate total loading to the  lake  due to the
close proximity of sources to P sampling sites. For  comparison,  earlier model-
ing estimates of P loading to the southern basin of Lake Michigan suggest 9-18
X 103 kg/yr by the atmospheric route (Winchester and  Nifong,  1971).   The present
estimate of dry deposition loading is then seen to be an order  of magnitude
above modeling estimates and an order of magnitude below bulk sampler estimates„
Also, dry deposition of gaseous components, such as NOX and  862,  are not
accounted for in the annual loading estimates of this  worke   Removal of gas
phase pollutants in raindrops would be included in bulk precipitation measure-
ments.  There are no extensive wet deposition data from the  midlake  area to
support this contention, however, and the bulk P loading of  1.05  X 106  kg/yr„
remains a best available estimate.  Based on this, the particulate dry  deposition
estimates account for at least 15 percent of atmospheric inputs of nutrients  to
the southern basin„  Of more importance than this  percentage implies, however,
is the direct particulate deposition input of nutrient materials  to  the open
waters of the lakes, far removed from shore sources.   As Delumyea (1977) has
pointed out, all atmospherically input nutrients must enter  the surface water,
i.e. the zone of photosynthetic activity.  This is not the case for  runoff
water and direct discharge„  Not only the total amounts of nutrients atmospher-
ically input, but also the relative proportions of biologically active  chemicals
in aerosol, are of significance.  As Kilham and Titman (1976) point  out, a
nutrient supply of changed composition can have disruptive effects upon the
lake's biota.

     Appendix A describes the basic theory and use of factor analysis in the
identification of air pollution sources.  Factor analysis  is a  technique which
provides a way of recognizing and aiding in the interpretation  of the underlying
patterns of interrelationships in any multiple-parameter data base.   Besides
ascertaining the relationships, it condenses large sets of data into uncorrelated
sets of factors.  These factors can then be subjectively labeled  or  classified
with a physical significance attached to the group of variables that constitute
that factor.

     The application of factor analysis to the midlake data  base  results in  the
specification of 6 uncorrelated factors, as shown  in  Table 15.  The  labeling  or
classification of the 6 factors, as well as a statement of primary and  secondary
member variables, is shown0  A primary member is one  for which  that  factor
explains 75 percent or more of the variable's variance.  A secondary member  is
one for which that factor explains 25 percent or more of the variable's variance.
The percent variation row in Table 15 states the percentage  of  total variance in
the midlake data base that is explained by each factor.   In  factor 1, the  trace
metals Ca, Mg, Al, Fe, Mn, Pb, Ti and Zn are primary  members.  The Chicago/Gary
source region (C/G) and thermal stability (AT) over the  lake enhance the con-
centrations measured at midlake for the trace elements above.  This  is  the
basis for the interpretation given to factor 1. Notice  that despite the nutri-
ents' secondary membership in factor 1, they are primary members  of  factor 4,

                                     49

-------
along with aerosol mass.  The fine participate metals Pb, Mn, and Zn are
secondary members of this factor, but there is no wind direction association.
This strongly suggests a longer range transport of the nutrients and that this
factor is best described as fine particulate aerosol  mass, unrelated to any one
shore source regionu  The fact that mass is a primary member of this factor also
suggests that the variance in midlake mass (not the mean itself but mass varia-
bility) is largely controlled by fine particulate (d < 1 ym) aerosols.,  Factor 3
shows that wind direction (but not necessarily change in wind direction) is not
a controlling variable for metals' and nutrients' concentration variability at
midlake.  The elements associated with the coarse particulate fraction, viz..
Ca, Mg, and Al are, however, shown to be mildly dependent upon wind direction by
the factor 5 result,,  Temperature stability is shown in factor 5 to have a
controlling influence over N02/N03 concentration variability and is of secondary
importance to 864 and soil-derived coarse particulate.  Finally, the second
factor points out a negative though secondary dependence of Ti on wind speed.
Upon closer scrutiny (Sievering, et^al., 1979a) this appears to be attributable
to ship contamination under light wind speed conditions.  More importantly,
factor 2 suggests that no other metals data is contaminated by ship effluento

     A more in-depth interpretation of several different midlake data base
factor analyses is found in Sievering, et al, (1979a)»  The degree of surface
layer turbulence, the time to reach snore, and air mass type are additional
data (see Table 16) available for characterization of meteorological effects
upon midlake elemental  concentrations.  The bulk Richardson number, Rig, was
calculated to parameterize turbulence by data set as shown in Table 16, column 4,
Data sets were then aggregated into 19 sets, having Rig > 0.03 and 20 sets,
having Rig  <  0.03.  Factor analysis results are shown in Tables 17 and 18. Since
larger Rig values indicate a surface layer reduced in turbulence and mixing,
whereas smaller Rig values indicate enhanced turbulence and mixing, the results
are quite reasonable,,  It was instructive to reintroduce Ti in this factor
analysis by bulk Richardson number.  In the case of reduced turbulence (Table 17),
all the metals are members of factor 1, which explains over 50 percent of the
variance in the 19 filter sets'  data.  Only Pb is more strongly a member of
factor 4 and then along with Zn.  Factor 3 and 1  together suggest that so long
as turbulence is reduced, higher wind speeds are not necessary for high metal-
related aerosol concentrations at midlake.  Factor 2 shows that even with
reduced turbulence, strong surface layer temperature stability is additionally
necessary before mass concentrations and the concomitant large aerosol Al and Mg
concentrations can be substantial.  Table 18 shows that under conditions of
enhanced turbulence mass is a member with the small  aerosol Pb, Zn, and Mn, as
well as Fe and Al.  Given enhanced turbulence situations, neither surface layer
wind speed nor temperature stability alone affect this 20 filter set metals'
data base.   Rather, the increased turbulence, along with lower wind speed,
allows (as factor 3 points out)  backwash of ship effluent to contribute to Ti
loadings.  Factor 2 shows that large aerosol, probably soil derived Al, Ca, Fe,
and Ti, is present but does not contribute to mass concentration variance,,  Factor
suggests that Mg has its own unique source -- during  enhanced turbulence conditions,
The lake appears to be that source although it is unclear why Ca is not also a
member of this factor.   This may be because only 33 of the 40 data sets have Ca

                                     50

-------
concentrations above analytical detection limits.  The lake source region data
sets also suggested Mg to be lake derived.  In Table 15, AT stands alone.  It
does not affect the metals'  data during these enhanced turbulence conditions,

     The third column of Table 16 indicates the average trajectory time for
aerosols to reach shore for each of the 35 data sets for which trajectories
could be drawn.  By splitting the sets' data base into 15 with 3.5 hr. or
greater trajectory times and 17 with 3.0 hr. or shorter trajectory times, the
factor analysis results of Tables 19 and 20 are obtained.

     In the case of the long travel time trajectories, Ca5 Mg, Mn, Pb and Zn
metal concentration variations explain over 40 percent of the variation in the
data base.  The third factor suggests Al, Fe, and much of the Mg and Mn concen-
trations are dependent upon stability but without contribution to mass variance,,
There is some contradiction for Mg and Mn relative to the first factor.  Wind
speed again stands alone.  For the 17 short travel time trajectory data sets
no significant subgroupings were found.  Wind speed is a factor on its own,  not
contributing to any of the other variables.  Factor analysis as a function of
the air mass type designations of column 5, Table 16, showed no difference
between cP and mT air masses.

     Together, the factor analyses as a function of Richardson number and
travel  time show that higher midlake concentrations will generally be observed
with increased temperature stability except under conditions of enhanced turbu-
lence.   It was verified for this data base that wind speed is not linearly
related to midlake concentration variations.

     Several general conclusions can be drawn from the factor analyses of the
midlake data base:

1.  The variance in midlake metal concentrations are linearly independent of
    wind speed.  The only apparent exception is the possible dependence of fine
    particulate Pb and Zn upon high wind speeds under the condition of several
    hundred kilometers over lake fetch.

2.  Midlake metal  concentration variability is strongly and directly dependent
    upon temperature stability of the surface layer.

3.  Midlake mass concentration variability is usually more dependent on fine
    particulate aerosol with high Pb, Zn and Mn content than on the coarse
    particulate fraction.  The Chicago/Gary and southeast shore source regions
    are close enough to the sampling site to occasionally negate this conclusion,

4.  The processes of sedimentation and aerosol aging appear to be of secondary
    consequence when compared to temperature stability of the surface layer  as
    regards aerosol concentration variability within that surface layer„

5.  Source region related metal concentrations clearly show that midlake  metal
    concentration is more dependent upon nearshore sources than upon long range
    transport.

                                    51

-------
    The third and fifth conclusions together are most interesting  in that the
fine participate fraction dominates midlake mass concentration  despite the
fact that nearshore sources also dominate midlake mass concentration.   Of
greatest significance, however, are conclusions  1, 2 and 4.   The controlling
influence of temperature stability has implications for the  consideration of
resistance to transport across the air/water interface.   For example,  the
unusually large scatter in the enrichment factor diagrams (Figures 13 through 18)
can be attributed to the variance in Al  concentration caused by variations in
temperature stability.  Additional factor analysis results  can  be  found in
Sievering, et al. (1979a).

     The method of factor analysis, coupled with trajectory  analysis,  has
sufficed to determine some general aerosol/meteorological relationships and
to provide some information concerning the physical and chemical types of
aerosol associated with each.   If this method were to be used to provide much
greater specificity of results, a vastly larger  data set would  be  needed.  An
identification of the source of aerosol  by an examination of the physical and
chemical characteristics found is a tantalizing, nearly direct  method, but
requires more knowledge of transport, deposition and transformation mechanisms
than is currently available.

    A promising source typing  method may be numerical modeling.  Trajectory
models could increase the specificity of source  typing, at least on a case-
study basis.  The reliability  of these methods are highly dependent on the
density of the data available  for initialization and testing of models, and
the most useful models tend to be expensive of computation and  programming time0
Many of the atmospheric processes directly involved in transport,  particularly
turbulence, are not well understood.  Nevertheless, these methods  should provide
more specificity and reliability than the manual methods used.   Further source
region/type information may be gathered from the consideration  of  chemical
characteristics of aerosol.  A materials balance calculation (Gatz, 1975) can
give gross estimates of the natural and anthropogenic fractions of the midlake
concentrations, by first calculating:

               Soil derived     e C/Cref (1n soi1)                     {27)
               percentage   -    c/Cref tin aerosol)       10°

               Lake derived  _  n C/Cref (in lake water) ^              (2g)
               percentage         C/Cref (in aerosol)

and then finally subtracting the results of equations (27)  and  (28) from 100.
The remainder can be considered of "non-natural" origin, for soil  derived and
lake derived metal concentrations should constitute all of the  natural sources
of these metalsu  The remaining percentage should then, indeed, be "non-natural"
and, very likely, anthropogenic.  This assumes no unknown natural  causes for
higher atmospheric surface layer concentrations.  One rather uncertain natural
cause is represented by 5 and r\ of equations (27) and (28)  respectively:  ?
represents the fractional on that can occur in the process  of soil breakup to

                                     52

-------
aerosol  generation; n represents the fractionation that can occur in the
generation of atmospheric aerosol from the bubble bursting process at a sea
or lake air/water interface (Windom and Duce, 1976).  Only rarely has r, been
found to be greater than one (Rahn, 1976)0  However, the fractionation in the
transfer process from bulk water to the air/water interface (especially when
a surface microlayer is present) and to an atmospheric aerosol  can enrich
certain metals from 5-to 25-fold and in rare instances 100-fold or more.
For example, Bloch, et al. (1966) found an enrichment for Pb of 2-to 7-fold
in water-generated atmospheric aerosol.  £ will  here be assumed to be 1 and
n to be 25 or less.

     An estimate of the "non-natural" or anthropogenic percentage of midlake
atmospheric surface layer metal concentrations is shown in Table 21„  Bowen's
average midwestern soil metal  concentrations were used.  The average Lake
Michigan bulk water metal concentrations were based on over 40  samples taken
at the ship sampling site»  Each lake water sample was actually a mixture of
3, 5 and 7 m depth samples.  Although this bulk  water metal concentration
should be more representative for the analysis at hand, the water concentra-
tions measured differ insignificantly from previously reported  values (Torrey,
1976).  Al is used as the soil reference metal and Ca is used analogously for
bulk lake water,,  The significance of Table 21  results is that  a majority of
Pb, Zn, Mn and probably Fe at the 50-100 km offshore sampling site are very
likely anthropogenic in origin.  The tabulation  additionally shov/s the lake
not to be a significant source for Ti.

    Source type contributions to the remaining anthropogenic component of
midlake aerosol mass shown in the last column of Table 21 may be estimated by
materials balance analyses performed on the chemical composition of the
collected aerosol samples (Gartrell and Friedlander, 1975).  The method consists
of (1) estimating certain primary source contributions at a point (here midlake)
using chemical elements as tracers, and (2) supplementing these estimates with
emission inventory data for those primary sources for which no  characteristic
tracers are available.  Step (1) was used above  in calculating  the percent
soil-derived (with Al as the tracer) and percent lake-derived (with Mg or Ca
as the tracer) contributions to the overall midlake massu  Emission inventory
data on the utility, transportation, steel, and smelter industries could be used
as chemical composition fingerprints.  Kowalcyzk, Choquette and Gordon (1978)
have done this using emission inventories on the coal, oil, refuse and motor
vehicle source types in Washington, D.C. with Fe, V, Zn and Pb  as the respective
source type reference metals.   The method was found to identify source types
better than in any previous urban materials balance calculation.  This was
attributed to the fact that monitoring stations  were removed from any one
source type.  These findings suggest that the 1977 midlake data base, also well
removed from the influence of any single source  type, has potential for this
method.  This approach also seems promising since the relative percentage of
each element is essentially invariant with transport from source to midlake as
shown in Appendix C.  The observed 2-to 3-fold variation in vj values with
aerosol size could, however, change source type fingerprints.  Use of this
balance approach will then have to await the presently ongoing analysis and

                                    53

-------
 interpretation of aircraft size distribution data as a function of downwind
 distance  from sources, as well as the accession of the available emission
 inventory data for  the Lake Michigan basin.

     A graphic approach to identification of aerosol from different source
 regions  (and possibly types) has been suggested by Stolzenburg and Andren (1979).
 Figures 40, 41, 42, 43, 44, and 45 are plots of the elementaljnass fraction in
 aerosol,  normalized to the most abundant in the overall mean C data, sulfur:

                                      n (element of interest)
           Relative mass fraction  =  	
                                      n (sulfur)

The order of the elements along the abscissa is that of decreasing abundance
in the overall-mean data, with no other significance implied by the abscissa
labeling.  Because of a lack of other mid-Lake Michigan aerosol data, nothing
has yet appeared in the literature with which to directly compare these figures*,
In light of this, Figure 40  might be defined as a "baseline" plot, since it
represents the overall-mean C for all data collected during the 1977 sampling.
The slope of the line segment between two  elements corresponds to the relative
abundance between two elements, but no particular significance is assigned to
the value of this slopeu   Mhat i_s_ of importance in these figures is the general
"shape" of the plots.   In the baseline plot, Figure 40 , note that the trend
(by definition,  since the order of the elements was chosen for that to be true)
is from upper left to lower righto   Contrast that regularity to Figure 42, the
plot of C data from the West Shore source  region (see Figure !)„  Clearly, for
the West Shore aerosol, ratios of N/S, Mg/Fe, P/Zn, and Ti/Mn are quite different
from the baseline trend.   For the overlake transport (Figure 41 ) and East Shore
(Figure 43) region aerosol,  the plots differ noticeably from the baseline.  The
Southeast Shore  (Figure 44 )  and Chicago/Gary (Figure 45') source region plots
show much similarity to the trend of the baseline plot.  This similarity points
to the fact that, for over 50 percent of the year, aerosol from those two regions
passes overlake,  and was  thus an important fraction of the "baseline" determination
plot.  This graphic method,  or refined and subsequent developments based on it, may
hold promise of  becoming  a source type identification tool.   The largest obstacle
to fulfillment of this  promise is definition of a suitable baseline aerosol  compo-
sition and of uniquely  different source type compositions in the Lake Michigan
basin.  Beyond that, an expanding aerosol  data base in the Great Lakes literature
will  allow for better comparison between aerosol  from more than five source
regions; comparisons to source type plots  are an avenue for future research efforts
                                    54

-------
  43° 30'
O
ro
o
00
00
O
O
o
00
co
  43° 00'
o
ro
o
C-
00
O
O
o
r-
oo
                                SOURCE
                           1
                                  Overlake
                                 Transport
o
ro
o
to
CO
o
o
o
CD
co
                                                               LMUSKEGON
        MILWAUKEE <
           West
          Shore
          SOURC
  42° 30
                           LAKE
                        MICHIGAN
                                              May 18-20;
                                              June 6-10;
                                              August 14-19;
                                              September 26-30,
                                              1977 Location of
                                              R/V Simons
                                                   SOURCE
                               May-December, 1978
                               Location of Sampling
                               From 68th Street Crib
               FIGURE 1.  Location  of  Source  Regions and Sampling Data.
                                  55

-------
en
01
                                                                                                          100
                                            1                               10

                                             Aerodynamic Aerosol Diameter
100
                     FIGURE  2.   Aerosol  Collection Efficiencies  for Three-Stage Cascade Impactor.

-------
FIGURE 3.   R/V Simons and Sampling Boom.

-------
en
00
                          FIGURE 4.  68th St. Water Intake Crib and Sampling Windows,

-------
01
10


5 .
H.
in
B a.
oc
£

-------
            ^  X
                                                     * OF SETS « 24

                                                     MHX CONC. = 3.S1SBH
          3..X
CT>
o
       in

       c
       m
       u.
       a
I ..
                   X  XX   X  X  X X
                          I '  I I I  I I
                                       -t—I—»-
                                                             H - 1 - 1 - 1
                                                                        1 - 1-
            ZERD
                                                                              > '  i—i



                                                                           MHX.  CDNC.
                       FIGURE 6.  Distribution of Occurrences of Ca Concentrations,

-------
/ —
E.
S.
E M
1C
i 3-
u.
2.
I .
» 4





r\
X






\



K

* DF SETS « 37
MflX OJNC. « 2.30305 UB/M**3


i
X X


i
K
K
XX X
XXX X X 3
ZERO
MflX.  CDNC,
               FIGURE 7.   Distribution of Occurrences of Fe Concentrations.

-------
CTl
1 .

3.
in
B 2.
IE
i
u.
a
I .



•






X










X X


X




-4—




x:






X






X













X
— 1—
r\








* DF SETS B 3B
MflX CDNC. B 0.7B7BH0000 UH/M**3


X


X



XXX XX XX X X X XX X

            ZERO
MflX.  CDNC.
                           FIGURE 8.  Distribution  of Occurrences of Mg Concentrations.

-------
cr>
00
           TX X
         5.
      in
         3..
      * 2..
          I ..
                                                    * DF SETS  « 35
                                                    MflX OJNC.  « 0.11701      UE/M**3
                    XX
X   X   XX
XXX
                                            t  t—i—t-
X   X
                                                                         -t—i   t
                                                                                    t-M	1—-I
            ZERD
                                                                  MBX. CDNC,
                           FIGURE 9.  Distribution of Occurrences of Mn Concentrations.

-------
en
_l -
2.
E
in
E
u.
C3
* 1 -
l



•\ J
X




r* *



"V 4
X




•^ •
3

1
K
<

l
K
1 •
•»
* DF SETS « 37
MflX 
-------
        3..
en
CJl
in

m  2 J.   X  XX

a:
        I  ..XX
              X  XX
                                             * DF SET5 = 3E
                                             MflX CDNC.  = 0.03E1S
                                                                            UE/M**3
                                          XX
XXXXX  X
                                                 -t—i-
XX
                                                                    H	1	1	1—I	»	1	1	»
          ZEFIO
                                                                             MflX. CDNC.
                       FIGURE 11.   Distribution of Occurrences of Ti  Concentrations.

-------
           ^    X
         B..  X
                                         * DF SETS B  3B

                                         MflX  CDNC. B  0.48028     UE/M**3
         £..
      in
         H..X
      £



CT>     f£
CTl     1C
         2..
         I ..
                  XX
           XX   X
X   X
                                                                    1	1-4-!	1	1	1	1	1	1	1



                                                                                        . <:DNC
ZERD
                      FIGURE 12.  Distribution of Occurrences of Zn Concentrations.

-------
   10000
    1000
(T
LJ
O
o:
LU
     100
      10
\
*I7N
•3;
'36
•18
N
\



41 K?9
•50 *54
\.5I
\-4740
\ 53
\
•27
\ -24
v
\
^


22-:^i3
30
.55
34
,26 *I5
•21
\
\
•25
\
\
\

•8 "7
•6
•28 "5
•10
•14
•20
\
\
\
\
\
         10               100             1000


            Al CONCENTRATION, ng/m3

          FIGURE 13.  Pb Enrichment Factor vs. Al Concentration.
                      67

-------
   10000
o
o
h-
LJ
O
a:
z
Ld
c
N
     000
      100
       10


36% *
,7\
••37
•18
50
•
33


\
\
31
„, *47
^] *52 -54
\ *9
\2J -29
\ *24
•45 .\
53 V^
40'



\
,,\
22* -30\
1321
26
55.-2I
34 15
•25
\
\
\
\



•7
.6
J18 -5
\
•20 \
•14
\
\
\
         10
100
1000
             Al CONCENTRATION, ng/m3
          FIGURE 14.  Zn Enrichment Factor vs. Al Concentration.
                     68

-------
   100
o
V)
    10
o

o
LU
O

cr

UJ
    i.o
    O.I

\ *
\

•36
•50

w















41
•54
•9
\ -53
\47 27


19 -51 \
s

•24












22"..2,
• •(SFS
OQ. &°
30 -15
26
34



^
S
\
\
X
\
S
\







•7.5
•8 *6
• 20

•14


•10






N
\
\
s
\
s
s
       10                100              1000



           Al CONCENTRATION, ng/m3

         FIGURE 15. Mn Enrichment Factor vs. AT Concentration.
                    69

-------
   100
-J

o
CO
     0
o:
o

o
h-
z
LJ



O
cr
1.0
LJ


-------
   100
-J

o
en
^  10
cr
o


§
LU
O

o:
z
LU
    I.O
    O.I

\
\
\
\
\
V
\
\
\
Ov3*

49 «
37

•
17
36
42*
39














50
vs
\
\
V. •
\
\
27 0 *,X
47« 34
31 .
51* 24 26
53'
54
*

41



















\
25\ 28
. V 55
30 \
\
21 13
22- ••
11 *






















2&3
\20 5.
* 8 \ *
v .
14\ 6
\
\
       10
IOO
IOOO
         Al CONCENTRATION,  ng/m3
        FIGURE 17. Ti Enrichment Factor vs. Al Concentration,
                    71

-------
    100
<
C£
     10
o
LU

O
cr
LU
     0
    O.I
•37
*50
\
Xv
•I8\
•
49



.54
•4O
•51 HU
•1
45 -47
•53
•9
•27
•
29
\ -24
v
\


op
.CO
II*.
34 *55
I5-*13
26 30- '21
•25
•22
\
\


•8
•7
•10 *6
*20 14 "5

\
s
       10
                        100
1000
          Al CONCENTRATION, ng/m3
        FIGURE 18.  Mg Enrichment Factor vs. AT Concentration.
                    72

-------
IO
 E
 o
 c>

 <
                                             .0
10.0
                                  r.jim
         FIGURE 19.  AN/A (log r) vs. log r for 17-20 May 77.
                               73

-------
o
                                                             10.0
        FIGURE 20.  AN/A (log r) vs. log r for  7-9  June  77.
                               74

-------
     10"
     10
ro
i _
 £


 <
     10
     10
     10

      0.01
                 -Itf/
                                   AUGUST l|4-19. 1977
                                   AVERAGE SLOPE •

                                   WITH r = 0.97
                           -3.3
O.I
1.0
10.0
                                r,jim
         FIGURE 21.   AN/A  (log r)  vs.  log r for 14-19 Aug  77.
                               75

-------
ro
i
 o
 £

 <
                                    AVERAGE SLOPE

                                    WITH r=b.99
                                                           10.0
          FIGURE 22.  AN/A  (log r) vs. log r for 26-30 Sept 77.
                               76

-------
10
  0.01
        FIGURE 23.  AV/A (log r)  vs.  log r for 17-20 May 77.
                               77

-------
0.01
                            r,jim
10.0
      FIGURE  24.  AV/A (log  r) vs. log r for  6-9 June 77.
                           73

-------
     icr
K>     r,

I
 o>
 o
     10
     10°
       0.01
O.I
.0
10.0
                                      r,
             FIGURE 25.  AV/A (log r) vs. log r for 14-19 Aug 77.
                                    79

-------
fO
 o
       0.01
10.0
             FIGURE 26.  AV/A (log  r)  vs.  log  r  for  26-30  Sept  77.
                                    80

-------
10
u
E
k_
o>
o
<
\
5 ,o'
10°
AT = 8.7*1
Uh= 4.6 * (
WD=243
_ VDh= 0.38




.9
3.3
*5


«
<
*
•


%
• •<
r








Se
On
Off


•.
.•
•
: no. 20050
5172215
5180253





0.01
O.I
1.0
10.0
 FIGURE 27.  AV/A (log r) vs. log r for 22:15 CDT,  17 May  77  to
             02:53 CDT 18 May 77.
                             81

-------
    icr
0>
o
    io
    10'
    10°
AT= 5.9 *
Uh= 4.5 * 0
WD = 240 ±
\/ - n /H
VDh- 0.41




.4
.2
10

.
«
V
•"
•


?S " *
1



/
•



Set
On
Off


•1
'.
•
no. 20060
5180316
5180618





     O.OI
O.I
I.O
IO.O
                                  r,)im
     FIGURE 28.  AV/A  (log  r)  vs.  log r for 03:16 CDT, 18 May 77 to
                06:18 CDT  18  May  77.
                                  82

-------
ro
 o>
 o
<
     10
    I0
    10

AT =6.7 ±3.6
Uh= 3.6 * 0.3
WD=239 *8
-V0h»0.22_








.
I
«
^B
V
•
*






•
•l» w
•
•
^








»% •




i
Set no. 20070
On
Off





1 •
.

5180820
5181120








      0.01
                           O.I
1.0
                                                                     10.0
                                   r.jim
      FIGURE 29. AV/A  (log  r)  vs.  log  r for  08:20 CDT,  18  May  77  to

                 11:20 CDT  18  May  77.
                                   83

-------
    10"
10

 E
 o
    10
o>
o
    10
AT=9.5 *o
Uh= 1.9 * 0
WD=174 t.
Dh




i.5
5
i6


(
y
•*
.


•
* •
••
\
••





*
•.

Set
On
Off



•

i 	
no. 20080
5181200
5181725





      O.OI
                           O.I
                                                                    IO.O
                                    r.jim
      FIGURE  30.  AV/A (log r) vs. log r for 12:00 CDT, 18 May 77 to

                 17:25 CDT 18 May 77.

-------
ro
 0>
^

<
•^
10
.o2
I01
.0°
	
AT= 11.4 ±;
Uh= 3.8 t C
WD=140±<
-VDh= 0.«




3.7
).3
*o


•
i
V
•
•


f.
• •<
*





AT
•
• 1
•"

Se
Or
Of


•
..

t no. 20090
5181746
f 5182110





      0.01
O.I
1.0
10.0
                                   r.jim
        FIGURE 31.  AV/A (log r) vs. log r for 17:46 CDT, 18 May 77 to

                   to 21:10 CDT 18 May 77.
                                   85

-------
    ICf
ro

 E    2
^  10
to
    I0
    10°
AT= 13 * 2
3 °C
Uh= 37*09 m/s
WD= 212 *
8
VHK= 013 cm/s
- un 	










• •


•
•





*
•
•
* •
• * •
• •_


V








*

'•**
• . ••

Se
it no. 20100
On 5182210
O1







•
•
1 5190120









       O.OI
        FIGURE 32.  AV/A (log r)  vs.  log r for 22:10 CDT, 18 May 77 to
                    01:20 CDT 19  May  77.
                                                                         IO.O
                                     86

-------
    icr
ro
    io
 o>
 o
    10'
    ,0°
AT= 72 * 2
Uh= 27*0
WD= 212*
-VDh=002c




.7 °C
4 m/s
?6
m/s

•
•;
•


• i
./ *•
>



<
i •



Set
On
Off
0 •
•




no 20110
5190144
5190616





       0.01
O.I
1.0
10.0
        FIGURE 33.  AV/A  (log r) vs.  log  r  for 01:44 CDT,  19 May 77 to
                    06:16 CDT 19 May  77.
                                     87

-------
                                       Neutral Drag Coefficient  CD(5m)
                                                   o
                                                   o
 cr>
 O
a
 o
 0)
 3
 a.


 o
 ai
 rt


 cn


 3

 CO
 to

 re
 
-------
CO
ID
                 2.0 3jQ   
-------
u
o
JO
o
                                                                                    UNSTABLE >S0>O °rTs-TA>O
                                                            STABLE 'S0<0 or Ts - TA 
-------
-16
-12
-8
          FIGURE 37.  Cumulative Frequency of Occurrence of AT Over Southern
                     Lake Michigan.

-------
  900
  800
„  700
CJ
   400
^ 300
   200
   100
                                    *
                              _ _y
   500
 I
I

     46   47   48   49   50   51    52   55   54   55   56   57   58
                             TIME  (MINUTES)
         FIGURE 38.  Aerosol  Number Concentration  (cnf3)
                     Measured in Size Range 0.25  < d < 20 \m
                     for  the  Midway Airport to Midlake Flight
                     Leg  Collected on 30 Sept 77  from 06:46 CDT
                     to 06:58 CDT.
                                92.

-------
       SMOOTH  FLOW                      ROUGH FLOW

     TURBULENT LAYER, KD^ to                 TURBULENT LAYER, KD^ KM

      BUFER LAYER, KD ^ AKM                   BUFER LAYER, KD ^
        LAMINAR SUBLAYER
AAAAAAAAAAAAAAAAAA
           FIGURE 39.  Diagramatic Representation of The Air/Water Interface.

-------
2.0  T
I.S  -.
I .0  ..
0.5  ..
0.0
0.20^
0.15..
0.10..
0.05.
                                        -H	1	»	1	1	1	1




           S     N     CH    FE    ME    HL    PB    ZN    P     MN    Tl    CU






                             ELEMENT  DF INTEREST
        FIGURE 40.  Normalized Aerosol  Mass Fraction for All  1977 Sampling.

-------
           2.0
0.212.
                I
tn
uc i .5 .
in
a
1 ,..:
1 ;
c
m
i 0.5 .


0.0 .,
a.isi


0.10-L
\
i
\
\
\ 0.0SJ
\ i
L \ |
'• t
^ \
^ _ M — - — — *• i
.. 	 1 	 1 	 1 	 1 	 — * 	 ,L_
                                                               \
                            N
CR    FE    ME    HL    PB    ZN     P
                                 MN
OJ
                                         ELEMENT OF  INTEREST
                 FIGURE 41.   Normalized Aerosol Mass Fraction for Overlake Transport Samples.

-------
           2.0 ^
o.
a i .5 .
si
in
a
^~
Ui
—
£
d i .0 .
ne
a
c
1C
a:
u.
in
iP
£ 0.S .



0.0
1
0.1S.
'
x^
\
\ !
\ i
\ i
/ \ 0.10^




>
!
V--'" 0 05 J
1
i
]
__ , 	 *- ... ... - A. ,. 	 *_ 	 . *. 	 ... 	 4- 	 J
                                                                  V
                                                                        A
                                                                                      \
                             N
CH    FE    ME    HL
Pfi    ZN
MN
                                                                                           a:
                                         ELEMENT OF  INTEREST
                   FIGURE 42.   Normalized Aerosol Mass Fraction for West Shore Source Samples.

-------
         2.0 T
          1.5 ..
IT)
     si
     d
     a:
c
i
     1.0 ..
          0.S ..
          0.0
                                      0.20-r
                                      0.15.
0.10..
                                      0.05..
                    _l	1	1	1	I-


                     5     N     Cfl    FE   ME    flL    PB    ZN    P     MN    Tl     CU



                                       ELEMENT DF INTEREST
                FIGURE  43.  Normalized Aerosol  Mass Fraction for East Shore  Source Samples.

-------
          2.0 ^
           1.5 ..
           1.0 ..
t-O
C3
1C
          0.5 ..
           0.0
                                      0.20^
                                  CH    FE    ME    BL    PB
                                                           ZN
MN
Tl
                                        ELEMENT DF  INTEREST
                     FIGURE 44.  Normalized Aerosol  Mass  Fraction for  Southeast Shore
                                Source Samples.

-------
    2.0 T
      .S ..
ti
CK
I
i
1.0 ..
     0.5 ..
     0.0
                                 0.15..
0.10..
                                 0.05..
               H	1	1      I	I—




                5     N     Cfl    FE    ME   HL    PB    ZN    P     MN    Tl     CU






                                  ELEMENT DF  INTEREST
           FIGURE 45.   Normalized Aerosol Mass Fraction for Chicago/Gary Source Samples.

-------
                              Table 1.   Listing  of Meteorological  Parameters Measured and
                                        Type  of  Sensors  Used  in  the  Lake Michigan  Study
o
o
Wind Velocity at 5 m height (u, j
Ambient Air Temperature     (T^)
Relative Humidity           (RH)
Vertical Wind               (W)
Wind Direction              (WD)
      Water Surface Temperature   (T  )
      Water Surface Current       (UQ)
      Bulk Water Temperature      (T  )
3-cup anemometer/Weather Measure Grp (WMC)  #W103
YSI thermistor probe/WMC #IS6/TP621  HF-35
Capacitance type/WMC#IS/HM-lll
Gill propeller anemometer/R.  M.  Young Co. #27103-21282
Lightweight vane/WMC#W104
Ship's log of WD, semi-hourly report
Hand-held Infrared Thermometer/Barnes Engineering Co.  #PRT-10
Floating drifter; Optical range  finder
Bucket thermometer

-------
        TABLE 2.  Detection Limits (Ld)  Filter Blank Corrections  (3),

                  and Typical Sample Elemental Concentrations  (C)  in


                  the ICAP-AES Liquid Sample  (all  yg-JT1K




                                                                Percent of Sets
Elements          L.                 3              C               C  >L  .
                   °                                                     d
Al
B
Ba
Ca
Cd
Co
Cu
Fe
Mg
Mn
Mo
Na
Ni
Pb
Ti
V
Zn
50
50
5
700
2
5
6
170
100
5
5
15000
5
20
15
12
60
140
1850
5
1200
2
5
9
380
240
16
6
15000
13
25
19
12
125
810
t
*
8150
*
*
18
1700
1890
110
*
*
*
170
50
*
175
97
77
46
85
31
5
88
92
95
98
48
10
40
97
90
5
98
* No C   value, over 50 percent of sets below L .


t Possible contamination of some data sets.
                                   101

-------
                         BULK WATER
SURFACE
Element
Al
B
Ba
Ca
Cd
Co
Cu
Fe
Mg
Mn
Mo
Na
Ni
Pb
Ti
V
Zn
%sb = (c
* Less than
pH
Tw°C
Overall May June
13 * *
79 12 13
10 11 10
29240 33000 34200
0.5 * *
0.4 * *
2.6 * *
29 54 *
8710 10200 10300
0.8 * *
2.2 * *
4870 4580 4630
3.1 * *
3.5 * *
1.5 * *
3.8 * *
17 15 27
surface/C bulk) - 1.0
3 samples > L ,
8.3 8.2 8.3
N/A 2.7 10.3
August
38
140
10
25400
*
0.9
3.0
25
7700
1.9
2.8
5860
6.6
2.4
1.0
11
17


8.5
21.9
September
*
68
10
29800
0.9
0.2
3.0
24
8590
1.0
2.9
4160
1.8
6.2
2.7
*
15


8.2
17.4
September
*
55
10
30900
1.0
0.6
4.0
37
8870
*
3.0
4330
3.0
8.0
3.0
*
20


N/A
N/A
EIb
*
-0.19
0.00
0.04
0.10
2UQO
0.30
0.50
0.03
*
0.03
0.04
0.70
0.30
0.10
*
0.3




TABLE 3.  Geometric Mean Concentrations  (yg-£~1)  of Certain
          Elements in Lake Michigan Water  at  87°00'W by 42°00'N
                            102

-------
      TABLE 4.  Geometric Mean Concentrations in Air (ng  -  m~3} of
                Certain Elements at 87000'W by 42°00'N  During  1977
                Sampling Periods
Element
Overall
June
August
September
Al
Ca
Cu
Fe
Mg
Mn
Pb
Ti
Zn
Total P
N02/N03 - N
so4 - s
Mass
180
770
6
320
200
20
140
8
55
35
1000
1800
32000
— - if.
470
1900
13
790
420
45
350
13
240
40
2300
3500
66200
190
1500
4
990
220
25
85
10
95
20
1400
1200
21000
90
340
3
190
150
10
70
8
25
30
560
1400
21100
60
550
4
90
210
6
60
7
15
30
720
1000
22700
                                  103

-------
             TABLE 5.  Characterization of Mid-Lake Michigan Aerosol
                       with Respect to Several Chemical Species
     Element
It*
FC*
[rp***
Al
Ca
Cu
Fe
Mg
Mn
Pb
Ti
Zn
Total P
N02/N03 - N
so4 - s
0.56
2.4
0.02
1.0
0.63
0.06
0.44
0.03
0.17
0.11
3.1
5.6
1.2
0.43
3.9
1.6
0.64
2.0
17.3
0.69
6.1
4.2
8.0
27.5
= 1.0
22
170
3.5
23
10
5200
1.4
530
—
— —
  * n   =  % of total aerosol mass
 ** FC  =  fractional degree of association of element with d < 1.0 ym aerosol
*** EF  =  Al-based enrichment factor, using Bowen's (1966) soil composition
                                       104

-------
Ca   .44



Mg   .81   .43



Al   .88   .85   .49



Fe   .90   .79   .91   .44



Mn   .89   .78   .86   .97   .41
Mo   .35   .25   .33   .38   .41    .03



Pb   .71   .69   .69   .72   .78    .63   .47



Ti   .71   .58   .67   .68   .62    .33   .51    .13



Zn   .84   .73   .79   .90   .95    .49   .88    .68   .45



C    .45   .50   .55   .54   .63    .14   .52    .28   .59    .43
 m                                                        ——



u,-  -.13  -.09  -.14  -.10  -.14   -.00  -.12    .29  -.08   -.12    .11
 b                                                              	
AT   .50   .60   .65   .49   .46   .20   .50   .16    .40    .52   -.35    .38



      Ca    Mg    Al    Fe    Mn    Mo    Pb    Ti     Zn     C      u,    AT
             3                                              m      5
      TABLE 6.  Correlation Coefficients  Matrix Among  Trace  Element,

                Total  Aerosol  Mass,  Wind  Speed, and Air/Lake Surface

                Temperature Difference  Using 40 Filter Sets'  Data.
                                  105

-------
TABLE 7a.   Trace Element Analysis  Results  for Data  Sets
           Collected at the Nearshore Site in 1978.
Element
Al
Ba
Ca
Cd
Co
Cu
Fe
Mg
Mn
Mo
Na
Ni
Pb
Ti
V
Zn
	 o
C (ng-nf )
290
12
1800
6
1
24
750
600
60
4
250
10
450
10
8
200
n
0.41
.02
2.5
.009
.001
.03
1.10
.85
.09
e006
.36
.01
.64
.01
.01
.28
FC
1.8
2.0
0U9
14.4
6.9
6.4
3.3
2.2
3.1
6.8
0.9
5.0
12.7
1.9
7U3
4.4
EF
= 1.00
6.2
28
40000
43
300
4U5
27
15
720
12
50
11000
0.5
24
980
                          106

-------

Cu

% Total
Data Set Soluble Recovery
6061*0
60650
60670
60680
60690
70700
70710
70720
7071*0
70750
70760
70770
70780
70790
70800
70810
70820
70830
7081*0
70850
MEAN
a
Excluding
7071+0-70810
MEAN
a
0
5U
1*1
1*5
1*1*
19
111
39
105
1U6
91
U3
62
65
121*
86
62
18
35
36
52.8
36.5
36.2
16.8
0
109
61*
62
77
25
7^
7^
105
1U6
92
1*3
62
65
121*
92
77
90
81*
97
78.1
32.8
69. 1*
30.1
Fe

% Total
Soluble Recovery
1*0
109
86
109
90
96
121*
78
101*
153
75
22
65
6l
ll*l
109
9U
51
7U
61
8U
37.7
8U. 6
2l*.5
1*0
119
86
115
98
96
121*
89
ioi*
153
75
22
65
61
lUl
109
97
81*
93
90
93
31
9U
21-5
Pb

% Total
Soluble Recovery
90
62
55
53
1*8
90
65
39
105
163
96
1*0
97
96
122
101
87
63
U7
66
79.2
31.2
63.8
17-2
95
120
97
105
107
109
102
103
105
163
97
1*0
97
96
122
101
103
97
116
113
ioi*
21
105-6
7-8
Zn

% Total
Soluble Recovery
19
36
78
28
18
57
29
21*
86
135
88
19
52
82
115
88
31*
11
22
20
52
36.7
31.3
18.8
19
129
135
107
81
97
93
91
86
159
92
21*
52
87
115
ioi*
108
102
96
112
9U
33
97-5
29.1
Mn

% Total
Soluble Recovery
0
1*2
1*6
U7
28
1*3
^9
31
83
116
53
19
56
56
163
85
58
25
1*3
19
1*8.1
3U.3
35.9
15.9
0
121*
ioi*
126
83
99
98
110
95
155
73
39
57
63
198
111
95
9U
131
110
98.2
1*1.8
97.8
3U

Wind
Direction
69
91
25
90
105
1*1
1*5
203
215
191
271
272
277
290
257
265
220
variable
variable
variable


TABLE 7b.  Results of Passive Aqueous Extraction of Nearshore Aerosol  Samples.

-------
     TABLE  8.   50  percent  Collection  Efficiency Diameters for Hi-Volume
               Sampler  Integrating  Nephelometer and Active Scattering
               Aerosol  Spectrometer (ASAS)
  Aerosol  Sizing                                    Range of Diameter
    Instrument                                         Sampled, ym
  Hi-Volume Sampler*              <. 0.1 to 10 or 20

  Integrating Nephelometer**      0.1 or 0.2 to 1.0


  ASAS             SR  0           SR  1           SR 2	SR  3
Channel #
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15

0.67-0.87
0.87-1.06
1.06-1.26
1.26-1.45
1.45-1.65
1.65-1.85
1.85-2.04
2.04-2.24
2.24-2.43
2.43-2.63
2.63-2.83
2.83-3.02
3.02-3.20
3.20-3.37
3.37-3.53

0.32-0.35
0. 35-0 „ 38
0.38-0.41
0.41-0.44
0.44-0.47
0.47-0.50
0.50-0.53
0.53-0.56
0.56-0.59
0.59-0.62
0.62-0U65
0.65-0U68
0.68-0.71
0.71-0.74
0.74-0^76

0.17-0.19
0.19-0.21
0.21-0022
0U22-0.24
0.24-0.25
0.25-0.27
0.27-0.28
0.29-0.30
0.30-0.32
0.33-0.33
0.33-0.35
0.35-0.37
0.37-0.38
0.38-0.40
0.40-0.41

0.09-0.11
0.11-0.12
0.12-0.13
0.13-0.14
0.14-0.15
0.15-0.15
0.15-0.16
0.16-0.17
0.17-0.17
0.17-0.18
0.18-0.19
0.19-0.19
0.19-0.20
0.20-0.21
0.21-0.22
 * Based on 50 percent collection efficiency of glass  fiber filter for
   small aerosol  and of Hi-Vol  hood for large aerosol

** Charlson, et al.  (1968)
                                 108

-------
                          Meteorology
Aerosol Mass  (pg/m3

Data Set Run Time (CDT)
_ ,
20050 5/17 22:15-5/18 03:00
20060 5/18 03:00-5/18 06:00
20070 5/18 08:20-5/18 11:20
I
20080 5/18 12:00-5/18 17:25

20090 5/18 17:45-5/18 21:00
20100 5/18 22:00-5/19 01 : 30
20110 5/19 01 :45-5/19 06:15

AT u(5) WD
(°C) (m/s) (° )
8.7 ± 1 .9 4.6 ± 0.3 243 ± 5
5.9+1.4 4.5±0.2 240 ±10
6.7 ± 3.6 3.6 ± 0.3 239 ± 8
9.5±3.5 1.9±0.6 174+46

11:4 + 3.7 3.8 ± 0.3 140 ± 40
13.3 ± 2.1 3.7 ± 0.9 212 + 8
7.2 + 2.9 2.7 + 0.4 212 + 26
ASAS
bscat Hi-Vol Mass
Mass Mass <1 .0 ym
117 /TX 46
125 I 160
226 ' 174
141 132 29

110 29
47 , 19
141 X/ 133

Total
51
169
193
31

34
23
333
TABLE 9.  Sample  Time,  Meteorological Parameters,  and
          Aerosol  Data  for Data Sets 20050  -  20010.

-------


Profile Set
23 June 78
13:40-14:30

v =0.6 cm/s
d
crib site
16 August 77
20:45-21:30

v =1.3 cm/s
d
ship site
28 September 77
17:00-18:15

v . = 0
d
ship site

19 May 77
13:30-16:30
ship site
v°
THEN
19 May 77
17:00-18:30

v , = 0.25 cm/s
d


u(5)
Cm/s)
4.0

±

0.2

7.9

-t-

0.3

1.3

±

0.3


1.6

1
0.5

4.2

±
0.3

Wind
Direc-
tion( °)
45

+

5

355

+

10

305

+

7


105

±
5

105

±
5


AT
(°C)
2.0

+

0.3

2.4

±

0.2

2.8

±

0.3


8.8

±
1.4

8.5

+
0 .5

Ratio of High
to Low Position
Aerosol Counts
1.34 ± 0.19

1 .19 ± 0.17

1.72 ± 0 . 24

1.30 ± 0.18

1 . 30 ± 0.14

1.47 ± 0.13
1 .50 ± 0.18
0.95 ± 0.08

0.93 ± 0.08

0.96 ± 0.06

1.01 t 0.03
0.77 ± 0.11

0.94 ± 0.05
0.79 ± 0.12
0.98 ± 0.05
1.34 ± 0.08

1.09 ± 0.03
1.07 ± 0.06
1.02 ± 0.06

Size Range
(wn)
0.6 < d < 1 .5

0.3 < d < 0.6

0.1 < d < 0.2

0.6 < d < 2.0

0.3 < d < 0.6

0.2 < d < 0.4
0.1 < d < 0.2
0.6 < d < 2.6

0.3 < d < 0.6

0.2 < d < 0.4

0.1 < d < 0.2
0.6 < d < 3.2

0.3 < d < 0.6
0.2 < d < 0.4
0.1 < d < 0.2
0.6 < d < <>.6

0.3 < d < 0.6
0.2 < d < 0.4
0.1 
-------


Profile Set
THEN
19 May 77
20:00-21:15

v = 0.2 cm/s
- THEN -
19 May 77
22:45-24:00

«d = 0.23 cm/s

u{5)
(m/s)

3.3

±
0.6
_
3.9

±
0.6
Wind
Direc-
tion(°)

146

±
15
_ _
198

±
25

AT
(°C)

8.5

-
0.9
.
8.6

-
1.5
Ratio of High
to Low Position
Aerosol Counts

U.8 - 0.55
0.95 r 0.10
1.06 r 0.05
1.12 ± 0.12

0.8 ± 0.26
0.85 r 0.16
0.98 t 0.04
0.99 ± 0.04

Size Range
Um}

0.6 < d < 2.7
0.3 < d < 0.6
0.2 < d < 0.4
0.1 < d < 0.2

0.6 < d c 3.0
0.3 < d < 0.6
0.2 < d < 0.4
0.1 < d < 0.2

RiR
D

0.07




0.05





0.

0.55

"
0.15

0.55
4-

0.15

h] \* (h2V*
? ;5,

0.29

*
0.09

_ _ _




V. /*H
dp d

0
0
0.1 ± 0.04
0.2 ± 0.05

0
0
0
0
TABLE 10.   Continued.

-------
TABLE 110   Binned Data Sets,  Using  AT
           as the Defining  Parameter
AT
Fo
vd














(°C)
(Yr)
(cm-s-1)

AT
Ca
Cu
Fe
Mg
Mn
Pb
Ti
Zn
Total P
N0;/N0o
(L -3
SOJ
Mass
<-0.9
.38
J2

40
140
5
60
100
5
25
5
10
15
1200
1400
1 3500
-0.8 - 0U9 +1
.19
.73
	 O
C (ng-rtf )
90
620
5
210
300
15
60
10
20
35
3800
3700
27900
.0 - 207
J5
.51

100
710
5
190
200
10
80
15
40
35
3000
5000
22500
2.8 - 8.2
.18
.39

630
2700
20
1400
520
80
600
20
300
60
7900
11400
58900
> 8.3
.02
J5

450
1200
5
420
340
20
190
10
80
20
9500
10300
68000
                    112

-------
TABLE 12.  Binned Data Sets,  using  u5
           as the Defining Parameter
IF5 (m-s"1)
Fo (Yr)
Vj (cm-s~ )

Al
Ca
Cu
Fe
Mg
Mn
Pb
Ti
Zn
Total P
NO;/ NO:
L 6
so=
Mass
< 2.0
.04
0.0

200
790
10
300
200
20
140
6
50
20
6950

4850
32900
2.1 - 3.5
.16
.23
"C (ng-m"3)
140
780
5
360
460
10
60
0
30
20
4300

5800
28200
3.6 - 5.2
.15
.47

300
2000
10
550
380
40
230
10
90
45
7400

7900
37000
5.3 - 6.9
.20
.76

110
540
5
330
240
20
120
10
40
30
3300

4050
30800
> 7.0
.41
1.17

50
320
2
60
70
5
70
10
20
25
1800

4100
18700
                113

-------
TABLE 13.  Binned Data Sets, Using Source Region
           as the Defining Parameter.
SOURCE AREA
F0 (YD
v. (cm-s~ )

Al
Ca
Cu
Fe
Mg
Mn
Pb
Ti
Zn
Total P
NO^/NO'-N
SO=-S
Mass
OVERLAKE
U22
U50

25
70
2
50
80
5
50
6
10
10
3750
3650
28600
WEST
SHORE
.16
.72
C (ng-nf 3)
55
375
2
160
180
5
55
10
15
30
2400
1100
17300
EAST
SHORE
.09
.62

105
255
2
208
120
8
42
10
27
40
1700
2000
20100
S.E.
SHORE
.35
.50

150
700
5
250
215
15
120
10
40
50
8750
10900
43200
CHIC./
GARY
.20
.42

575
1900
15
980
400
50
325
21
260
40
6200
9650
50300
                     114

-------
        TABLE 14.  Annual Loadings of Certain Trace Elements to the
                                                      o        _i
                   Southern Basin of Lake Michigan (10  kg - yr~ )0
ELEMENT
Al
Ca
Cu
Fe
Mg
Mn
Pb
Ti
Zn
Total P
WO™ / Mfi~ M
iiv**)/ llv/rt 1 1
DRY DEPOSITION
Minimum Mean
300
1000
{10}
600
400
30
300
{15}
100
75
2000
650
2900
{20}
1200
730
60
520
{30}
200
150
4350
% Total
± 2
< 1
{6}
1 20
< 1
>_ 10
_> 60
{10}
>_ 30


Precipita-
tion (2)
560

20
950

50
90
60
50
100 (5)

Run-Off
(3)
1300
490000
140 (4)
1450
1 34000
250
100
100
180 (4)
-3000 (6)

   so4-s
3500
7850
Brackets' { } enclosing results indicate caution should be used in  interpreta-
tion — see Section 5.
(1)  This work.
(2)  Gatz, 1975b.
(3)  Winchester and Nifong, 1971.
(4)  Robbins, et al., 1972U
(5)  Murphy and Doskey, 1976.
(6)  International  Joint Commission on The Great Lakes, 1976.
                                    115

-------
FACTOR 1



"LABEL"




PRIMARY
MEMBERS


SECONDARY
MEMBERS

VARIATION
Transport of soil ,
lake and anthropo-
genically derived
aerosol enhanced
by temperature
stability and
Chicago/Gary
Source Region
CA, MG, AL,
FE, MN, PB,
TI, ZN

P, N02/N03, S04
C/G
AT
41
2
TI ship contaminant,
other elements' cone.
not significantly
contributed to by
ship effluent.



U5


-TI

8
3

Stand Alone
Wind
Direction Factor




C/G
SES
ES

-L
.JS
9
4

Fine
Particulate
with
Nutrients



P, N00/N0,
SO,
Mass
MN
PB
ZN
20
5

Soil and Lake
Derived Coarse
Particles con-
tributing
CA, MG, AL


C/G
WS
L

CA
MG
AL
10
6

Temp. Stab.
brings out
reactive and
coarse par-
ti cul ate


N02/N03
AT

MG
AL
so4
MASS
12
TABLE 15.   Factor Analysis  Results  and   Interpretation of  Factors.

-------
Filter
Set#
5
6
7
8
9
10
11
13
14
15
17
18
19
20
21
22
24
25
26
27
28
29
30
31
33
34
35
36
37
40
41
45
47
49
50
51
53
54
55
Source *
Region
nC/G
nC/G
nC/G
-
L
nC/G
nC/G
nSES
nSES or nC/G
-
L
L
SES
nC/G
nC/G
-
ES
ES
ES
ES
ES or nSES
SES
nC/G
-
L
US
L
L
-
WS or nC/G
WS or C/G
WS
WS
L
L
SES
SES
SES
nSES
Time to Reach
Shore
(hours )
2.7
3.0
2.8
-
3.4
3.2
3.5
2.5
3.5
-
6.5
9.9
3.5
2.8
6.1
-
4.4
3.8
2.9
5.3
1.5
1.2
2.6
-
4.1
5.4
6.4
6.6
-
1.7
3.0
2.1
2.0
4.0
3.6
2.9
1.9
1.5
1.8
Bulk
Ricnardson
Number, R-n
0.05
0.04
0.06
0.32
0.09
0.12
0.12
0.07
0.13
0.11
0.03
0.05
0.03
0.01
0.18
0.09
0.01
0.01
0.013
0.03
0.012
0.01
0.03
0.022
0.01
0.012
0.01
0.014
0.016
0.018
0.023
0.008
0.005
0.007
0.12
0.033
0.022
0.015
0.013
Air Mass
Type
.
-
mT
-
mT
mT
mT
mT
-
-
cP
cP
cP
cP
cP
-
_
-
_
_
-
mT
mT
-
cP
cP
cP
cP
-
_
_
cP
cP
cP
cP
cP
cP
_
-
  * C/G is  Chicago/Gary,  WS is West Shore,  L is Lake, ES  is  East Shore,  SES  is
  Southeast Shore and n refers to nearshore source domination  (vs. longer  ranqe
  source domination, for  which no prefix  is stated)                        9
TABLE  16.  Meteorological  Influences Upon  the 39  Filter Set Data  Base,
                                    117

-------
     Table 17.  Factor Analysis of Nineteen Filter
                Sets'  Data with RiB  > 0.03.

                 UNCORRELATED FACTORS
Variable
Al
Ca
Fe
Mg
Mn
Pb
Ti
Zn
Mass
U5
AT
1
(.82)
(.86)
(.93)
(.76)
(.91)
(.52)
(.95)
(-79)
.17
.24
.16
2
(,,54)
.34
.24
(,,59)
.22
,27
.02
.15
(.93)
.20
(.98)
3
.17
.19
.21
.07
.12
.09
.28
.18
.18
(.95)
(.09)
4
.09
.33
.17
.28
.34
(.81)
JO
(.56)
.28
.08
.02
% TOTAL
VARIATION   52.3
23.5
9.9
12.3
     Table 18.   Factor Analysis  of Twenty  Filter
                Sets'  Data with  Rig <  0.03.

                 UNCORRELATED FACTORS
Variable
Al
Ca
Fe
Mg
Mn
Pb
Ti
Zn
Mass
U5
AT
% TOTAL
VARIATION
1
(.50)
.22
(.76)
.22
(.87)
(.92)
.38
(.95)
(.99)
.08
.08

36.4
2
(.82)
(.94)
(.57)
.24
.43
.18
(.58)
.31
.06
.19
.27

30.2
3
.05
-.06
.08
.01
-.02
.07
(.67)
-.01
.01
(-.97)
.19

12.1
4
.23
.22
.28
(.92)
.24
.08
.19
.05
.09
.04
.22

9.9
5
.17
.15
.08
.22
.01
.34
.22
.03
-.06
-.11
(.91)

9.4
                         118

-------
TABLE 19o  Factor Analysis of Fifteen Filter Sets'
           Data with Travel Time to Shore ^3.5 hrs,

               UNCORRELATED FACTORS
Variable
Al
Ca
Fe
Mg
Mn
Pb
Zn
Mass
WS
T
% TOTAL
VARIATION
1
,35
(.84)
.30
(.63)
(.63)
(088)
(.92)
(.97)
.12
(.70)
43.9
2
007
.26
oil
.12
U40
.04
.12
U08
(.99)
.32
13,5
3
(.94)
048
(.95)
(.77)
(-67)
.48
.37
»22
.11
(.64)
38.6
                       119

-------
TABLE 20»  Factor Analysis of Seventeen Filter Sets'
           Data with Travel Time to Shore <_3»0 hrs,

               UNCORRELATED FACTORS
Variable
Al
Ca
Fe
Mg
Mn
Pb
Zn
Mass
WS
T
% TOTAL
VARIATION
1
(.94)
(.96)
.98
(.96)
,98
o95
.97
(.99)
.19
(.94)
85.3
2
.25
,30
J8
.21
J9
.31
.23
,08
(.98)
.37
13.1
                        120

-------
Element
Al
Ca
Mg
Fe
Mn
Ti
Zn
Pb
Average
Soil Cone.*
(ppm)
7100U
13700
6000
38000
850
4700
50
10
% Soil
Deri ved
100
3
6
24
11
(134)
0.2
0.02
Average
Bulk Water
Cone.**
(ppb)
38
29900
8730
46
2.6
2.4
18
7.5
% Lake
Derived***
(.08)
97
94
0.4 to 10
0.4 to 10
(10)
1.0 to 25
0.2 to 4.5
%
Non-natural
-
-
-
> 65
>. 80
-
>_ 75
> 95
  *  Bowen, 1966.
 **  Grand average of more than 40 bulk water samples.
***  Assumes a maximum of 25 X enrichment of element
     of interest over that in bulk water.
      TABLE 21.   Percentages of Soil, Lake and Non-Naturally Derived Mid-Lake

                 Michigan Atmospheric Surface Layer Metal Concentrations.

-------
                                REFERENCES


Andren, A.Wo, AoW. Elzerman, and D.E.  Armstrong,  1975,  Proc.  1st Spec.  Symp.  on
   Atmos. Contribution to the Chemistry of Lake Waters.,   IAGLR,  101-110.

Batchelor, G.K0, 1967, An Introduction to Fluid Dynamics.   Cambridge Univu
   Press, Cambridge, Mass., 615 pp.

Binkowski, F0SU, 1979, Atmos. Env..  13, 247-253„

Blackader, A.K., 1975, Experiments With Simplified Second-Moment Approximations
   for Use in Regional Scale Models:   Second  Annual  Rep.,  SRG; CAES, Penn.
   St. Univ.

Blifford, I.H. and G.O. Meeker, 1967,  Atmos.  Env.. 1, 147-157.

Bloch, M.R., D. Kaplan, V.  Kertes  and  J.  Schnerb,  1966,  Nature,  209. 802-803,,

Bowen, H.J.M., 1966, Trace  Elements  in Biochemistry.  Academic Press, N0Y.

Brtko, W.J. and R.L. Kabel, 1976,  Water,  Air. Soil  Poll.,  6.,  71-79.

Businger, J»A., J.C0 Wyngaard,  Y.  Izumi and E.F.  Bradley,  1971,  J. Atmos, Sci\,
   28, 181-189,

Cadle, R.D., 1975, The Measurement of  Airborne Particles.   Wiley, N.Y.,  p0  81  ff.

Chamberlain, A.C., 1968, Quart. J0 of  the Royal Meteor  Soc..  94,  318-324.

Dave*, M., D. Dolske and H0  Sievering,  1979, Atmospheric  Aerosol  Mass Concen-
   tration and Scattering Coefficient  in  a Midlake Region:  Accepted for
   publication in Atmos.. Env.

Davidson, C.I. and S.K0 Friedlander, 1978, J. Geophys,  Res..  83^,  2343-23520

Deacon and Webb, 1962, in M0N0  Hill  (ed.) The Sea,  Interscience,  N.Y.,  pp.  43-87.

Deardorff, J., 1968, J. Geophysical  Res0, 93^, 2549-2554.

Delumyea, R0G<, and R.U Petel,  1977, Atmospheric  Inputs  of Phosphorus to
   Southern Lake Huron. April-October  19750  EPA-600/3-77/038.

Donelan, M.A., 1977, Manuscript Rept.  Series  #43  Symp. On  Modeling of Transport
   Mechanisms in Oceans and Lakesa  Available from the Marine Sciences"Directorate,
   Deptu of the Environment, Ottawa, Canada.
                                    122

-------
Donelan, M.A., K.N.  Birch and D.C.  Beesley,  1974,  Proc.  17th  Conf.  Great Lakes
   Res. Int.  Great Lakes Assoc.,  369-388.

Dyer, A. and  B.B.  Hicks, 1970, Quart.  J.  of  the Royal  Meteor.  Soc.» 715-721.

Eisenreich, SUT.,  P.J.  Enroling and  A.M.  Beeton, 1977,  J.  Great Lakes Res.,  3_,
   291-304.

Eisenreich, S.T.,  A.W.  Elzerman and D.E.  Armstrong,  1978, Env. Sci. and Tech.,
   12, 413-417.

Elzerman, A.U., 1976, Surface Microlayer- Microcontainment Interactions in
   Freshwater Lakes.  Ph.D.  Thesis, U. of Wisconsin.

Elzerman, A.W., and D.E. Armstrong, 1979,  Limnol.  and  Oceanog.. 8_,  133-139.

Friedlander,  S.K., 1977, Smoke, Dust and Haze:   Fundamentals  of Aerosol  Behavior,
   Wiley-Interscience,  New York,  N.Y., 317 pp.

Gaarenstroom, P.O.,  S.P. Perone and J.L.  Moyers, 1977, Env. Sci. and Tech., J_l_,
   795-800.

Garland, J.A., D.  Atkins, C. Readings  and S. Caughey,  1974, Atmos.  Env., 8, 75-79.

Garrett, W.D., 1965, Limnol. Oceanog., JJ3, 602-605.

Gartrell, G., Jr.  and S.K. Freidlander,  1975,  Atmos.  Env.. 9.,  279-288.

Gatz, D.F., 1975a, Atmos. Env., 9_,  1-18.

Gatz, D.F., 1975b, Water, Air and Soil Poll.,  5_, 239-251.

Gatz, D.F., 1978,  J. Appl. Meteor,  V7, 600-608.

Gillette, D.A. 1972a, J. Appl. Meteor.,  JJ9,  1977-1985.

Gillette, D.A. 1972b, Atmos. Env.,  6_,  451-462.

Gillette, D.A. and J.W. Winchester, 1972,  Atmos. Env., (5, 443-450.

Hall, C.D., 1954,  Bull. Amer. Meteor.  Soc..  35>, 105-107.

Harmon, H.H., 1967,  Modern Factor Analyses,  2nd Ed.,  U.  of Chicago  Press, 389  pp.

Hess, G.D. and B.B.  Hicks, 1975,  Interagency Comm. on  Marine Science and Engin-
   eering Report:   2nd Conf. on the Great Lakes, 238-247.

Hicks, B.B.,  1972, Boundary Layer Meteor., 3,  201-213.
                                    123

-------
Holzworth, G.C., 1972, Mixing Height Wind Speeds  and Potential  for Air Pollution
   in the Contiguous United States.   Office of Air Programs  Publ.  No.  AP-101.
   United States Environmental  Protection Agency,  Washington,  D.C.

Hopke, P.K., 1976, J. Environ.  Sci.  Health, 6^ 367-383.

Hunt, J.C.R., 1973, J. Fluid Mechanics,  61_, 625-635.

Hunt, J.C.R. and P.J. Mulhearn, 1973, J.  Fluid Mechanics,  61_,  245-254.

International Joint Commission, 1976, Hater Quality Board  Report,  App. B.,
   Surveillance Subcom. Rep.

International Joint Commission, 1973, Environmental  Management Strategy for
   the Great Lakes System., Final  PLUARG  Report IJC, Windsor,  Ont.

Irwin, J.S., 1979, Atmos.   Env., 13, 191-194.

Izumi, Y. and J.S. Caughey, 1976,  Minnesota 1973  Atmospheric Boundary  Layer
   Experimen t Data Report.   Air Force Cambridge Laboratory,  AFCRL-TR-76-OQ38.

Jirka, A., M.J. Carter, D.  May, and  F.  Fuller.  U1tra-Mi cro  Semi-Automatic
   Method for the Simultaneous  Detection  of Total  Kjeldahl Nitrate in  Waste Water.
   USEPA Central Regional  Laboratory, Chicago, 111.

Kondo, J., 1975, Boundary  Layer Meteor..  ^, 91-112.

Kondo, J., Y. Fujinawa and G. Naito, 1973, J.  Phys.  Oceanog.,  3^,  197-209.

Kraft, R.L., 1977, Predictions  of  Mass  Transfer  to Air-Land and Air-Water
   Interfaces.  Penn. St.  Univ., CAES Publ. #472-77.

Kramer, J.R., June 1976, "Atmospheric Chemical Loading  of  the Great Lakes"
   Annual Meeting of Great Lakes Chapter, AICHE,  Northwestern Univ.

Lyons, W.A., 1972, J. Appl. Meteor., Vl_,  1259-1270.

Lyons, W.A. and L.E. Olsson, 1973, Monthly Weather Review, 101, 387-403.

Lyons, W.A., C.S. Keen and R.A. Northouse, 1974,  Symposium on Atmospheric
   Diffusion and Air Pollution, Sept. 9-13, 1974.   American  Meteorological
   Society, Boston, Mass., 273-280.

Meszaros, A., 1977, Atmos.  Env., V]_, 1075-1081.

Montieth, J.L., 1973, Principles of  Environ. Physics.  Amer. Elsevier  Publ. Co.,
   New York, N.Y., 241 pp.

Moyers, J.L., R.A. Duce and G.L. Hoffman, 1972, Atmos.  Env., 6^, 551-556.


                                     124

-------
Murphy, T.J. and P.V.  Doskey,  1976,  J.  Great Lakes  Res., _2,  60-70.

Nageb, 1978, personal  communication.

National Oceanic and Atmospheric Administration  (NOAA),  1975,  Summary  of
   Synoptic Meteorological  Observations for Great Lakes  Areas,~Vo1.  3,
   Lake Michigan.Nat'l  Climatic Center,  Asheville,  N.C.

Owen, P.R. and W.R. Thomson, 1963, J.  Fluid Mechanics, ]5_,  321-334.

Owen, R.M., P.A. Meyers,  and J.E. Mackin,  1979,  Geophy.  Res. Letters,  6.,  147-150.

Panofsky, H., 1963, Quart.  J.  Roy. Meteor. Soc., 89_,  85.

Pena, J.A., J.M. Norman and D.W. Thomson,  J. of  the Air  Poll.  Control  Assoc..
   27., 337-341.

Priestly, C., 1959, Turbulent Transfer in  the Lower Atmosphere,  U.  of  Chicago
   Press, 130 pp.
    , K., 1976, The Chemical Composition of the Atmospheric Aerosol.  Grad.
    chool of Oceanography, Univ. of Rhode Island, Kingston, R.I.,  265,  pp.
Rahn
   School
Robbins, J.A., 1972, Proc. 15th Conf. Great Lakes Res.  Inter.  Assoc.  Great
   Lakes Res., 270-290.

Robbins, J.A. and D.N. Edgington, 1975, Proc. 2nd ICMSE Conf.  on Great Lakes.
   Interagency Committee on Marine Science and Engineering, Argonne, 111.,
   378-390.

Raynor, G., P. Michael, R. Brown, and S. SethuRamon (1974) Symp. on Atmos.
   Diff. and Air Poll., Santa Barbara, Cal., Sept. 9-13, pp. 289-294.

Rummel, R.O., 1970, Applied Factor Analyses, Northwestern Univ. Press, 404  pp.

Schlichting, H., 1968, Boundary-Layer Theory, 6th ed., McGraw-Hill, New York,
   747 pp.

Schmict, J., 1977, Selected Metals in Air Particulates Over Lake Michigan.
   M.S. Thesis, U. Wise. Madison.

Sehmel, G.A., 1971, J. of Colloid and Interface Science. 37, 891-898.

Sehmel, G.A. and S.L. Sutter, 1974, Pacific Nirl Lab. Annual Rep, for 1971,
   Vol. II. Part 1.  Atmos. Sciences, Battelle NW Labs, Richland, Wa.  99352.

Sehmel, G.A. and S.L. Sutter, 1975, J. Rech. Atm., III. 911-918.

SethuRaman, S., 1976, Monthly Weather Rev., 104. 1040-1045.


                                     125

-------
SethuRaman, S. and J. Tichler,  1977,  J.  Appl.  Meteor.,  1_6,  455-461.

Sheperd, J.G., 1974, Atmns.  Env.,  8,  69-73.

Sheppard, P.A., 1963, Abst.  in  Proc.  135th Assembly,  IUGG-IAMAP,  Berkeley,  Ca.,
   pp. 117-121.

Sievering, H., 1976, Hater.  Air and Soil  Poll..  5.,  343-351.

Sievering, H., 1979, Proc.  4th  Symp.  Turbulence,  Diffusion  and  Air Pollution.
   Reno, Nev., 518-521.

Sievering, H. and C. Williams,  1975,  Potential  Loading  of Southern Lake  Michigan
   in Proc. of^Workshop  on  Atmospheric Transport  and  Removal  Processes0   2nd
   ICMSE Conf. on Great  Lakes,  Argonne,  111.,  pp. 265-275.

Sievering, H., M. Dave",  P.G. McCoy, K. Walther,  1978, Env.  Sci. and  Tech.,
   1_2, 1435-1437.

Sievering, H., M. Dave",  D.  Dolske, P.  McCoy,  1979a, Trace Element. Concentrations
   Over Mid-Lake Michigan as a  Function  of Meteorology  and  Source Region:
   Accepted for publication  in  Atmos.  Env.

Sievering, H., M. Dave,  D.A. Dolske,  H.A. Roberts,  1979b, Deposition and Trans-
   formation of 504 During  Stable  Atmosphere  Transport  Over Lake  Michigan:
   Accepted for publication  in  Atmos.  Env.

Skibin, D., 1973, Hater, Air and Soil  Poll..  2.,  405-407.

Slinn, W.G.N., 1975, Atmos.  Env.,  9^,  763-765.

Slinn, W.G.N., L. Hasse, B.B. Hicks,  A.W. Hogan,  D. Lai, P.S. Liss,  K.O. Munnich,
   G.A. Sehmel and 0. Vittori,  1978,  Atmos.  Env., VZ_, 2055-2087.

Stolzenburg, T.R. and A.W.  Andren, 1979,  Spatial  and  Temporal Concentration
   Variations of Certain Aerosol Constituents  in  a  Mixed Urban-Rural Watershed.
   22nd Conf. on Great Lakes Res.. May,  1979,  Rochester, N.Y.

Stulov, L.D., F.I. Murashkevich and N. Fuchs,  1978, J.  Aerosol  Sci., 9_t  1-7.

Sutton, N., 1977, Lake Michigan Lab Manual -  unpublished data.

Taneda, S., 1956, J. Phys.  Soc. Japan, jj_, 302-309.

Technicon Instruments Corp., Automated Methyl-Thymol  Blue Auto-Analysis  Procedure
   Method #CRL453.  Technicon Instruments Corporation,  Chicago,  111.

Torrey, M., 1976, Chemistry of  Lake Michigan;  Vol.  3  of Environmental Status of
   the Lake Michigan Region.ANL/ES-40,  Argonne  Nat!.  Labs, Argonne, 111.


                                    126

-------
Twomey, S., 1977, Atmospheric Aerosols.  Elsevier Publ.  Co., Amsterdam,  302 pp.

USEPA, 1974, Methods for Chemical Analysis of Mater and  Waste.   USEPA Office
   of Technology Transfer, Washington, D.C., 78-144 (metals), 207-214 (N02/N03),

Vugts and Businger, 1977, Boundary Layer Meteor., 1J,  295-305.

Wieringa, J., 1974, Comparison of Three Methods for Determining Strong Wind
   Stress Over Lake Fleva, Boundary Layer Meteor., _7_»  3-19.

Winchester, J.W. and G.D. Nifong, 1971, Water, Air and Soil  Poll.,  50-64.

Windom, H.L. and R.A. Duce, 1976, Marine Pollutant Transfer, Lexington Books,
   Lexington, Mass., 391 pp.

Whelpdale, D.M. and R.W. Shaw, 1975, Tell us. 26_, 196-199.

Wu, J., 1969, J. Geophys. Res.. 74, 444-455.

Wu, J., 1972, Phys. Fluids. 15, 741-747.
                                     127

-------
                                APPENDIX A

                     FACTOR ANALYSIS - A TECHNIQUE FOR
                      INTERPRETING ENVIRONMENTAL DATA
INTRODUCTION

     Factor analysis, initially developed to  explain  intelligence  test scores,
is widely used in the social  sciences.   There have been a  few and  limited appli-
cations in the environmental  field.   Gaarenstroom (1977) has  applied both
pattern recognition and factor analysis  techniques to characterize atmospheric
particulate composition.  Gatz (1978) has used factor analysis in  the identifi-
cation of aerosol sources using the  large METROMEX project data baseu  Hopke
(1976) has used the technique to interpret chemical  and physical analyses of
lake sediments.  Finally, Blifford (1967) has used the National Air Sampling
Network data to produce a pollution  model with the use of  factor analysis.

     All efforts thus far in  applying factor  analysis have been "results"
rather than "user" oriented.   The primary aim of this paper is to  introduce
factor analysis in a way that will help  the researcher understand  the scope and
dimension of this technique.   The explanation is conveyed  through  an application
to an ongoing research project.  Several  conclusions  are drawn using the factor
analysis results.  The technique is  applied to the data base  to exemplify the
procedures one must follow in order  to perform a factor analysis on any data
base.  The same data base has been considered in a "results"  oriented approach
to interpret trace metal and  meteorological data (Sievering et a1„. 1979a)0

     The basic mathematical  concepts are presented in conjunction  with examples
to clarify understanding,,  The data  base used here is part of a study to deter-
mine atmospheric dry deposition to Lake  Michigan.  The data consist of several
trace metal, nutrient, micro- and meso-meteorological, and total suspended
particulate mass parameters.   The trace  symbols used  in the tables are self-
explanatory.  The symbols P,  N03, and $04 stand for total  phosphorus, nitrate,
and sulfate respectively.  MASS represents the total  suspended particulate mass0
The symbols WS, WD1, WD2, WD3 and DELT represent various meteorological parameters,
WS stands for the wind speed  monitored at a 5 m height above  the water surface.
The WD1, WD2, and WD3 symbols stand  for  three wind direction  sectors.  Figure A-l
shows the wind direction sectors along with the ship  location with respect to
Lake Michigan.  DELT is a measure of the temperature  stability.  It is the air
temperature at the 5 m height above  the  water surface less the water surface
temperature.
                                    128

-------
THE METHOD OF FACTOR ANALYSIS

     Factor analysis is a technique which provides a way of recognizing and
aiding in the interpretation of the underlying patterns of interrelationships
in any multiple-parameter data base»  It differentiates the separate patterns
of relationships in the data matrix by factoring the matrix into its basic
dimensions,,  Aside from ascertaining the relationships, it condenses large sets
of data into uncorrelated sets of factors„   These factors represent the total
variability in the data and may be interpreted in place of the data matrix.
These factors can then be labeled or classified with a physical  significance
attached to the group of variables that constitute that factor,,   Each factor
or characteristic is assigned a weighting that shows to what degree a factor
represents the variation in the data base0   This weighting may be used to quan-
titatively compare one factor with another„

     The basic outline of a factor analysis  is shown in Figure A-2.  The following
sections are numbered with Roman numerals corresponding to appropriate blocks of
the flow chart,,

                          Block I - The Data Matrix

     One starts with a data matrix, Dm, n,  with m being the number of data points
or sets and n the number of variables or parameters.

                     Block II - Standardize  Data Matrix

     The first step in any statistical analysis of non-compatible variables is to
standardize the data matrix*  By standardizing the data matrix one eliminates
scale differences between measured variables,,  Equation (1) is used to standardize
the data.


                                I.  =  Xl" ~  X               0)
                                          a

                   where, X.j = ith data value

                          X  = average of all data values

                          a  = standard deviation

                          I- = standardized  value of Xi

     The standardized data values are then  used to calculate a matrix of correla-
tion coefficients,,
                                    129

-------
                 Block III - Correlation Coefficient Matrix

     The coefficient of correlation expresses  the degree  of relationship  between
any two variables.  The correlation coefficient matrix of the  40 data  sets,  each
with 17 variables from the Lake Michigan data  base,  is shown in Table  A-l.   The
principal diagonal contains the squared multiple correlation coefficient,.  This
number times 100 measures the percentage of linear variation that can  be  produced
(predicted, generated, explained) for one variable from the others.   The  squared
multiple correlation (SMC) for MN is 0.49.  This indicates that 49 percent of
the linear variation in MN is dependent on the other 16 variables.  WD1,  WD2,
and WD3 are fairly linearly independent of the other variables (Rummel,  1970).

                  Block IV - Does One Do A Factor Analysis?

     One must now make a decision whether or not to proceed with a factor analysis.
There are three major considerations which influence that decision.   First,  one
must have approximately twice as many data points as there are variables.,  There
are, however, certain exceptions, discussed in the section on  the Factor  Score
Matrix, for which one may use as few data points as there are  variables.   Second,
the squared multiple correlation of about half the variables should  be approxi-
mately 0.40 or greater„  This is a good indication of linearity in the data,,
Finally, since the factor analysis routine only detects interrelationships  in the
correlation matrix, if one can decipher all the possible  relationships directly
from the correlation matrix, the need to do a  factor analysis  does not arise.
Based on these three points, one should be able to decide whether or not  to
perform a factor analysis,,

                       Block V - Common Factor Matrix

     If one decides to go ahead with the factor analysis, the  next step  is  to
calculate the common factor matrix.  The correlation matrix is used to extract
initial factors.  The common factor model and  the principal axes technique  is
used to calculate the factors.  Later, the results of this factor matrix  are
used to rotate the factors,.  To avoid confusion, this initial  factor matrix  will
be referred to as the "unrotated factor matrix,,"  Equation (2) shows the  common
factor model.  The basic assumption is that each variable can  be described  as a
linear combination of factors»

              V]  =  an F1 +a12 F2+ • • • + «lp Fp + O] u F]u


              V2  =  a21 F] + a22 F2 + • • •   a2p Fp + a^ F2u
              V   =anF,+a^F0+*'«   a   F+a   F
               n      nl  1    n2  2            np  p    nu  nu

              where, F   = pth common factor

                                    130

-------
                     a   = a loading measuring the contribution of F

                           to the common variance of V

                     p   = number of common factors (p.. j< n)

                     n   = number of variables

                     V   = variable n


                     Fnu = a uniclue factor contributing to the unique

                           variation of V
                     anu = a

     The aim here is to solve for the anp'su  The factors (Fp's) can then be
determined*  To do this, the unique factors (Fnu's) must first be omitted.  The
first stage of the principal axes method involves the selection of the first
factor coefficients (an-|'s) in order to make the sum of the contributions to that
factor a maximum,,  This sum is given by:


                   Sl   =  al

     Using the method of Lagrange multipliers, one may determine the values of
the ani's.  This may be continued until all the anp's are obtained.  The first
common factor accounts for the most general pattern of relationship in the
data; the second factor delineates the next most general pattern that is uncor-
related with the first; the third represents the third most general pattern
uncorrelated with the first two; and so on (Harman, 1967).

     The Lake Michigan data base is used to obtain the common factor loading
matrix shown in Table A-2U  The number of factors can be as many as the number
of variables, but not all factor loadings are calculated,,

                  Block VI - Number of Significant Factors

     The number of significant factors are considered to be the factors having an
eigenvalue of greater than or equal to CL6o  They are the number of uncorrelated
patterns of relationships between the variables,,  In Table A-2, the first six
factors are considered to be significant.

                  Block VII - Interpretation of Communal ity

     The factor loadings, a's, measure which variables are involved in what factor
and to what degree.  They are the "correlation coefficients" between the variables
and each factor,,  The communal ity of each variable, or the measure of the common

                                    131

-------
variance of a variable, is the sum of the squared loadings  for that variable,,   By
subtracting the communality from one, the uniqueness  of a variable is determined,,
For CA, the communality indicates that 92% of its variation has been accounted for.
The remaining 8% is either unique and/or is explained by the remaining nine factors
which have not been extracted.  The eight factors in  Table  A-2 account for 96%  of
the variation in the data base.

                     Block VIII - Rotate Common Factors

     Since most of the variability in the data base is accounted for by the first
factor and the amount of variability accounted decreases in descending order from
the first factor to the last, it is difficult to interpret  the factors„  Using
factor one, Fl, and factor two, F2, of Table A-2 as the x and y axes respectively,
the common factor loadings may be plotted for all the variables.

     This graphical representation is shown in Figure A-30   As one can see,
several clusters are formed,and if the axes can be orthogonally rotated, the
variance on a factor can be maximized.  This is done  simultaneously for all six
factors that are to be rotated.  The varimax criterion is used to orthogonally
transform (rotate) the common matrix.  The varimax criterion is a function of
the variation of the column of factor loadings.  As there are more high and low
loadings on a factor, the variance of the squared factor loadings is larger.
Therefore, an orthogonal rotation can be computed by  maximizing the variance of
the squared factor loading.  By rotating the common factor space, the adequacy
of the solution is not affected.  Once the rotated factor matrix is calculated,
one may proceed to interpret factors and the variables constituting each factor.

     The rotated factor matrix, as presented in Table A-3, is the most important
since it is used for interpreting the factor loadings and identifying the factors.
The factor loadings convey the same meaning as in the unrotated factor matrix.
At the bottom of the table the percent variation explained by each factor is shown.,

                    Block IX - Interpret Factor Loadings

     As a rule of thumb, absolute loadings of less than 0.5 are not considered to
be a member of a particular factor, though they may be considered as subordinate
or secondary members.  For example, in Table  A-3, factor two is made up of WS as
a primary member and TI as a secondary inverse member.  The various primary and
secondary members constituting each of the six factors are shown in Table  A-4.

     The "labels" attached to the six factors in Table A-4 are based on the
combination of primary and secondary members making up that factor.  This  is
where the researcher needs to draw on his experience to properly label the factors.

     In factor one, the trace metals CA, MG, AL, FE, MM, PB, TI, and AN are
primary members; WD1, DELT, and the nutrients P, N03, and 804 are secondary
members.  The  trace netals and nutrients represent anthropogenic, soil, and lake-
derived sources.  The presence of WD1 indicates that these sources are from the
Chicago and Gary wind direction sector.  DELT being a secondary member shows that

                                     132

-------
temperature stability enhances the transport of aerosol over the lake's surface
and to the midlake region.  Based on this interpretation, factor one is then
appropriately labeled.

     The two members of factor two form a very interesting combination.  The most
likely cause for the TI concentration to rise with low wind speeds or the TI
concentration to decrease with high wind speed is ship contamination.  With low
wind speeds, the ship effluent may very likely be carried up to the bow sampling
site by turbulent eddies„  With high wind speed, and given the ship pointing
into the wind, there would be little possibility for the ship effluent to
contaminate the sampling site0  Therefore, TI is seen as a ship contaminant.
Other constituents have an inverse loading of no more than one-third displayed
by TI on factor two, so TI is isolated in this regard.

     Factor three is made up of the three wind direction sectors only and thus no
dominant effect of source regions is present in the data base.

     Factor four is the most interesting of all factors,,  The constituents that
comprise this factor are associated with fine particles (particle diameter <_ 1 urn)
and yet total suspended particulate mass (TSP) is also a member of this factor,,
In urban and land-based environments, the opposite is usually the case, with
coarse particulates (particle diameter >_ 1 urn) contributing the majority of the
mass.  The physical explanation for this is that coarse particles are lost to
the lake prior to reaching the midlake sampling location.

     In factor five, WD1 and WD2 wind direction sectors are co-members with
CA, MG, and AU  Calcium and magnesium are quite abundant in the lake and can be
considered as lake sources as well as soil-derived, which explains the WD2, lake,
and wind direction sector presence.  Aluminum is abundant in soil and can there-
fore be considered as a soil-derived source contributed to by the WD1 wind direc-
tion sector.

     Factor six is made up of the reactive constituents N03 and $04 associated
with temperature stability*  Total suspended particulate mass is a secondary
member,,  This suggests that temperature stability is needed to bring coarse
particulate MG and AL to the midlake region and thus contribute to TSPU

                     Block X - Determine "Odd" Variables

     Looking at the results of the rotated factor matrix, three variables -- WD1,
WD2, and WD3 do not seem to lend added information.  In fact, they form to make up
a factor, namely factor 3, on their own.  It would appear warranted to perform
the factor analysis by omitting these "odd" variables.

                       Block XI - Factor Score Matrix

     There may be concern that a single or a few data points may skew the factor
analysis results, or one may want to evaluate the relation between the data points
and factors.  The factor score matrix provides a scale on which one may compare
data points with each other.   Equation (4) is the basis for calculating the
factor score matrix.,
                                    133

-------
                              G  =   (B'BHB'Z                 (4)

              Where, G       = factor score matrix

                     B       = rotated factor loading matrix

                     Z       = standardized data matrix

                     B1      = transpose of matrix B

                     (B'B)"1 = inverse of product of B and B1


     The factor score matrix presented in Table A-5 gives a score for every case
or data point for every factor*  The score shows how each data point has affected
the factor.  The mean, standard deviation, minimum, and maximum values for each
factor are shown at the bottom of the table.

                       Block XII - Plot Factor Scores

     The factor scores have been normalized and plotted in Figure A-4 showing the
mean and standard deviation^  The plots are a much clearer representation of the
factor scoresu  The figure actually represents six separate plots; one for each
factor with the factor scores being the abscissa and the data points the ordinate.

                   Block XIII - Determine "Odd" Data Points

     A data point would be considered "odd" if its factor scores consistently
lie outside the standard deviation for each factor.   In Figure A-4, the data
points starting with the number 20050 to 20150 inclusive were  all  collected  in
the month of May.  With the exception of number 50560, the May data set is
consistently skewed away from the remainder of the data points.  Therefore,  it
would be wise to break the data base in two parts — one consisting of the 10 data
points collected in May, and the other consisting of the remainder of the data sets
excluding data point 50560,  Data point 50560 is clearly an "odd" data point.

                      Block XIV - Re-do Factor Analysis

     Three "odd" variables — WD1, WD2, and WD3 — and one "odd"  data point  —
50560 — have been determined.   The factor score plots suggest breaking the  data
into two parts.   One must now decide whether or not to re-do the factor analysis
based on these results.  If one does decide to do so, the "odd" data points  and
variables must be deleted and the procedure must be started again at Block I.

                 Block XV - Delete "Odd" Data and Variables

     The three wind direction sectors and the data point 50560 have been eliminated
from the data base.  The data base is then broken up into two parts — one with the
May data and the other with the remaining data0   The following sections are  results
of these two parts.
                                     134

-------
RESULTS FOR THE MAY DATA

     The May data consists of 10 data points and 14 variables,.  Statistically,
it is not valid to perform a factor analysis since one needs at least as many
data points as variables.  The correlation coefficient matrix is shown in Table A-6.
The squared multiple correlation for all the variables is well below the
necessary 0.40 mentioned earlier.  This indicates that the data is not well
related on a linear basis and it is not justified to do a factor analysiSo
Based on these arguments, a factor analysis of the May data is not pursued.

RESULTS OF THE REMAINING DATA

     This data set is made up of 29 data points and 14 variables,,  There are just
enough data points and variables showing a squared multiple coefficient above
0.40 to justify a factor analysis„  The correlation coefficient matrix is shown
in Table  A-7.  WS is the only variable displaying a very low squared multiple
correlation,,  The unrotated common factor matrix is shown in Table A-8*  The
eigenvalues suggest that only the first five factors are viable and, thus, to
be considered for rotation,,  The seven factors explain 93% of the variation  in
the data.  The communality of the variables varies between 84 and 99 percent*
The results of transforming the first five factors are shown in Table A-9.
primary and secondary members of each factor are outlined in Table A-10.

     Looking at the first factor, CA, MG, AL, FE, MM, TI, ZN, and N03 represent
pollutant, soil, and lake-derived sources.  DELT being a secondary member, it
enhances the transport of land-based aerosols to the midlake region,,

     Factor two is made up of two parts.  The first is the inverse relationship
between wind speed and TI, which again suggests that TI is a ship contaminant.,
The second is temperature stability related with nitrate*  This relation suggests
that higher temperature (and thus temperature stability over the colder summertime
lake) contributes to the formation of nitrates.

     The conclusion that fine particulates contribute to TSP is again reflected
in factor three.  There is, however, the added component of temperature stability
necessary to transport those fine particulates to midlake.  A second dominant
aspect of factor three is the higher temperature creation of N03 and $04.

     A combination of coarse and fine particulates contributing to TSP is present
in factor four.  The coarse aerosols associated with MG may be largely lake-derived.

     The least weighted factor,  factor five, shows a two-fold relation.  Only FE,
TI, and P very likely represent soil-derived sources.  The temperature dependence
of the chemical transformation to produce $04 is secondarily present*

CONCLUSIONS

     Factor analysis is a powerful  technique to aid in the interpretation of large
data bases*   The method is not limited to any particular field and is widely adapt-
able to a broad range of applications*  Using the Lake Michigan data base, several

                                    135

-------
relationships evolved that would not have been detected without the use of factor
analysis:  the inverse relationship between titanium and wind speed leading to
the conclusion that titanium must be a ship contaminant; and the important
conclusion that fine particulates primarily contribute to TSP in the midlake
region,,

     There is, however, concern in using factor analysis.  The common factor
model is a linear model and therefore assumes linearity in the datau  This may be
an incorrect assumption,and the researcher must consider this problem carefully.
The relationship within the factors themselves, however, may be non-linear and
still allow the use of factor analysis.
                                     136

-------
              WISCONSIN
                                LAKE
                              MICHIGAN
SCALE IN KILOMETERS
;=K"s=i2"9
0   25   50  75   100

 SCALE IN MILES
 ^^BES
0  10 20 30 40 50
                                                      Location of
                                                      R/V Simons
                 ILLINOIS

                 WD1
  42°00*
                                87°00'

      FIGURE  A-l.  Wind Direction Sectors Over Southern Lake  Michigan.
                                  137

-------
DATA MATRIX  Im
  Dm.n
m=*OF DATA PTS.
n=** OF VARIABLES
 STANDARDIZE
 DATA MATRIX
    Sm,n
CORRELATION
 COEFFICIENT
   MATRIX
   Rn,n
DELETE "ODD"
DATA AND
VARIABLES
1DOES ONE
DO A FACTOR
  ANALYSIS
                                /

\

f
COMMON FACTOR
MATRIX i
Fn,p
JD =4* OF FACTORS

DETERMINE -^
"ODD"
VARIABLES
>
(
FACTOR SCORE 2E
MATRIX
FSm,j

	 JJ
f
s
V,
^

YES
J.^OP 21
FACTORS HAVING
EIGENVALUE
^O.G

INTERPRET 3Z
FACTOR
LOADINGS
>0,5

PLOT FACTOR23E
SCORES
VS.
DATA PoiMTS

^
•»,


INTERPRET ^E
COMMUNALITY

i
ROTATE c
FACT
Tn,

f
:OMMON
'ORS^^
• son
J

DBTERMIKE 2IE
"ODD"
DATA POINTS
                       FINAL
                    INTERPRETATION
     FIGURE A-2.  Basic Outline of Factor Analysis,
                       138

-------
                            /

                         i.o -

                          .8 -
                          •6 -
                     WD3 •
                          .4-
i
^	1	H
  -i.o   -.8  -.6  -A-  -.2
                         -.44
                 •WD2      '
                         -.6 I
                             FS  .Fa
                                            . DELT
                                          MASS
                                          .WDi
                                                 so
                                           •p   CA7 'MM —
                                            r
                                                 FE
                       WS/
                          r-8+
                         ;w
             FIGURE A-3.  Rotation of Co-ordinates,
                            139

-------
•£»
O
        5055

        5053<
                o
       4030
       
-------
CA
MG
AL
FE
MN
PB
TI
ZN
P
N03
so4
MASS
WS
WD1
.33
.81
.88
.90
.89
.71
.71
.84
.37
.53
.55
.45
-.13
.53
.17
.85
.79
.78
.69
.58
.73
.35
.59
.56
.50
-.09
.46
•ID
.91
.86
.69
.67
.79
.40
.55
.63
.55
-.14
.53
.39
.97
.72
.78
.90
.53
.52
.65
.54
-.10
.48

^9
.78
.72
.95
.59
.61
.70
.63
-.14
.50


J12
.51
.88
.47
.64
.68
.52
-.12
.37



J£
.68
.51
.24
.46
.28
.29
.28




.16
.62
.57
.70
.59
-.8
.43





.36
.49
.62
.71
.11
.20






.13
.79
.79
-.34
.33
                                                                .79   .69
                                                                                  •H
WD2   -.41  -.45  -.41   -.34   -.39  -.31  -.19  -.31   -.18   -.43  -.41  -.38   .29   -.61    .20.
WD3   -.17  -.5   -.18  -.19   -.17  -.10  -.12  -.18   -.04   -.06   .8   -.05  -.07   -.51  -.37   .01
DELT   .50    .60   .65   .49    .46   .50   .16   .40    .09    .70   .63   .52  -.35    .32  -.43   .09   .£Q_
       CA    MG    AL    FE     MN     PB    TI    ZN    P    N03   S04   MASS   WS   WD1  WD2   WD3    DELT
                          TABLE  A-l.   Correlation Coefficient  Matrix.

-------
    VARIABLES
       c A
       MG
       AL
       FE
       MN
       PB
       TI
       HK1
        P
       N03
       S04
       MASS
       WS
       WDf.
       WD2.
      DELT

X VARIATION
E IGENVALUES
         FACTORS

1.  2.  *3.  4.   5.   6.  7.  8.
                              COMMUNAL)TV
.89
.85
.91
.92
.95
.83
.68
.91
.fal
.75
.81
.73
-.18
.57
-.50
-.13
.£4
-.18
-OZ
-09
-25
-.18
705
756
724
7)9
•45
.2^
.25
763
.09
751
.43
.52
.19
.09
.17
.07
.02
706
711
-.05
74fi
-.20
733
730
-45
.63
-.01
764
.06
-23
727
717
712
-:02
.OZ
-.24
.04
•41
.23
.id
.43
717
.1$
.37
759
713
706
710
-.11
-.06
-.01
-20
.12
706
.30
713
-04
.14
.17
.47
-58
.09
735
.05
715
-13
.09
JG
.1<2>
-.01
.22
.|3
-.08
-.03
k20
-.50
714
703
.17
734
704
.05,
715
-r1G
705
.46
-.18
.16
-.M
.06
-.04
-.07
.21
.10
-08
702
707
.07
.30
X4
-.00
.05
-:08
722
700
.10
.06
-.30
.13
-.00
707
.03
.04
713
54
9.2
2.0
      1.6
7
i. a
 5
0.9
 a
sP-4
 2
0.3
                                           PROPORTION OF VARIANCE
                                           OF EACH VARIABLE
                                           INVOLVED IN THE FACTOR.
                                           SPACE.
                                             A LOADING '•
                                            CORRELATION OF A
                                            VARIABLE WITH THIS
                                            FACTOR	

                                            PERCENT OF VARIATTIOM
                                            AMONG ALL THE VARIABLES
                                            THAT is ACCOUNTED FOR
                                            BY THE FACTOR.S  •.
  SUM OF THE COLUMN OF SQUARED
  FACTOR. LOADINGS. ALGEBRAIC ROOT
  OF A CHARACTERISTIC EQUATION!.
                °t;
                             PERCENT OF VARIANCE AMONG
                           'ALL THE VARIABLES THAT is
                        ACCOUNTED FOR BY THAT FACTOR:
                                m   .£
                                                          X 100
                                               m
                      TABLE A-2.  Common Factor Loading Matrix.

-------
              TABLE A-3.  Rotated Factor Matrix.
VARIABLES

  CA
  MG
  AL
  FE
  MN
  PB
  TI
  ZN
  P
  N03

  S°4
  MASS
  WS
  WD1
  WD2
  WD3
  DELT

% VARIATION
FACTORS
1
.93
.83
.85
.94
.89
.84
.89
.89
.38
.33
.48
.25
.00
.32
.22
.13
.32
2
,06
-.04
-.00
.02
.10
.21
-.41
.11
-.15
.27
.11
.00
.97
.16
.16
.04
.20
3
.10
.00
.14
.11
.09
.01
.02
.08
.01
-.07
-.12
.09
.06
.54
-.33
.98
-.07
4
.12
.16
.18
.27
.39
.44
.17
.43
.88
.69
.76
.88
.02
.13
.17
-.02
.19
5
.23
.26
.23
.13
.15
-.02
.10
.04
.04
.13
.10
.20
.18
.74
.87
-.09
.16
6
.23
.47
.41
.15
.11
.25
-.11
.05
-.23
.57
.40
.35
.16
.10
.17
-.07
.89
41
20
10
12
                             143

-------
FACTOR
"LABEL"
PRIMARY
MEMBERS
SECONDARY
MEMBERS
PERCENT
VARIATION
HIFRHBrHV
1
Transport of soi 1 ,
lake, and anthropo-
genically derived
aerosol enhanced by
temperature stability
and Chicago/Gary
Source Region
CA, MG, AL
FE, MN, PB
TI, ZN
P, N03, S04
WD1
DELT
41

2
TI ship contaminant,
other metals' cone.
not significantly
contributed to by
ship effluent
WS
-TI
8
K
1
3 . 4
Fine
Stand Alone Particulate
Wind with
Direction Factor Nutrients
WD1 P, NO,, SO.
WD3 Mass
MN
-WD2 PB
ZN
9 20
c o
5
Soil and Lake
Derived Coarse
Particles
contributing
CA, MG, AL
WD1
WD2
CA
MG
AL
10
n
6
Temp. Stab
brings out
reactive and
coarse
particulates
N03
DELT
MG
AL
so4
MASS
12
•)
TABLE A-4.   Interpretation of Factors.

-------
SAMPLE SET
  NUMBER
  20050
  20060
  20070
  20080
  20090
  20100
  20110
  20130
  20140
  20150
  30170
  30180
  30190
  30200
  30210
  30220
  40240
  40250
  40260
  40270
  40280
  40290
  40300
  40310
  40330
  40340
  40350
  40360
  40370
  50400
  50410
  50450
  50470
  50490
  50500
  50510
  50530
  50540
  50550
  50560
  AVERAGE
  SIGMA

  MINIMUM
  MAXIMUM
                             FACTORS
   1

 -.781
 -.670
-1.243
-1.183
 -.932
 -.922
-1.206
 -.952
 -.804
 -.825
 -.308
 -.269
 -.366
 -.501
 -.646
 -.747
 -.251
 -.292
 -.310
 -.358
 -.498
 -.583
 -.697
 -.442
 -.294
 -.411
 -.241
 -.279
 -.300
 -.262
 -.289
 -.279
 -.261
 -.307
 -.383
 -.221
 -.511
 -.745
-1.389

 -.558
  .323
-1.389
 -.221
 .038
-.021
 .260
 .284
 .144
 .170
 .317
 .096
 .051
 .062
-.102
-.129
-.073
-.008
 .047
 .106
-.136
-.113
-.103
-.082
-.051
-.055
-.016
-.101
-.111
-.049
-.141
-.127
-.109
-.124
-.105
-.114
-.119
-.109
-.090
-.096
-.140
-.001
 .080
 .269
-.012
 .130
-.141
 .317
3
,443
.427
,535
,422
397
337
,376
521
444
459
184
191
197
229
284
309
218
199
182
230
306
401
464
363
209
233
223
239
186
178
194
188
207
238
269
227
156
215
316
649
301
122
649
156
4
1.668
1.405
2.839
2.470
1.912
1.719
2.408
2.226
1.718
1.782
.091
.036
.239
.595
1.026
1.243
.009
.096
.131
.264
.778
1.196
1.534
.731
.085
.367
-.021
.094
.096
-.013
.037
.029
-.017
.150
.404
.292
-.118
.620
1.321
3.429
.872
.939
-.118
3.429
-1
   730
   634
   129
-1.054
 -.863
 -.788
-1.036
 -.933
 -.754
 -.781
 -.244
 -.228
 -.296
 -.416
 -.561
 -.636
 -.204
 -.256
 -.282
 -.297
 -.467
 -.589
 -.677
 -.420
 -.236
 -.334
 -.195
 -.232
 r.255
 -.206
 -.223
 -.223
 -.203
 -.253
 -.333
 -.312
 -.202
 -.454
 -.658
-1.285

 -.497
  .304
-1.285
 -.195
             1
1.162
 .906
2.166
2.082
1.524
1.511
2.137
  534
1.218
1.268
 .177
 .091
 .305
 .601
 .915
1.138
 .038
 .142
 .184
 .275
 .566
 .726
 .971
 .434
 .145
 .402
 .026
 .107
 .162
 .071
 .135
 .115
 .076
 .170
 .330
 .271
-.002
 .633
1.117
2.422
 .706
 .690
-.002
2.422
                         TABLE  A-5.   Factor Score  Matrix.
                                          145

-------
cn
CA
MG
AL
FE
MN
PB
TI
ZN
P
N03
S°4
MASS
WS
DELT

.05_
.85
.77
.94
.98
.64
.90
.94
.84
-.16
.23
.04
.50
-.50
CA
.87
.80
.80
.62
.75
.78
.75
-.03
.10
.06
.20
-.30
MG
•1Z
.86
.77
.44
.80
.66
.55
-.41
.02
.09
.49
-.15
AL

•1Z
.97
.56
.97
.87
.69
-.37
.24
-.09
.57
-.41
FE


.Jl
.68
.94
.96
.81
-.22
.32
-.04
.53
-.53
MN



.30
.53
.83
.87
.27
.57
.11
.22
-.37
PB




J6
.83 ._1_7
.62 .93 Jl
-.32 -.07 .14 .00
.33 .43 .35 -.05 .J0£
-.21 -.00 .18 .34 -.56 .29
.50 .45 .29 -.56 .06 .30 ._31
-.38 -.58 -.57 .15 -.48 .36 .05 .19
TI ZN P NO, SO. MASS WS DELT
                    TABLE A-6.  Correlation Coefficient Matrix for the May Sample Sets.

-------
CA
MG
AL
FE
MN
PB
TI
ZN
P
N03
so4
MASS
WS
DELT

09
.51
.90
.75
.73
.38
.52
.60
.12
.49
.24
.18
-.06
.30
CA
•Ji
.54
.56
.57
.34
.43
.43
.26
.42
.18
.39
.18
.13
MG
.J5
.87
.86
.56
.65
.81
.34
.64
.39
.22
-.02
.44
AL
.44
.89
.52
.62
.74
.39
.73
.42
.29
-.08
.47
FE
.J58
.64
.48
.81
.26
.83
.49
.40
-.21
.51
MN

J6
.47
.82
.30
.72
.82
.53
.08
.56
PB


.22
.54
.59
.38
.51
.34
.44
.16
TI



.58
.37
.69
.66
.28
-.02
.52
ZN




•!§
.24
.42
.28
.26
.22
P





-.34
.58
.54
-.19
.53
NO.,






•.34
.45 .3JL
.04 .10 .05_
.45 .19 -.29 .30
SO. MASS WS DELT
TABLE A-7.  Correlation Coefficient Matrix.

-------
     VARIABLE
FACTORS
                                                                                                   COMMUNALITY
-IS.
CO

CA
MG
AL
FE
MN
PB
TI
ZN
P
N03
so4
MASS
US
DELT
% VARIATION
EIGENVALUES
1
.73
.60
.89
.89
.91
.81
.69
.89
.46
.83
.68
.50
.00
.57
51
7.2
2
-.16
.20
-.10
-.11
-.26
.04
.56
-.07
.53
-.23
.15
.24
.87
-.39
13
1.8
3
-.52
-.37
-.37
-.28
-.14
.45
-.16
.07
.13
.20
.56
.38
-.11
.35
11
1.5
4
-.02
.45
-.14
-.06
.10
.02
-.19
-.19
-.33
.23
-.14
.64
-.01
-.27
7
1.0
5
.15
-.12
.06
-.14
-.04
.30
• .01
.20
-.57
-.04
.16
-.16
.32
-.15
5
.7
6
.11
-.38
.05
.05
.05
-.09
.18
-.04
.02
.08
.15
.14
-.23
-.44
4
.5
7
-.20
.26
-.07
-.01
.07
.06
-.15
.22
.11
.01
.14
-.25
-.14
-.32
3
.4

91
96
97
90
94
96
90
92
97
84
89
97
94
99
93

                                TABLE A-8.   Unrotated  Common Factor Matrix.

-------
VO
VARIABLE



  CA

  MG

  AL

  FE

  MN

  PB

  TI

  ZN

  P

  N03

  so4

  MASS

  WS

  DELT


% TOTAL
VARIATION

1
.99
.73
.94
.89
.82
.31
.57
.68
.13
.55
.13
.06
.06
.30

2
-.02
-.20
.00
.14
.24
-.07
-.52
.01
-.12
.30
-.08
-.06
.98
.60
FACTORS
3
.13
-.08
.31
.29
.41
.91
.33
.72
.19
.61
.94
.36
-.02
.69

4
.02
.64
.01
.16
.29
.28
.11
.03
.13
.48
.19
.92
-.05
-.05

5
-.04
.12
.17
.29
.08
.04
.53
.14
.96
.06
.25
.13
-.18
.26
                          37
13
27
13
11
                                 TABLE A-9.   Rotated Factor Matrix

-------
en
O
FACTORS
"LABEL"
PRIMARY
MEMBERS
SECONDARY
MEMBERS
% VARIATION
HIERARCHY
1
Transport of
soil , lake and
anthropogenically
derived aerosol
enhanced by temp.
stability
CA, MG, AL,
FE, MN, TI,
2N, N03
DELT
37
1
2
TI - ship
contaminant
-TI, WS
DELT
N03
13
3
3
Fine participate
transport by temp.
stability enhances
mass cone, at mid-
lake. Higher temp.
enhances S04, NO,
cone.
PB, ZN, N03,
S04, DELT
MN, MASS
27
2
4
Coarse parti -
culate as
well as fine
parti culate
mass.
MG, MASS
MN, PB, N03
13
4
5
Soil derived
sources and
DELT reaction
of S04.
TI, P
FE, S04,
DELT
11
5
                 TABLE A-10.  Meteorological and Aerosol Data for the Profile Data Sets.

-------
                                 APPENDIX B

                         MICROMETEOROLOGY RELATED TO
                   DIABATIC DRAG COEFFICIENT DETERMINATION
     Results of modeling efforts to date clearly point to the flux of aerosol
across the air/water interface and its discrimination by aerosol size as the
most crucial parameter needing experimental measure.  Other researchers
(Andren, 1976; Gatz, 1976; Hess and Hicks, 1975; Kramer, 1976) also have
identified this flux or deposition rate as the major unknown quantity needed
to resolve the question of atmospheric contribution to Great Lakes loading.
The deposition velocity, v ,, Is given  by

                         Flux of constituent of interest              /, %
                         Concentration of that constituent

and, more specifically for atmospheric aerosol

                   Flux of aerosol as a function of radius
         v (r)  =
                   Concentration of aerosol as a function of radius
     v(j(r) was used to approximate fluxes into Lake Michigan.  Using wind
tunnel data of Sehmel and Sutter (1974) for aerosol v,j(r) and having C(r) of
aerosols at or near the water surface, the flux can be approximated by

                       Aerosol Flux  =  vd(r)  •  C(r)                (2)

     It is clear that a field measurement of Vd(r) under varying meteorological
conditions and for constituents of interest such as the trace metals and phos-
phorus is the most crucial piece of field data to resolving the controversy
over lake loading by atmospheric aerosols.

     Defining a tractable experiment is not a simple task.   In general,  the
flux of a constituent S can be related to the mass fraction of the constituent
S per unit mass of air (s), as well as the vertical velocity of s (We) and the
density of air (p).  The rate of vertical transport, i.e. the flux, is expressed
by pWss.  If this transport is measured a few meters above  a horizontal  rigid
surface, upward and downward currents are found to be distributed in space and
time in the chaotic and random fashion we know to be turbulence,,  The net
vertical transport of air, however, must be zero if the total mass at the
surface is constant (as we know it to  be at the earth-atmosphere interface).

                                    151

-------
However, the average value of pWss is not zero,  for the downward currents
may systematically carry larger amounts of s than the upward currents.  The
rate of average vertical transport of S, the flux Fs, is then dirnensionally
given by
                     yg of ai
                     m3 of ai
ir / m \

ir \sec/
yg of S

yg of air
                                                   yg of  S
                                                   m   sec
      It  can be shown that the flux of aerosol A is given by


                               F   -   oK  —
                               FA  ~  -pKA
where a is the mass fraction of aerosol  A, and Ka is the aerosol  mass transfer
coefficient.  Priestly (1959) and others have pointed out that the sheer stress,
T, can be related both to the change in  u with height through the momentun
transfer coefficient K^, as well  as to wind speed at one height alone (ui) through
the momentum drag coefficient, CQ, i.e.

                                     3 U
                                     _  .  pCDu?


     Because we are concerned with aerosols of radius <_ 10 ym in
terminal velocity is less than the vertical velocity fluctuations
aerosols  <_ 1 ym radius.  This is also true
for some in the size range 6  <  r  < 10  ym,
density „  For this case
                                                                  all cases, the
                                                                  (Wa) for all
                                            for most aerosols  _< 6 ym radius and
                                            depending on their shape, surface and
                                  K
so that
                                                                      (6)
                                                                      (7a)
 or, in finite difference form
                                =  -p
                                                                      (7b)
 where  u2  is  the  wind  speed  at  the  height  for which  the second aerosol
 tion a has  been determined.
                                                                       concentra-
                                     152

-------
     Gilette et a1. (1972) have used this equation to study wind erosion with
some succesSo  If CQ can be well specified and if high enough winds prevail,
statistical significance in U2 - u-j and 82 <- ai over a soil surface can be
demonstrated.  It was for these narrow meteorological conditions on certain
"days of opportunity" that Gillette obtained F/\ for four different size ranges
of aerosol, using cascade impactors.

     The problem in using the above equation for F/\ generally comes in achieving
statistical significance in &2 - a-j and U2 - U] over soil surfaces.  Over a water
surface a^ - a-) is relatively easier to measure because one has essentially no
re-entrainment (until wave whitecapping) but the measurement of U2 - U] becomes
more difficult.  Since a water surface presents less wind resistance, the differ-
ence between uo and u-j, no matter what the height separation (remembering to
confine our attention to the surface layer of 30-50 m) will be even less than
it is over landH  The measurement of u? - ih over land has been accomplished on
a number of occasions, usually with difficulty and requiring sophisticated equip-
ment.  For example, Businger et ai. (1971) obtained U2 - u-| for a wide range of
wind speeds and stability.  That experiment required the use of eight cup anemome-
ters at 2, 4, 5.7, 8, 11 .,3, 16, 22 and 32 meters in a stationary position for
three months of once per second data0   Given the still more difficult task of
obtaining statistically significant u;? - UT  measurements over a water surface,
the use of equation (7) for F^ determination appears infeasible.  The good
fortune of a "meteorological day of opportunity" with relatively high winds of
sufficiently sustained steadiness to obtain characteristic mass loadings does
not suffice for the expensive and thus relatively few sampling occasions afforded
by EPA shipboard sampling.

     If the sampled portion of the surface layer (here, the first 10 m) can be
assumed to be a constant flux layer, one can integrate equation (7) from the
surface to any one height of interest  (i.e., a single sampling height) and
thereby obviate the measurement of U2  - u-|.   Several  indirect flux experiments,
that of Dyer and Hicks (1970) chief among them, have convincingly shown the
surface layer to be a constant flux layer.

     Integration of equation (7) with  constant F. gives


                                                1                     (8a)
or

                           If   -   » I D I /T    ~\                      (Qh\
                           rn--pjDKa-a;                      ^oD;


where Bm is the bulk turbulent transfer coefficient for momentum.  Priestly (1959)
has shown that Bm can be related to the momentum drag coefficient, Cpj, by
                             Bm
                                    153

-------
where UH is the_average of the horizontal  wind speed at any one height in the
atmosphere and UQ is the average speed at  the roughness length height (i.6o, we
specify the lower height for integration 2]  = HQ).   The roughness length height
over a lake surface is less than 0.05 cm.

     Considering the lake surface to be an efficient sink for aerosols, a-| = a~n
is approximately equal to 0.  This should  hold as long as whitecapping of waves
(requiring u _> 12 m/s) does not occur, for then one might expect significant
aerosol release from the surface.  Given the above  considerations, equation (8b)
results in                  _
                            FA  =  -p?CD(U- ZT0)                      (10)


     Measurement of u" and a" is relatively  simple^  (Size fractionating a~ by
using a cascade impactor will, of course,  give F/\ as a function of aerosol size.)
A value of UQ can be obtained by measuring the time it takes a spherical object
floating half in/half out of the water to  pass from stem to stern of the sampling
ship.  A value for CQ is all that remains  to be determined.

     When the roughness elements of the surface are greater than the roughness
length and for neutral conditions, Deacon  and Webb  (1962) approximate CQ to be

                     CD  =  [i + o.07(IF10  -ig]  x  io-3               (ii)

     Experimentally, the neutral case_certainTy seems to give a tractable
solution.  One needs only to measure U-JQ and UQ.  These measurements can be
done fairly easily.

     However, equation (11) assumes the water surface to be fully rough.  We
need to relax this assumption as well as the assumption of neutral stability,
so that both mechanical and thermal stability variations can be considered.
This is so because neutral thermal stability and mechanical instability of the
surface occurs less than 5 percent of the  time over Lake Michigan.

     Incorporating smooth, transition, and rough surface conditions has already
been done in the micrometeorological literature of  the 1960's.,  The resultant
curve is the heavy line of Figure 34with the straight line portion being given
by equation (11).  The diabatic case (thermal instability) is much more difficult
to resolve.

     Deardorff (1968) proposed a way to handle stability effects on drag coeffi-
cients.  However, he had to assume the drag coefficient for momentum, CD, to be
equal to CH and Cjr, the analogous coefficients for heat and water vapor.  Further,
his method raises the problem of determining the Monin-Obukhov length -- a very
difficult parameter to obtain in field experiments*  Thus, a derivation of the
drag coefficient for the diabatic case (CDQ) in terms of measurable quantities
is essential.
                                    154

-------
      A basic definition of CQQ in terms of the ty functions of Panofsky (1963) is


                                 CDD  =  k?/^M

 where k is the von Karmen constant and ^ is  the dimensionless momentum
 function given by
/ d>..
 M
—  db,
 h
                                       'b0  b
                                                                       (13)
where  bQ  =  zfl/L,  and  b  =  z/U   Dyer  and Hicks(1970)  have  found,  for the  unstable
case,

                             4>M  =   0  -  16b)"1/4                      (14)

     For  the  stable case
                                             b                          n c\
                             *M  =   1+6
                                           1  +  b

      Integration of  |v|/b  for  b  <  0  (the  unstable  case) yields:


                      /•b db    A / 1         x          1
                                    ll + x     1 + x2     1 + x2
                                                                 A     (16a)
                                                                 dx    x    '
                  b        (x^ + 1) (x0+ I)2
           =  In —  +  In- - — --  + 2 (tan-:x - tan-xxn )    (16b)
                 b0        (x2 + 1) (x + I)2

where

                    x  =  (1 - 16b)1/4       XQ   =  (1 - 16b0)1/4      (17)

Integration of M/b for b > 0  (the stable case)  yields


                       ^  =  In b/b0  + 6 In

     These expressions for ^M (and for ^ and ip£, calculated in a similar way)
can then give the diabatic transfer coefficients.  Details of the calculation
are:
                                     155

-------
1„  Take an arbitrary value of neutral  wind speed u(n).   A friction velocity
    is calculated by

                              u*  =  CD u(n)                         (19)

2.  Take arbitrary values of the characteristic temperature and specific humidity
    Values for b, b0  and ik. can then be calculated from equation (13)0

3.  CDD can then be obtained from

                               CDD  =  k2/^, .                        (20)

     By this procedure a set of values  of u, T0 -~fl+  0.61 0(qo  - q^) and CDD is
obtained.  For conditions over Lake Michigan, TQ  - T  + CL61 0 (qo - qj - T  - Tj .
Calculations were made for several score sets and values plotted.  A sketch is
shown in Figure 36 where the ratio of CQQ to CD is plotted against S0,  a stability
parameter defined by


                         C   -
                          0     ,.2r, A •»„„  AnYl,                     (21)
     Experimentally, one needs to measure TQ  and Ji (in addition to u, u0  and a).
The surface temperature can be obtained through use of an infrared sensing
thermometer.  The 10 m height temperature is a standard measurement.  The differ
ence between the two must, however, be measured with at least 0U5° C accuracy.
The IR thermometer used in this experiment is accurate to ± 0.2° C and the
thermistor to ± 0.1° C.
                                    156

-------
                                 APPENDIX C

             TRACE ELEMENT LOADING OF SOUTHERN LAKE MICHIGAN
                  BY DRY DEPOSITION OF ATMOSPHERIC AEROSOL


INTRODUCTION

     The densely populated and heavily industrialized southwestern shore of
Lake Michigan represents a significant and expanding source of anthropogenic
aerosol to the atmosphere.  The combustion of fossil fuels in residential,
industrial, and transportation activities, as well as manufacturing processes
such as steel-and cement-making, are principal sources.  Pollutant aerosol
emitted in nearshore urban/industrial areas provides an input potential  for
loading of Lake Michigan via wet and dry falloutu   Fifty percent or more of
the time, prevailing winds give rise to greater than 80 km long trajectories
over the lake for air masses passing through the Chicago/Gary area (Williams
and Sievering, 1975)u  Not only the health of the  urban population, but  also
the health of the lake ecosystem,  may therefore be linked to atmospheric pollu-
tion by wet and dry deposition.

     Earlier workers sought to quantify the extent of atmospheric route  loadings
for several elements by estimation schemes which were based on assumed transfer
efficiency (Et) values, where:

                            pollutant loaded to lake
                     E   =  	               (1)
                            pollutant emitted at source

Winchester and Nifong (1971) estimated loadings by calculating an emission
inventory for the southwest shore  source region and assuming an Et value of OJOU
It was soon pointed out, however,  that this value  of E^ may be too low.   In con-
sideration of mesoscale circulation effects on aerosol  trajectories,  as  well  as
temperature stability and wind speed effects on deposition, Skibin (1973) suggested
that an Et of 0.25 or larger would be appropriate.  Gatz (1975a) presented a
refined characterization of Chicago source region  aerosol composition, which was
used in loading estimates (Gatz, 1975b)u  Sievering (1976) described  a model  of
the transport and deposition of aerosol  over the lake,  and proposed an Et of
0.20 to 0.40, dependent upon season, for combined  wet and dry deposition.
All of these studies, however, were based upon extrapolation of shore-collected
aerosol chemistry data and, at best, wind-tunnel measurements of aerosol  dry
deposition rate (Sehmel and Sutter, 1974).  Very little has appeared  in  the
literature regarding overlake aerosol composition  and the deposition  rate of
that aerosol.  Eisenreich, Emmling, and Beeton (1977) used primarily  shore-

                                    157

-------
collected bulk preparation samples to estimate lake loading  from field measure-
ments of deposition rates,,  The bulk samples,  i.e., wet and  dry fallout combined,
give no better than an order of magnitude accuracy for the dry depositional
component of total loading (Cadle, 1974).  The objective of  the research presented
here is to consider the dry deposition contribution to lake  loading  separately.
By the collection of an aerosol composition and deposition rate data base  at
midlake, the annual dry route loading rate to  southern Lake  Michigan can be
better estimated„

     In order to approach this objective, three aspects of concurrent investigation
were pursued at a midlake site (87°00'W by 42°00'N), about 55 km ENE of Chicago^
First, the concentration and composition of atmospheric aerosol were determined.
Data describing the chemical  makeup (for certain trace lements) and  the physical
nature of the aerosol were collected.  Second, surface-layer meteorological  data
were obtained for the estimation of aerosol deposition velocity (vd).  Values
for vd based upon midlake data not only improve upon wind-tunnel  vd  data (because
all lake-surface effects cannot adequately be  duplicated in  the laboratory), but
also can be determined as a general function of meteorological parameters  and
changes in those parameter   Through climatological information, vd results from
limited sampling periods may be extrapolated into an estimation of the entire year's
loading rateu  Third, meteorological observations were utilized in the calculation
of mesoscale aerosol trajectories.  These collection site-to-shore back trajectories
provide information on the lake's effect on transport and deposition of aerosol,
as well as provide identification of general source regions  on shore for the
aerosol„

METHODS

     During four several-day-long periods in the summer and  autumn of 1977,
aerosol samples and meteorological data were collected from  the UUS. Environmental
Protection Agency  (USEPA) R/V Roger R. Simons, anchored at a mid-Lake Michigan
station (see^Figure  1 )„  This site was chosen to be fairly  representative of
most of the southern basin lake surface.  The nearest shore  is 40 km distant,
so that effects upon loading rates due to point sources and  the land-water
boundary zone are minimized.,  The 40 km or greater upwind fetch allows a 10 to 30  m
thick  "surface layer" of air, just above the water, to reach equilibrium with
respect to surface effects„  As long as the lake surface appears constant to
the air passing over it, the mass flux within the surface layer is constant
throughout its vertical extent (Kraft, 1977).   Thus, under most conditions,  the
aerosol is well-mixed within the atmospheric surface layer by the time it  is
transported to the sampling site  (Gillette and Winchester,  1972K  The site is
also well away from heavily travelled shipping Ianes0  Reports from several  National
Weather Service and UUSU Coast Guard stations were added to  R/V Simons, Governors
State  University,  and Argonne National Laboratory meteorological data, forming a
network of data points around the southern end of the lake»

     Aboard the R/V Simons, meteorological instruments (Table 1) were mounted
at the tip of a 6  m aluminum boom which extended forward of  the bow horizontally
at a mean height of 5 m above water.  With the ship bow-anchored on station, the
boom faced upwind; this allowed for measurement of ambient conditions.  The results

                                           158

-------
of Hunt and Mulhearn (1973) show that even a 3.5 to 4.0 m boom length would have
been sufficient to avoid turbulence effects due to the ship's structure,,  During
very light winds  (u5  < 2.0 m-s~1) or periods of quickly changing WD, which
made  the windward orientation of the boom questionable, sampling was halted,,
Analog output from the U5, WD, Tg, and RH sensors was sampled once per second
by a data acquisition system  (Weather Measure Corp., #SC601/M733).  For each
15 minutes of the 1 Hz data,  arithmetic mean and standard deviation (a) values
were calculated by the system.  These mean and a values were hardcopy printed
and stored on digital magnetic tape (Texas Instruments, Inc., #ASR733).  The
vertical wind (1^5) analog output was continuously monitored on a strip chart
recorder as a check on the vertical motion of the boom.  The infrared thermo-
meter was operated manually by an observer standing on the upper foredeck.

     Shipboard data collection was done around the clock.  In order to divide
each sampling period into manageable time segments, data collection was done
within 3 to 6 hour periods, referred to as "data sets".  Each data set, because
of its limited duration, corresponds to a period of fairly constant meteorological
conditions.  The occurrence of rain or fog events precluded any sampling, since
this study was directed at dry deposition only.  At the beginning of each data
set, aerosol sampling media were replaced.  Thus, aerosol samples can be classified
according to values of meteorological  parameters which prevailed during each set.
In general, a minimum three hours elapsed time for a data set was required to
provide:  1) statistically meaningful  definition of prevailing meteorological
conditions, and 2) sufficient loading of the Hi-Volume filters to be above
detection limits of the chemical analysis procedure.  Past that minimum time,
rapid and persistent changes  in U5, WD, T5, or TQ were cause for the termination
of a data set.  Such changes were defined as two successive 15 minute mean values
for a given parameter being > la removed from the running mean.  A maximum run
time of about 8 hours was allowed so that no data set was biased by a widely
differing run timeu

     Aerosol was collected on three-part cellulose filters (Sierra/Misco #P252A,
P810A) in a three-stage cascade impactor.  The impactor utilized the first,  third,
and backup stages of a Sierra Series 230 Cascade.  This configuration provided
resolution of the aerosol  analyses results into fine particulate (diameter  D  <
1.0 ym) and coarse particulate (D > 1.0 urn) fractions (Sievering, et al., 1978).
The filters were exposed at a flow-controlled (Sierra #310) rate of 1.13 nr-min"1
(40 ft3-min-l),  on a standard Hi-Volume sampler (General  Metal  Works,  #GMW1000)U
The Hi-Vol was located 4.5 m  ahead of the bow, on the instrumentation boom.
This location, and the addition of a 4 m long exhaust tubing to the Hi-Vol
motors, helped reduce contamination of aerosol samples (Schmidt, 1977; Moyers,
Duce, and Hoffman, 1972).  An additional Hi-Vol was operated on the upper fore-
deck with 20 X 25 cm glass-fiber filters (Sierra #C305) for gravimetric analysis
of total aerosol mass concentration.

     Aerosol samples on the three-part cellulose filters were analyzed for trace
elements at the USEPA Central Region Laboratory in Chicago.  The filters were
placed in acid-washed fused quartz trays and low temperature ashed in 03 at  75
watts.  The residue was then  dissolved in HN03 and distilled deionized water.
Trace element analysis was done by Inductively Coupled Argon Plasma atomic

                                    159

-------
emission spectroscopy (ICAP-AES)  (Jarre!1-Ash,  Plasma  AtomComp  750) „   In  more
than 85 percent of the data sets,  concentrations  of Ca,  Mg,  Cu,  Fe,  Mn,  Pb,
Ti, and Zn were above ICAP-AES instrumental  detection  limits (Lj).   The  limited
3 to 8 hour exposure time of the  filters  resulted in concentrations  of Na,  Ba,
Cd, Co, Ho, Ni, and V which were  below LH for 50  percent of  the  data  sets
(Table 2).  Therefore, loading calculations  will  not be  done for these metals.
Post-analysis statistical  review  of the data revealed  contaminant or  otherwise
anomalous data for B, so that this  element was  not used  in loading estimates.
Filter blanks were carried throughout the handling and analysis  procedures,
except the collection of ambient  aerosol.   The blank correction  (3-j)  applied
to the ICAP-AES results for the exposed filters was defined  as:

                               3,-   =  Cbi  +  la                      (2)

           where:  Cu- is the mean concentration  of the  element  i
                    bi
                       in filter blanks,

                    a  is the standard deviation  in that mean value.

                       (both ng - JT1)

For those elements which were above Lj in > 85 percent of the data  sets,  concen-
trations are reported here within the < 20 percent expected  reproducibility
confidence for ICAP-AES analyses.

     Several optical particle-measuring instruments were also used  at the ship
site to collect information on the physical  character  of midlake aerosol.  An
Integrating Nephelometer (IN) (Meteorology Research, #1550)  was operated on the
upper foredeck to continuously monitor aerosol light scattering coefficient and
thus, variations in  total aerosol mass concentration (Dave,  Dolske,  and  Sievering,
1979).  An Active Scattering Aerosol Spectrometer  (ASAS) (Particle  Measuring
Systems, #ASAS 300-PMT) was mounted at the proximal end of the sampling  boom,,
The ASAS counts aerosol number concentration in 60 adjacent size channels in the
range 0.1. <  D  <  3U5 ymu  The size distribution  of midlake aerosol  was  thus
recorded, generally  at 15 minute counting intervals during each data set.  Size
and volume plots yield much qualitative information regarding aerosol transport
and depositional processes  (Slinn, 1974; Sievering, 1979)u  ASAS counts  were
usually  taken at the  5 m mean height above water,,  When vertical ship motion was
minimal,  attempts were made to measure vertical gradients in the aerosol number
concentration.  The  instrument's intake was alternately placed at heights of
3.6 and  6.4 m during  successive counting periods of from 6 to 15 minutes.  In a
limited  number of cases, the measurement of number concentration difference
between  the  two  heights was statistically significant (Sievering, 1979).

     Aerosol  deposition velocity  (v^) was estimated for each data set by means of
a  diabatic drag  coefficient,  C^,  and  the relation:


                                     160

-------
                            vd s Cdd •  (U5 - U0}
                   where:   v, = deposition velocity

                           Tic - mean wind speed at 5 m height

                           TL = water surface current

                                 (all in cm-s"1)

Note that v^, as given by this method, is independent of aerosol  size.   The
limited results of ASAS-measured number concentration gradients  indicate that v^
may not vary strongly with aerosol  size in the range OJ <  D <  2 ym.   Theoretical
models of deposition processes and  wind tunnel  data predict as much as  an order
of magnitude variability in v^ across this size range.   The ASAS  vertical  gradient
measurements suggest the size dependent variability of vj is somewhat  less than  a
factor of three for this size range of aerosol  (Sievering,  1979).  These vertical
gradient measurements further suggest the Vd determined by  relation (3) has an
uncertainty of roughly 2-to 3-fold, which is the same order as the size-dependent
variability.  Aerosol size distribution and total  mass concentration data from
the midlake site indicate that most of the aerosol mass was accounted  for by the
the 0.1 <  D  < 2 ym size range (Dave, Dolske, and Sievering, 1979K  A bulk
estimate of v^ from relation (3) is therefore a reasonable  value, although
not size dependent, for use in loading calculations.

     The value of Cdd used ""n (3) is itself, a function of surface layer wind
speed and atmospheric thermal stability (AT = TS - T0).  The vj  determined for
each data set is thus strongly dependent upon meteorological conditions prevailing
during that period of time.  From th£ most stable cases (AT > 8.3°C) to the most
unstable cases (AT  < -009°C), mean v^ from (3) varies from 0.15  to 0.72 cm-s~l
(Table  11).  For very low wind speeds (u^  <  2.0 m-s~'), aerosol deposition is
impeded by a continuous laminar sublayer immediately above  the air/water interfa£e.
For such cases, the v^ estimate of  (3) is no longer physically reasonable and a_Vd
of zero is assumed.  For the highest wind speed cases (us > 7.0  m-s~'J_, a mean vd
of 1.17 cm-s~' was determined (Table  12 ).  The great variability in vj due to
changing meteorological conditions  must then be considered  in the estimation of
annual loading rates for the lake.   At present, techniques  for direct  measurement
of aerosol deposition, such as eddy-flux correlation and concentration-profile
methods, require much development before V(j can be routinely determined (Hicks,
1979; Sievering, 1979K  Until the  direct measurement methods become reliable for
over-water field sampling, the C^d  method yields meaningful estimates  of lake
loading via dry deposition, within  the approximate factor of two  accuracy indicated,
While this may seem to be a rather  large uncertainty, it represents a  significant
refinement in the estimation of the dry route loading component  over previous
estimates given by extrapolation of shore-collected data or bulk  sampling resultsu

     Two approaches were followed in the calculation of estimated annual dry
loadings to the southern basin of Lake Michigan.  The most  basic  approach is the
overall-average method:

                                     161

-------
                                     ci
                           A   =  vd - -  -  A  -   t                   (4)
                                       d

            where:   X-  is  loading  to  lake of element  i  (kg-yr~')

                    7,  is  mean bulk  deposition  velocity  for all data

                       sets (m-s~^)

                    C-  is  mean concentration of element  i  (kg-m~^)

                    A  is  total  water surface area  of southern  basin

                       of  Lake Michigan (2.9 X  1010 m2)

                    t  is  part of  the year  when no  precipitation

                       occurs  overlake  (s-yr~'}

Equation (4), although  straightforward, entails an  underlying assumption that
deposition at the midlake  point  represents  deposition over the  entire  southern
basinu  For single-site sampling,  this  is an unavoidable assumption.   However,
the surface layer conditions  at  midlake are certainly more similar  to  the majority
of the lake surface than are  conditions on  shore  0£ near to shoreu  Moreover,  (4)
assumes that the data sets which made up  the mean v
-------
              F    is annual frequency of occurrence for defining

                   range of parameter (AT, Ur, or source) - (yr)

              t    is time without precipitation overlake (s-yr~l )

In applying equation (5), it is assumed that the year in which sampling is done
is a climatologically representative oneu  How well the data fits prevailing
climatology is at least partially accounted for by_the inclusion of the F0j
values.  It is assumed, however, that the v^j and C^ values for each bin do
indeed represent the conditions of that bin averaged over a year0  Even so, (5)
is a refinement over (4) in that the variations in v^ and C-j due to surface
layer meteorology and source region are taken into consideration^

RESULTS

     The geometric mean aerosol mass concentration, TTm, observed at the midlake
site was 32. yg-m~3u  The range of values measured was 10.  <  Cm  < 94. yg-rrr3.
Trace element composition of midlake aerosol  is expressed in Table C-l by three
ratios of concentrations:
                                       -
                               P   =  _L . 100                      (6)

                                      Cm

                                  C\ (D < 1,0 ym)

                                   .  (D  <1.0 ym)
                        EF   -  Ci/Cstd (in aerosol)                (8)
                                Ci/Cstd (in soil)

In (6), Pi gives the mass percentage contributed by the concentration of element i
and is useful for comparison to the data of Gatz (1975a).  Two values of P-j  and A-j
are tabulated from the midlake data.  The first is  for the mean of all  data  sets
collected at midlake, while the second Pi and A-,- values are based only  on those
data sets for which the Chicago/Gary area was indicated as the probable aerosol
source region0  From (7), A-,- is a measure of how significant the fine particulate
(D < "LO ym) component of C-j is in relation to the  coarse particulate (D > 1.0 ym)
fraction.  A rough comparison between the composition of midlake aerosol and source
area aerosol is indicated by a transport composition ratio, R^c, which  is the ratio
between P-j for the Chicago-source midlake data and  P| from Gatz1 (1975a) composite
model  Chicago area aerosol data.  For the majority  of the elements Rtc  appears to
be close to 1.0, which suggests that the trace element composition of the aerosol
is not drastically altered during overlake transport.
                                    163

-------
     Enrichment factors, EF-j, from (8),  relate the  concentration  of an element to
the concentration which might be attributed to a  reference natural  source material.
In this work, the composition of midwestern soil  described by  Bowen  (1966)  is
used, with Al concentration defined as the source standard, Cg^.   EF^ values
close to unity indicate that element is  present in  soil-derived proportions;
large EF-j values indicate other major sources  for that element in  aerosol (Rahn,
1976).  It appears that at least Cu, Zn, and especially Pb are greatly enriched
in midlake aerosol, i0e. come from other than natural  sources.

     The dependence of trace element concentrations at midlake upon surface
layer meteorology and aerosol source region is discussed in detail  by Si even' ng,
et al.(1979a)0  The data "binned" in Tables 11, 12, and^ 13 can clearly_show that
midlake meteorology and source region strongly affect  C^u  Values  o^v^  are  also
                                                        ju
quite strongly dependent upon AT and U5_   The variability of C-jj  and vjj  are of
particular importance to the loading rates,  because these two parameters  are
producted in Xn- calculations (4) and (5)u   For example,  although  the Chicago/Gary-
source data sets give higher IT-j  results  than the Southeast shore  sets,  Xj  values
for the_ two source regions are about equal,  due to the compensating effect of
lower vj and F0 values for the Chicago/Gary  sets.   Both the AT (Table  11 ) and U5
(Table  12) data set aggregations reveal  that meteorological  conditions which lead
to high v^ values were not sampled.   The most unstable bin F0 accounts  for the
0U38 year with AT <-0.9°C; yet, the single  most unstable case sampled  was
AT = -2U0°C.  Cdcj, and thus v,j from (3)  increases  markedly as AT  becomes  more
negative^  If more unstable (AT  <  -2.0°C) data had been collected, the v,j for
this bin would likely have been somewhat higher.  Similarly, the  bin for  high u5
values has an FQ which accounts for the  0,41 year  with u5 < 7.0 m-s"1.  The
highest u5 case sampled was u5 = 8.3 m-s~l.   Again, the available data  sets miss
the cases where higher vj values might have  been determined.   The F0 values used
in the most extreme-case bins thus apply the available data to parts of the year
when loading rates would be greater^  Because of this, the X-j values calculated
here likely underestimate annual X^ „  Given  this consideration and the  2-to 3- fold
uncertainty in the C^d based VH determination, the least X-j estimate from the
overall-mean (4) and bin-method (5)  calculations is divided by two and  reported as
a "minimum" loading estimate.

DISCUSSION

     Two factors in the loading estimate calculations, (4) and (5), were  quanti-
fied through field observations and related  to overlake meteorological  variability:
VH and C-j.  Because al_l sampling was done at a single site, temporal variations
within data-set-mean v
-------
that aerosol number concentration remained fairly constant within each data set.
Variation about the overall set mean was generally less than 20 percent.   The
IN-monitored aerosol mass concentration also showed a similar degree of steadi-
ness.  Each data set thus represents a period of fairly homogeneous meteorologi-
cal and aerosol concentration conditions.  The data set mean v,j and C-j were then
used as input values for the calculations of this work.

     The emphasis, during sampling, upon constancy of parameters within each set
necessitated relatively short Hi-Vol filter run timesu  The gross amount  of
aerosol collected was thus such that 7 elements which were analyzed for by
ICAP-AES failed to appear above L^-j in many data sets (Table 2).  Contamination,
probably introduced by the analytical procedure, makes the data for B unreliable.
This leaves 9 elements as candidates for the A-j calculation; however, data for
2 of these elements must be used with caution.  Because the Hi-Vol  sampler pumps
use Cu-alloy commutators which are known to produce Cu-aerosol  (Moyers, Duce, and
Hoffman, 1972), precautions were taken in setting up the samplers to avoid
re-entrainment of the pump exhaust.  At the 1.18 m^ - min"' flow rate, given the
short run times and low filter matrix load, back-pressure contamination through
the sampler head is unlikely.  There is nothing in the Cu results to directly
suggest any significant contamination actually occurred.  To the contrary, PCU,
ACu, and Rtc results appear to be consistent with other elements (Tablet)-I),,
Even so, any Cu data from Hi-Vol sampling should be used with caution, and the
AQU results are so indicated.  Finally, the Ti data shows some  anomalous  behavior
that suggests probable contamination.  Sievering, et al. (1979a) showed that while
all other elements in this data base correlated inversely with  U5,  Ti was weakly
positively correlated.  In Table C-l, the R^c results clearly set Ti apartu  By
plotting Ti enrichment factors versus Al concentration (Figure  17), Ti is seen to
tend somewhat towards constant concentration (dashed diagonal line)u  All this
suggests the probability of contamination for Ti in collecting, handling, or
analyzing the aerosol samples^  The possibility is not conclusive,  however, and
the AJ.J results are still presented, with caution in interpretation suggested.
For the remaining 7 elements (Al, Ca, Fe, Mg, Mn, Pb, and Zn),  the uncertainty
introduced to the X-j calculation by the C-j factor is expected to be the ICAP-AES
instrumental reproducibility confidence of < 20 percent.

     It should be emphasized that the VQ used in loading calculations (4) and (5)
is derived from meteorological data through the C^^ method and  relation (3)u  This
derivation is based upon an analogy between momentum transfer and mass transfer.
The resultant v^ is a parameterized bulk estimate of the aerosol deposition rate
and not a direct measurement.  Methods which may, after further technique develop-
ment, lead to directly-measured v^'s (such as the concentration-profile method,
or the eddy correlation method) are not yet suitable for routine over-water use
(Hicks, 1979; Shepherd, 1974).  The C^d method was used here in order to  reduce
the order-of-magnitude uncertainty in dry deposition loading estimates from E-J--
based modeling (Winchester and Nifong, 1971; Gatz, 1975b) and bulk  precipitation
sampling (Eisenreich, Emmling, and Beeton, 1977) results.  As stated in a preceding
section, the v^'s used in this work should be viewed in the context of the stated
2-to 3-fold uncertainty.  Still, there are some data highly supportive of the C^-
based results.  Limited ASAS aerosol number concentration profiles  taken  on board
the R/V Simons indicate v^'s close, i.e. within a factor of two, to the C^-based

                                     165

-------
vd values (Sievering, 1979).  Delumyea and Pete! (1977), using a simple mixing-
box aerosol depletion model in a study of phosphorus deposition over Lake Huron,
found that vd for D« 1 ym aerosol  to be 0.6 cm-s~'.  Sievering, et a 1 „ (1979b)
used a similar depletion model and sulfate-aerosol  data taken during May 1977 at
the midlake site and in Chicago to calculate a vd of 0,3 cm-s~^.  For those midlake
samples, 90 percent of the sulfate was associated with the D < 1.0 ym aerosol.
These results agree well with the overall mean Cdd-based vd of 0.5 cm-s"1.

     Another apparent shortfall of the Cdd method is that VH is given as a  bulk
estimate and not as a function of aerosol size.  Theoretical consideration  of
aerosol deposition points to as much as an order of magnitude variation in  vd
in the particle size range 0.1 < D < 3.5 ym; wind tunnel studies of deposition to
a water surface also indicate strongly size-dependent v 7.0 m-s~') extrapolate
the data to more unstable-air and high wind speed periods which were not sampled in
the summer season.  If winter season data had been available, the vd's  for  these

                                           166

-------
extreme-case bins would likely have been somewhat largeru  The extrapolation
through the F0 values thus tends to make AJ values apparent underestimates of
annual loadings^  At the other end of the bins, an assumption of vj = zero for
all cases with U5 < 2.0 m-s~^ also leads to underestimation of annual loadings.
For such low wind speed cases (which include some very stable-air cases), a
laminar sublayer exists at the interface which impedes, but probably does jnot
completely block (Sievering, 1979) aerosol deposition,,  The dependence of C-j
and Vj on U5, AT, and source region is discussed by Sievering, et al. (1979a).
J_he influence of these dependencies upon A-j results is significant,  J_n Table 10,
Cn- tends to generally increase as AT becomes more stable (positive); vj tends to
be reduced.  Because Xn- depends upon the product of these two factors, a "self-
protection" mechanism of the lake surface from high stable-air C-j levels is
implied,,  For the fai 11-winter, unstable-air season this^ mechanism is reversed.
even though lower C-j values might be expected, higher vd's would likely result
in significant winter loadings.   In Table 12 , similar trends appear, although
weakly defined for C-j.  Again, the winter season overlake being the high wind
speed season (NOAA, 1975), summertime "self-protection" of the lake is reversed
in winter.  The source region bin data in Table 1_3 points to another aspect of
the dry deposition_ loading situation,,  Although vj is only weakly variable with
source sector, a C-j maximum is clearly indicated from the Southeast Shore and
Chicago/Gary areas (Figure 1 ),  not an unexpected result,,  The sum of F0 values
show that 53 £ercent of the year these two high-air pollution regions are sources
of increased C-j at the midlake site,,  As a consequence, the Chicago/Gary and
Southeast Shore areas contribute from 60 to 65 percent of the total dry deposition
A-j to the southern basinu   Table C-2 gives these dry deposition A-j values in the
context of the precipitation loadings estimated by Gatz (1975b) from about the
same general  source region.  It  appears that wet and dry route A-j values are
about equal,  except for Pb and Zn, where the dry route A-j are somewhat larger
than for precipitation.  Transfer efficiencies, based on the total  atmospheric
route loading and source strength model  (Gatz, 1975b), range from 4 percent (Cu)
to 14 percent (Zn).   The dry deposition A-j's of this work agree well with the
source strength Et results of Gatz (1975b), again excepting Pb and Zn where
loadings are larger than Gatz predicted.

     The transfer of trace elements from the lake water surface to the atmosphere
is assumed, in (4)  and (5), to be very small compared to the downward flux of
aerosol  to the lake.   For  over 90 percent of the year, wind speed over the lake
is less  than  lOjn-s'1  (NOAA,  1975) and aerosol  generation at the surface is  minimal
At the highest U5 cases observed in the 1977 sampling, no increase in Ca or Mg
content  of the aerosol  was observed (Table 12).  Also, Ca and Mg levels  are  much
reduced  in the overlake-source data sets (Table 13).   Ca and Mg are present  in
fairly high levels  in lake water and one might then expect to see an increase
relative to other elements if lake surface generated aerosol  were present.   The
Rtr value of  2U5 for Ca in Table C-l is probably not a significant indication of
increased relative  Ca;  if  the "all  data  sets"  P-j  value had been used,  Rtc  for
Ca would fall  in line  with the other elements'  values.  Although some  lake-
surface  aerosol  generation does  occur,  it seems reasonable to expect the trace
element  flux  to be  small relative to the atmospheric  input on a long term  average.


                                    167

-------
     The significance  of atmospheric  inputs of trace elements to the Lake Michigan
ecosystem is not well  known  (UC,  1978).   Dry deposition appears to be a major
contributing source of at least  four  elements on a percentage of total loading
basis (Table  14).   10 percent of  Mn,  20  percent of Fe, 30 percent of Zn, and
as much as 60 percent  of Pb  inputs to the lake are by dry deposition of atmos-
pheric aerosol.   Only  for those  elements  which have very large natural land
run-off inputs are  atmospheric loadings seen to be negligible.  More important
to consider than the gross total loadings, however, is the fact that all atmos-
pheric route loadings  are deposited directly to the surface zone of maximal
biological activity.  The impact of this  upon the well being of the biota must
surely be an object of future research, as must be the continuation of the task
of precisely quantifying atmospheric  route loadings to Lake Michigan, other large
lakes, and the oceans„
                                          168

-------
                                         Mid-Lake Michigan
en

ELEMENT
AI
Ca
Cu
Fe
Mg
Mn
Pb
Ti
Zn
All
Pi(%)
0.6
3.
0.02
1.
0.8
0.05
0.5
0.03
0.2
Data Sets
Ai
1.3
0.5
3.9
1.7
0.7
2.2
21.
0.7
6.4
EFi
=1.0
22.
170.
3.5
23.
10.
5200.
1.4
530.
Chicago Source Sets
P1(%)
2.
5.
0.09
3.
1.
0.1
0.9
0.06
0.7
Ai
0.6
0.4
2.9
1.3
0.6
1.2
13.
0.8
5.8
EFi
=1.0
11.
280.
3.2
8.6
6.0
3300.
0.9
560.
	Chicago	
Composite Aerosol t
   2.
   2.
   0.1
   3.
   0.9
   0.1
   0.6
   0.01
   0.6
!ic_
1.0
2.5
0.9
1.0
1.1
1.0
1.5
6.0
1.1
          t  Gatz, 1975a
       TABLE  C-l.   Aerosol  Trace Element Composition and Enrichment Factor.

-------
                 Percent of Dry Loading
                                                     Atmospheric Route Loading, (10  kg - yr" )
*   Precipitation Loadings and Area Source
    Strengths from Gatz (1975b)

**  Results in brackets {  } indicate caution
    in interpretation - see Discussion
Et*

Al
Ca
Cu
Fe
Mg
Mn
Pb
Ti
Zn
Total due to Chicago/N. W, Indiana
80
80
{65}
75
60
80
75
{60}
85
Dry Deposition
480
1600
{12}
750
440
50
370
{20}
170
Precipitation
560

20
750

50
90
60
50
Total
1040

{32}
1500

100
460
{80}
220
Percent
13

{4}
7

13
8
{9}
14
               TABLE C-2.   Atmospheric Route  Loadings  to the  Southern  Basin
                            of Lake  Michigan  Due to  the Chicago Shore/Northwest
                            Indiana  Source  Region.

-------
                                 APPENDIX D

                    AEROSOL SAMPLE PREPARATION PROCEDURE
                     FOR EMISSION SPECTROSCOPY ANALYSIS
EQUIPMENT
     1  Plasmod Low Temperature Asher
     1  D2A Vacuum Pump, 61  1/min
     1  Oxygen Cylinder with Two-Stage Regulator
     1  Hot Plate
     1  Top Loading Balance
     4 or 8 Quartz Combustion Podats, each with a permanent identifying  mark
SUPPLIES
     1  pair Teflon Coated or Plastic Scissors (dedicated)
     1  Oxford Pipet, 5 ml, with disposable oipet tips
     1  Eppendorf Pipet, 50 A (yl) with disposable pipet tips
     1  Graduated Cylinder, 50 ml
       60 ml  Polyethylene Bottles with caps (Monsanto), 5  for each filter set
       Disposable Plastic Funnels, 1 for each bottle
       Petri  Dishes, 100 X 100 X  15 mm square style, 1  for each filter set
       Glass  Rods, 3 mm diameter, "L" shape, 2 for each petri dish
     1  Glass  Plate
     2 Pair Forceps, Teflon coated or plastic
       Disposable Plastic Gloves  and ruler are also needed

REAGENTS

       Yttrium Nitrate, Purified  Crystals
       Deionized Water or Doubly-Disti lled-in-Glass VJater
       Redistilled-in-Glass Nitric Acid or Spectrograde Nitric Acid
       Spectrograde Hydrochloric  Acid

PREPARATION OF YTTRIUM STOCK SOLUTION:  1000 pprn

1.  Acid wash a 1 liter volumetric flask,,  Rinse several times with deionized
    water.  Dry flask in oven and let cool  to room temperature.

2U  Weigh a clean, dry weighing boat or bottle to the nearest 0,,1  mg,,   Weigh
    out 4.3080 g of Y(N03)3 • 6H20,  Quantitatively transfer the Yttrium
    Nitrate to the 1  liter volumetric flask using a cleaned powder funnel  and
    a wash bottle of deionized water.   Dissolve  all  of  the Y(M03)3 • 6H20  and

                                    171

-------
then dilute to mark with deionized water.   Store  the  1000  ppm  Yttrium stock
solution in a labeled,  clean (acid washed)  bottle.

REDISTILLED ~ 6N HN03

1.  Clean and assemble  an all  glass distillation  set-up.   Use  an electric
    heating mantle to heat the distillation flask.

2,,  Add a 50 percent solution  of concentrated  reagent grade  HN03 in  distilled
    water until the flask is half filled.

3U  Adjust the current  to the  mantle until  the nitric acid solution  just begins
    to boiK  Collect the condensate in a  cleaned,  labeled reagent bottle.
    Distill the solution down  to approximately one-fourth  of the original
    volume.  Add more of the 50 percent solution  of nitric acid  and  distill.

PROCEDURE I-A.  PREPARATION OF INERTIAL CASCADE IMPACTOR  FILTERS

    The cascade and backup filters are stamped with a five digit filter identi-
fication number.  The trace metal samples  for  ICAP  analysis  will be  identified
by a six digit number,,   The samples from the cascade  impactor  filters will  be
identified by the filter number plus a zero (0).  The backup filter  will be
divided into three samples.  The undisturbed portion  of the  backup filter will
be numbered with the filter number plus a  "-1".   The  samples from the portion
of the backup filter used in the extraction procedure will be  numbered with
the filter number plus  a "-2"  and the soluble  portion from the extraction will
be identified as the filter number plus a  "-3",

    The cascade filters and backup filter  are  stored  in a  filter set container.
Each filter within the  filter  set is numbered.  The cascade  filters  are approxi-
mately 5V X 6" with 11 impaction strips,   10  of  the  strips  are  impacted with
atmospheric particulates.,  The narrower end strip will  be  eliminated, since  it
is not impacted on.  The 10 impacted strips are to  be cut  and  ashed.  It is
imperative that the cutting procedure be performed  in a clean  work area, A  pair
of Teflon coated scissors should be kept solely for use on this  procedure,

1.  Clean the quartz boats by  adding ~5 ml  of  6N  HN03 to  each.  Place boats  on
    the hot plate and heat until acid begins to fume,,  Gently  (and carefully)
    swirl acid around inside surfaces.  Pour acid out and  rinse  several times
    with deionized water,,  Wipe inside dry with Kimwipes,  Do  not touch inner
    surfaces of boats after cleaning,

2.  Clean the pair of scissors with acetone prior to  cutting a sample.

3,  Record, along one row of the Plasma Asher  Log,  the date  and, in  the column
    designated for the  quartz boat, the filter number (see Figure D-l ).

•k  Wearing disposable  plastic gloves, cut the filter along  lines as indicated
    in Figure D-2.

                                     172

-------
    a)  Cut the filter along dashed line marked "cut !"„  Dispose of the
        numbered end strip.

    b)  Cut along the dashed lines marked "cut 2 and cut 3".  Dispose of the
        two side stripsu

    c)  Cut along the dashed line marked "cut 4".  10 impacted strips and a
        numbered end strip will remain.

    d)  Cut the 10 impacted strips in half.

5,,  Place the 20 impacted strips in the appropriate quartz boat.  Dispose of the
    numbered end strip0

6.  Repeat steps 2 through 5 for the other cascade filter within the set.
    NOTE:  If the ashing chamber can accommodate four quartz boats, repeat
    steps 2 through 5 for the cascade filters in the next filter set.

7.  Proceed to Procedure II - Low Temperature Ashing.  After ashing all  of the
    inertial  cascade impactor filters, proceed to Procedure I-B,

PROCEDURE I-B.  PREPARATION OF HI-VOLUME BACKUP FILTERS

    The backup filters are 8" X 10".  The exposed area is approximately 7" X 9"u
A 3-5/16" X 7" portion (cut into two equal  pieces) of the backup filter is used
in a soluble aerosol extraction procedure.   The extraction is performed and then
the extracted portion and the undisturbed portion of the backup filter are pre-
pared for ashing.

1.  Write the filter number and "-3" on the external  surfaces of the top and
    bottom of an unused square petri dish.   Use a grease pencil  to write number.
    Cover the number with an adhesive label.  (The numbers should be readible
    and legible when the two halves of the  dishes are resting on their external
    surfaces.)

2,  Acid clean (using diluted HN03) the "L" shaped glass rods.   Rinse thoroughly
    with deionized water and allow to dry in a clean environment.  Put two of the
    rods in each of the larger half of the  labeled petri dishes,,   Arrange the rods
    so that they form a square.  Set aside  the smaller half of the petri dish,

3.  Add 29 ml  of deionized water to each dish.  NOTE:  pH can be measured on a
    dish set  up as a blank,,

4.  Clean the pair of scissors prior to cutting the filters,

5.  Measure along  both sides of the exposed area of the filter  a distance of
    3-5/16" (see Figure D-3).   Cut along the line connecting the  two points.

6.  Put the larger piece (with number intact) back in the filter set container,

                                    173

-------
 7.   Measure 3-1/2"  along  the  bottom  edge  of  the exposed area of the smaller
     piece.   Cut the piece of  filter  in  half  at that  point.

 8.   Cut off the borders.   Two 3-5/16" X 3-1/2" pieces  should remain.

 9.   Put one of the  3-5/16" X  3-1/2"  pieces of the  backup  filter back  in  the
     filter set container.

10U   Using two pair  of forceps,  gently place  the other  3-5/16"  X 3-1/2" piece
     and filter onto the surface of the  water in the  petri dish correspondingly
     numbered to the backup filter.   NOTE:  The filter  should just  rest upon
     the glass rods  and should not be submerged nor should the  edges of the
     filter touch or drape over the side of the petri dish.

11.   Repeat steps 4-10 for the remaining backup filters.

12.   Cover the array of petri  dishes  with  a glass plate and  let them remain
     undisturbed for 24 hours.

13.   Carefully remove the  piece of filter  paper after 24 hours„  Use the  two
     pair of forceps„  Place the wet  piece or filter  in the  correspondingly
     numbered smaller half of  the petri  dish.

14.   Repeat step 13  for the remaining samples0  Set them aside  to dry  in  a clean
     area.  If possible, cover filters without touching them.

15U   Remove from the filter set container  the other 3-5/16"  X  3-1/2" section of
     the backup filter.  Carefully place it on the  surface of  the liquid  in the
     correspondingly numbered  petri dish.

16U   Repeat step 15  for the remaining filters.

17,   Cover the array of petri  dishes  with  a glass plate and  let them remain
     undisturbed for 24 hours.

18U   Place the dry filter sections from  step  14 into  their correspondingly
     numbered filter set containers.

19.   Repeat steps 13, 14 and 18 after the  second 24-hour extraction.

20.   Number a sample bottle corresponding  to  the first  petri  dish.

21.   Tare the bottle on the top loading  balance,,

22U   Carefully pour the contents of the  petri dish  into the  bottle  through a
     clean, unused plastic funnel.  The  glass rods  will adhere  to the  surface  of
     the petri dish.  Carefully remove  them with the  forceps and knock in the
     last drops from the dishu


                                        174

-------
23.  Set the bottle back on the balance.   Uash down with deionized water  any
     residual drops from the funnel,  (NOTE:   do not weigh the  funnel  — hold
     it up from contact with the bottle)  and add water until  the  weight
     registers 25 g, i.e. 25 ml of solution.  Measure and record  the  pH
     (be careful not to contaminate the sample).

24.  Cap the bottle and set aside.

25.  Repeat steps 20-24 for the remaining liquid samples.

26.  Record the filter number and "-2"  in the plasma asher log for the first
     backup filter.

27.  Remove the two (dry) extracted portions of the backup filter and cut them
     into strips small enough to fit into quartz boats.

28.  Put all of the strips in a clean (see Step 1  of Procedure I-A),  identified
     quartz boat.

29.  Record in the plasma asher log the backup filter number and  "-I".  Remover
     from the filter set container the undisturbed portion of  the backup  filter.
     Cut off and dispose of the borders.   Cut into strips small enough to fit
     into quartz boats.  Put all strips in a clean, identified quartz boat.

30.  Proceed to Procedure  II - Low Temperature Ashing.

PROCEDURE  II -  LOW TEMPERATURE ASHING IN TEGAL PLASMOD ASHER

 1.  Place  two  or  four boats into the inner glass chamber of the  low temperature
     asher.

 2.  Make  sure  that the oxygen tank, regulator, vacuum pump, and  low temperature
     asher  are  properly connected.

 3.  Turn on the oxygen supply and regulate to 10 psi.

 4.  Turn on vacuum pump.

 5.  Turn on the asher (push in AC button) making sure that the RF power  switch
     is OFF  (OFF is the down position).

 6.  While  holding the glass chamber up against its casing, turn  on (switch up)
     the vacuum switch.  A  tight  vacuum must be obtained for the  asher  to operate
     properly.

 7.  Shut the screen door,  turn on the RF power switch, and set RF level  to 75 watts,
     Adjust  the tuning knob until a minimum reading is obtained on the  RF power
     meter.  A  faint  blue  or pink hue of excited oxygen will appear in  the chamber.
     Turn RF power to maximum value and retune to obtain a minimum signal.
                                    175

-------
      Ashing times  are  related  to  the  amount  of  filter  paper  being  ashed.
      Complete destruction  of all  organic material  is required  for  proper
      sample preparation.   Complete  ashing  can be identified  by a very  faint
      blue glow after a period  of  a  bright  blue  glow  (the  bright blue corre-
      sponds to excited
  9.   Mhen the filter has  been  ashed,  turn  the  RF  level  down  to  zero  and  switch
      the RF power to OFF.   The door may  now  be opened  to  inspect  the filter.
      If the filter has  not been completely ashed,  repeat  step 7.

 10.   Turn off the vacuum  switch,,   Allow  the  purge  line  to  fill  the chamber
      with air and then  take out the inner  glass chamber.   If no more filters
      are to be ashed, turn off vacuum punp,  asher,  and  oxygen supply.

 11.   Place quartz boats on a clean asbestos  square  that is resting on a  hot
      plate (in hood).  Using an oxford pipette, add to  each  boat  5 ml  of
      redistilled 6N HN03.   Heat for approximately  10 minutes.   Do not let
      solution spatter.   Insure contact of  all  surfaces  by  gently  (and carefully)
      swirling the acid.  Let the boats cool.   Insure that  all of  the material is
      solubilized.

 12.   Label 2 or 4 bottles  with the appropriate filter  identification numbers.

 13.   Tare the first bottle on  the top loading  balance.

 14.   Quantitatively transfer the contents  of the  first  boat  into  the first
      bottle.  Be sure to  rinse the underside of the lip of the  boat.
      NOTE:  Many small  rinses  are better than  one  large rinse.  Use  a clean,
      unused plastic funnel in  the transferring process.

 15.   Set the bottle back  on the balance  and  wash  down  the  funnel  until a weight
      of 25 g is registered. Again note  that the  funnel should  be held out of
      contact with the bottle.

 16.   Repeat steps 13-15 for the remaining  boats.

 PROCEDURE III.  YTTRIUM SPIKE

     When all of the filters from one  outing  -- cascade  and backup ~ have  been
ashed and dissolved and  all extractions completed,  the  Yttrium spike  can  be added.

  1.   Arrange the bottles  according to filter sets  (5 bottles per  set).

  2.   Open the first bottle.

  3.   Pipette 50 X (yl )  of the  1000 ppm Yttrium stock solution into the bottle,,

  4,   Cap the bottle and place  identifying  mark to  note  fact  that  Yttrium spiking
      was done.

  5.   Repeat steps 2 through 4  for the remaining bottles,


                                        176

-------
                                     BOAT  3   |     BOAT  4
                                     FILTER #      FILTER #
FIGURE D-l.  Example of Plasma Asher Log.
                  177

-------
CUT

             FIGURE  D-20   Cuts  Made  During  Preparation of
                          Inertial Cascade  Impactor  Filter.
                               178

-------
CUT 6
                              CUT 6
                             Exposed
                             Area of
                             Filter
 CUT 3
CUT 2
CUT 3
            FIGURE D-3.   Cuts  Made  During  Preparation
                         of Hi-Volume  Filter.

                              179

-------
                                  TECHNICAL REPORT DATA     .
                           (Please read Instructions on the re\ene before completing)
 REPORT NO.
 905/4-79-016
                                                          3. RECIPIENT'S ACCESSION'NO.
 TITLE ANDSUSTITLE
 An Experimental  Study of Lake Loading by Aerosol
 Transport  and  Dry  Deposition in the Southern Lake
 Michigan Basin	
                                                          !)afe of Preparation:  July 1979
                                                          6. PERFORMING ORGANIZATION CODE
 AUTHOR(S)
 Herman Sievering,  Mehul  Dave, Donald A,
 Richard  L.  Hughes, Patric McCoy
                                                          8. PERFOI
                                                                             VTION REPORT NO.
                                          Dolske,
 PERFORMING ORGANIZATION NAME AND ADDRESS
 College  of  Environmental  & Applied Sciences
 Governors State University
 Park  Forest South,  Illinois  60466
                                                          10. PROGRAM ELEMENT NO.
                                                          2BA645
                                                          11. CONTRACT/GRANT NO.
                                                          R00530101
2. SPONSORING AGENCY NAME AND ADDRESS
 Great  Lakes  National Programs Office
 U.S. Environmental  Protection Agency, Region  V
 536 South Clark Street
 Chicago,  Illinois  60605
                                                          13. TYPE OF REPORT AND PERIOD COVERED
                                                          Final  6/1/76  - 7/31/79
                                                          14. SPONSORING AGENCY CODE
                                                          EPA-OGLNP
                                                          Office of  Great Lakes National
                                                          Programs	
 5. SUPPLEMENTARY NOTES
  Additional  support received from the National  Center for Atmospheric Research,
  Research  Aircraft Facility, Boulder, Colorado  80307   	^	^^
  and  $04 were obtained.  A strong
  in the  variability of all twelve
  ence upon wind speed was found.
6. ABSTRACT
 A Lake Michigan experimental  program to assess the contribution  to  Great Lakes loading
 by atmospheric transport  and  dry deposition of aerosol was conducted.   A midlake and
 nearshore trace element and nutrients data base with associated  meteorology capable of
 establishing a climatology for mass transfer to Lake Michigan was collected during
 1977 and 1978.  Significant data for Al, Ca, Cu, Fe, Mg, Mn, Pb, Ti,  Zn, total P, N03
                                   linear dependence upon atmospheric thermal stability
                                   aerosol constituents was found, but no linear depend-
                                   Source region and wind direction dependence are cause
 for as much as an  order of magnitude variation in concentrations.   Aircraft-based
 meteorological and aerosol data were collected in June and September,  1977 and May,1978

 Bulk deposition velocities as a function of overlake climatology were used to calculate
 dry deposition atmospheric loadings to Lake Michigan.  Limited aerosol  number profile
 measurements of the deposition velocity confirmed a three-fold or smaller uncertainty
 in the bulk deposition velocity estimates.  Minimum dry deposition  loading estimates
 compared to precipitation and surface runoff show that atmospheric  inputs by dry
 loading are at least 60%  for  Pb, 30% for Zn and over half the total  sulfate and nitrate
 input.  Dry loading of total  P is approximately equal to precipitation loading.
17.
                                KEY WORDS AND DOCUMENT ANALYSIS
                  DESCRIPTORS
                                              b.lDENTIFIERS/OPEN ENDED TERMS
                                                                           COSATI Field/Group
  GREAT LAKES
  Aerosols
  Meteorology
  Lake Michigan
  Trace Elements
  Fallout
  Air Pollution
                          Water Pollution
                          Lake Breeze
Atmospheric Loading
Dry Deposition
Aerosol Spectrometry
Nutrients
ASAS/Knollenberg
12A
07D
10B
13B
13. DISTRIBUTION STATEMENT

  Release Unlimited
                                              19. SECURITY CLASS (ThisReport)
                                              Unclassified
                                                                         21. NO. OF PAGES
                                             J20. SECURITY CLASS (Thispage)
                                             \ Unclassified
                                                                         22. PRICE
EPA Form 2220-1 (9-73)
                                             180
                                                               U. S. GOVERNMENT PRINTING OFFICE: 1980-653-168

-------