September 27,1991
        Office of Air Quality Planning and Standards
        Research Triangle Park, North Carolina 27711

Project MOHAVE

    Study Plan
U.S. Environmental Protection Agency
     September 27, 1991

              TABLE  OF CONTENTS

List of Figures	  iii

List of Appendices	•  iv

List of Acronyms  	  v

1. Introduction 	  1
      Reason for the Study	  1
      Goals of the Study	  2
      Project MOHAVE Organization	  2
      Study Planning and Review to Date .,	  3
      Study Schedule	  4
      Plan Organization	  5

2. Current Knowledge and Available Data  	  6
      Setting 	  6
      Transport Regimes	  7
      WHTTEX	  10
      SRP Study	  11
      MPP Emission Modulation Studies	  12

3. General Field Study Design	  14
      Selection of the Intensive Periods	  14
      Siting  of Monitoring Instrumentation	  17

4. Tracer	  23
      Purpose	  .  23
      Choice of Tracer	  23
      Tracer Release  	  25
      PFT Programmable Samplers 	  26
      Tracer sample analysis	  .  27

5. Air Quality Monitoring  	  32
      IMPROVE Samplers	  32
      DRUM Samplers	  33
      High Volume Dichotomous Samplers	  34
      Hydrogen Peroxide Measurements 	  34
      Methylchloroform Measurements	  35

6. Meteorological Monitoring  	  37
      Background	  37

      Objectives  	   37
      Field Study Plan  	   39
      Data Collection	   40
      Data Quality Assurance	   40
      Data Processing and Analysis  	   41

7.  Optical Monitoring	   42
      Overview  . . .	   42
      View Monitoring	   43
      Electro-optical Monitoring  	   44
      Monitoring Locations  and Sampling Frequency	   44

8.  Emission Inventory and Characterization	   47
      Purpose	   47
      Review of Existing  Data and Inventories	   47
      MPP Stack Sampling	   47

9.  Centralized Data Management and Validation	   48
      Overview	   48
      Aerosol Sampling (UC-Davis)	   49
      Transmissometer Data	   49
      Radar wind profilers and RASS	   50

10. Descriptive Data Analysis and Interpretation	   52
      Goals	   52
      Descriptive Statistics	   52
      Extinction Budget	   52
      Empirical  Orthogonal Function Analysis	   53
      Meteorological Classification	   54

11. Attribution   	   56
      Overview	   56
      Deterministic Meteorological Modeling  	   57
      Transport, Chemical and Deposition Modeling	   59
      Hybrid and Receptor Modeling  	   60
      Extrapolation of Intensive Study Periods to the Long-Term  ....   61
      MPP Emission Modulation Study	   62
      Framework for Interpreting Results	   63

12. Overall Quality Assurance	   65
      Approach	   65
      System Audits - Study Planning and Preparation	   65
      Measurement System  and Performance Audits	   66

References	   68

                   LIST  OF  FIGURES
Figure 1.    Map of the southwestern U.S. illustrating location of MPP.

Figure 2.    Synoptic flow patterns of concern.
            a. Dry summer southwesterly flow
            b. Summer monsoonal flow
            c. Winter storms

Figure 3.    Dri Mountain quarterly wind roses from 1990.

Figure 4.    Las Vegas moisture climatology
            a. Average specific humidity by month: 1951-1980.
            b. Average relative humidity by month: 1951-1980.
            c. Average percent daytime cloud cover by month: 1921-1950,
               1951-1980 and  1921-1980.

Figure 5.    Mean monthly dew point at Dri Mountain:  1982-1990

Figure 6.    Mean (resultant) wind direction by month at Dri Mountain: 1976-

Figure 7.    Location of air quality and tracer monitoring sites.

Figure 8.    Chromatogram of a 20 L sample of ambient air.

Figure 9.    Schematic of IMPROVE sampler.

Figure 10.   Meteorological modeling grids.

                 List of Appendices

1     Project MOHAVE Summary Table

2     Conceptual Plan

3     Project MOHAVE Planning Workshop Participant List

4     Scoping analysis of potential MPP impacts at GCNP

5     MPP Emission Modulation Study Plan

6     Colorado State University Regional Atmospheric Modeling System

7     Washington University cooperative agreement proposal: Development of
      an Interactive Data Analysis Tool Using the Monte Carlo Model

8     Project MOHAVE Response to NAS WfflTEX Comments

9     CMB, TMBR, and DMB Model Formulations

                   List of Acronyms
ABL        Atmospheric Boundary Layer
ARS        Air Resource Specialists
BNL        Brookhaven National Laboratory
CD4        Deuterated methane
CMB        Chemical Mass Balance
DMB        Differential Mass Balance
DRUM      Davis Rotating-drum Universal-size-cut Monitoring
ECD        Electron Capture Detector
EOF        Empirical Orthogonal Function
EPA        U.S. Environmental Protection Agency
EMSL      Environmental Monitoring  Systems Laboratory
GC         Gas Chromatography
GCNP      Grand Canyon National Park
INAA       Instrumental Neutron Activation Analysis
LIPM       Laser Integrating Plate Method
LQL        Lower Quantifiable Limit
LOD        Limit of Detection
MOHAVE   Measurement of Haze and  Visual Effects
MPP        Mohave Power Project
NAS        National Academy of Sciences
NGS        Navajo Generating Station
NOAA      National Oceanic and Atmospheric Administration
NPS        National Park Service
OAQPS     EPA Office of Air Quality Planning and Standards
PDCB      Perfluorodimethylcyclobutane
PDCH      Perfluorodimethylcyclohexane
PESA       Proton Elastic Scattering Analysis
PFT        Perfluorocarbon Tracer
PIXE       Particle Induced X-ray Emission
PMCP      Perfluoromethylcyclopentane
QC         Quality Control
RASS       Radio Acoustic Sounding Systems
SCE        Southern California Edison
SF6         Sulfur hexafluoride
SO2        Sulfur dioxide
SO4        Sulfate
SRP        Salt River Project
TMBR      Tracer Mass Balance Regression
WHTTEX    Winter Haze Intensive Tracer Experiment
XRF        X-Ray Fluorescence

1.   Introduction

      Reason for the Study

      In 1977, in Section 169A of the Clean Air Act, Congress set as a national
goal,  "the  prevention  of  any future,  and the remedying of any  existing,
impairment of visibility in  mandatory Class I Federal areas which results from
manmade  air  pollution."   Section  169A also required EPA  to promulgate
regulations  to assure reasonable progress toward meeting the national  goal for
mandatory Class I areas where visibility is an important air quality related value.
On November 30, 1979, EPA identified  156 areas, including Grand  Canyon
National Park (GCNP), where visibility is an important air quality related value.
On December 2, 1980, EPA promulgated the required visibility regulations.  In
broad outline,  the visibility regulations require the States to coordinate  their air
pollution control planning activities with the appropriate Federal Land Managers
to develop a program to assess and remedy visibility impairment from new and
existing sources.
      More recently, Congress reaffirmed its desires to address visibility issues
by adding section 169B to the Clean Air Act amendments of 1990. Section 169B
calls for a substantial research program to  study regional haze, and requires the
Administrator of EPA to establish a visibility transport commission for the region
affecting the visibility of GCNP.
      In January and February,  1977,  the National Park Service (NPS), acting
in its capacity as the Federal Land Manager for GCNP, conducted a study known
as the Winter Haze Intensive Tracer Experiment (WHTTEX). WHTTEX involved
a six-week long intensive monitoring period during which an artificial tracer was
released from the Navajo Generating Station (NGS) northeast of GCNP. National
Park Service analysis of optical, air quality and meteorological  data indicated a
significant fraction of the  haze in GCNP during this time period was due to
sulfates  resulting from NGS emissions (Malm et al, 1989).
      Salt River Project  (SRP), the operators of Navajo Generating Station,
conducted  a  study  during early 1990.   The SRP study also indicated  a
contribution of NGS emissions to haze  in GCNP,  but at a lower  frequency of
occurrence.  A difference in   prevailing  meteorological conditions during the
years of the NPS and  SRP  studies would at least partially  account for the
differences in magnitude and frequency of impacts identified by the two studies.
The results and limitations of  the NPS and SRP studies are described briefly in
section 2.
      Based on  these  studies and additional evidence presented, EPA  has
proposed regulations that would  require substantial reduction of sulfur dioxide
emissions from NGS.  While  NGS has  been linked to a portion of the haze at
GCNP,  it is generally recognized that a number of other area and point sources
also contribute to haze at  GCNP.  One potential source is the Mohave Power


Project (MPP), a 1580 Megawatt, coal-fired steam electric power plant located
in Laughlin,  Nevada, southwest of  GCNP  and operated by the  Southern
California Edison Company (SCE).  Like NGS, MPP has no pollution control
equipment for sulfur dioxide.  Congress, desirous  of additional information
concerning the sources of visibility impairment in GCNP,  added $2.5 million to
the fiscal 1991 appropriation for EPA to conduct "a pollution tracer study at the
Mohave Powerplant".  Project MOHAVE (Measurement Of Haze And Visual
Effects) is EPA's response to this  congressional mandate.

       Goals of the Study

       The primary goal of Project MOHAVE is to determine the contribution
of the MPP to haze at GCNP and other mandatory Class I,areas where visibility
is an important air quality  related value.  This implies a quantitative evaluation
of the intensity, spatial extent, frequency, duration and perceptibility of the MPP
contribution.  The  improvement in visibility that would resijlt from control of
MPP emissions is  included in the primary goal.  Secondary goals include an
increased  knowledge of the role  of other sources on haze in GCNP and the
southwestern United States in general.  Because knowledge of regional transport
and  air quality levels  is necessary to separate the effect of MPP from other
sources, meeting the primary goal will result in increased knowledge  about the
impacts from other sources.
       It is  hypothesized  that the maximum impacts of  MPP on visibility at
GCNP occur during periods with  clouds present (to facilitate transformation of
SO2 to sulfate) and  wind directions that transport the MPP  plume toward GCNP.
The study is designed to test this hypothesis.

       Project MOHAVE Organization

       The EPA Office of Air Quality Planning and Standards (OAQPS) in
Durham,  North  Carolina has overall management  responsibility for Project
MOHAVE.   Robert Bauman is  the manager of Project MOHAVE. and has
selected staff from the Environmental Monitoring Systems Laboratory (EMSL)
in Las Vegas as the technical advisors.  Staff includes Marc Pitchford, a National
Oceanic and Atmospheric Administration (NOAA) employee assigned to EPA and
Dr. Mark Green, a Desert Research Institute (DRI) employee working under a
cooperative agreement with EPA. To be advised in the overall direction of the
study, Mr. Bauman has formed a steering committee composed of government
and industry scientists. The steering committee includes:

             Dr. Carol Ellis     Southern California Edison Company
             Dr. William Malm   National Park Service
             Dr. Peter Mueller   Electric Power Research Institute
             Marc Pitchford     EPA (EMSL-LV)

             Dr. William Wilson EPA (AREAL)

      Temporary technical advisory panels provided recommendations during a
planning workshop, as discussed later in this section.  Coordination committees,
composed of Project MOHAVE participants and their contractors responsible for
various  components of the study, will meet on an ad hoc basis  to refine and
coordinate in the following areas:
      (1)    Monitoring
      (2)    Modeling
      (3)    Data Management
      (4)    Data Analysis

These committees will facilitate joint analyses with SCE and other contributing
participants.  The participants in Project MOHAVE  include Federal agencies,
universities and private companies. A list of the main participants and their areas
of responsibility is given in the summary table presented in Appendix 1.

      Study Planning and Review to Date

      The first significant planning effort was the formulation of a conceptual
study plan. The conceptual plan outlined the main components of the study and
gave generalized approaches for each aspect of the study. Preliminary monitoring
locations and schedules were also identified. The purpose of the conceptual plan
was to serve as a preliminary planning document to provide a common starting
point for outside review.  The conceptual plan was reviewed by (1) the Project
MOHAVE steering committee, (2) members of the Haze in National Parks and
Wilderness  Areas Committee  of  the  National Research  Council,  National
Academy of Sciences (3) participants in a Project MOHAVE planning workshop
(a group of about 40 experts), and (4) various other individuals. The conceptual
plan underwent several revisions; the most recent version,  which led to  the
current plan, is presented in Appendix 2.
      The Haze in National Parks and Wilderness Areas  Committee was briefed
on the conceptual plan on March 14, 1991 at the University of California-Irvine.
Individual members of the Committee asked clarifying questions and made some
suggestions on  the conceptual  plan. Several  of  the members made additional
comments at later dates.  The  Committee as  a whole did not comment on the
      During the week of April 23, discussions were held between SCE, DRI,
and EPA in Las Vegas to formulate conceptual models of conditions during which
MPP emissions may be transported  to GCNP. This included a  review of the
dynamic processes affecting MPP plume transport and  dispersion, and the diurnal
and  seasonal  variation of these processes.   Alsoj considered  were issues
concerning chemical transformation and deposition, in particular gas-phase and
aqueous phase oxidation and the roles of clouds and H2O2. These discussions and

a summary of the meeting provided by SCE and DRI helped in selecting the
intensive study periods as well as providing insight about the important physical
       A planning workshop was held April 30-May 2 in Denver.  Thirty-nine
individuals with expertise in one or more study components attended. A plenary
session was held first during which the conceptual plan was presented. Following
the plenary  session, subgroups met to make recommendations on the study
components.    The subgroup  topic  areas  were:  1) tracer,  2)  air quality
measurements,  3)  emissions,  4)  deterministic  modeling   and  upper  air
measurements,  and  5) quality  assurance.   Another plenary session followed,
during which clarifying questions were asked and different subgroups coordinated
their  plans.    The subgroups  again  met  to compile  recommended study
components; these were presented in a final plenary session.  After the workshop,
a small group met to evaluate the recommendations and plan the implementation
of the study.   A list of the participants attending  the workshop appears in
Appendix 3.
       In July 1991, a table summarizing the main components of the study and
the responsible persons for each component,  and a map showing expected
monitoring locations were prepared.  These were sent out to study participants.
The purposes of the summary table and map were to provide an update on the
plan and to ensure that the Project MOHAVE staff and other study participants
had a mutual understanding of the responsibilities and plans for each study
component.  The  summary table was updated after review by participants. It is
presented in Appendix 1.  More  detailed descriptions of the information in the
summary table appear in subsequent sections of this plan.

       Study Schedule

       The field measurement portion of the study will last for one year, from
September 1991 through August 1992.  Intensive monitoring and tracer release
periods are scheduled for January 4-31, 1992 and July 15-August 25, 1992.  A
list of milestones of and anticipated dates of completion major operational phase
is given below.  Coordination,  data review,  and planning meetings will  be
scheduled as appropriate.

             MILESTONE                   DATE

Deploy year-round monitoring equipment        9/91

Deploy winter intensive equipment               11/91-12/91

Winter intensive study                          1/92

Begin data processing                          3/92

Preliminary analysis of winter intensive          5/92

Deploy summer intensive equipment             6/92

Summer intensive study                         7/92-8/92

End monitoring                                9/92

Preliminary analysis of summer intensive         12/92

Receive final monitoring and modeling data      3/93

Draft report                                   7/93

Final report                                   12/93

       Plan Organization

       This plan is composed of  12 sections and  8 appendices.   Section 2
discusses current knowledge, including recent tracer studies and data available for
further study.  Section 3  provides an overview of the field study design in terms
of monitoring locations and schedules. Section 4 describes the tracer aspects of
the  study.  Sections 5-7 discuss  the air quality, meteorological,  and optical
monitoring  plans.   Emission inventory and source characterization are outlined
in Section  8.   In  Section 9, data management and validation are discussed.
Section  10  details the  descriptive data  analysis   and  interpretation study
components. The methods of attribution to be used appear in Section 11.  Section
12 describes the overall quality assurance plan.

 2.  Current  Knowledge  and Available Data


      MPP is located at Laughlin,  NV, about 125 km south-southeast of Las
Vegas, 350 km northeast of Los Angeles, and 340 km northwest of Phoenix (see
Figure 1).  The MPP is a coal-fired, base loaded generating facility with a 153
m high stack.  The base of the stack is at 210 m msl. It uses low sulfur (0.6%
by wt.) Arizona coal delivered by slurry pipeline. Its SO2 emission rate averages
about 150 tons/day at full operation  (Nelson, 1991).
                                                       GRAND CANYON
                                                       NATIONAL PARK £/\  A
                                                                 A-..  A
 Figure 1. Map of the southwestern U.S. illustrating location of MPP.

       The  topography  in  the vicinity of the MPP  is complex with sparse
 vegetation.  A portion of the Colorado River Valley, the Mohave Valley, lies to
 the north of the MPP between Davis and Hoover Dams.  The Mohave Valley is
 bordered on the west by the El Dorado and Newberry Mountains and on the east
 by the Black Mountains.   Long  north/south oriented valleys lie to  the east
 (Detrital Valley) and west (El Dorado Valley) of these ranges.
       The Mohave Valley walls are not symmetric with respect  to the valley
 axis.  Western slopes rise gradually, while eastern slopes rise slowly for the first
 few kilometers with steep walls further to the east.  The border between Nevada
 and Arizona also extends along the valley axis. The bottom of the valley is about

200-300 m msl and the ridges reach 1200 m msl. Toward the west, the Mohave
Valley extends into  a  high plateau and toward the east into the Detrital Valley
plateau (600 m msl). The Mohave Valley narrows significantly as it approaches
Hoover Dam.  At Lake Mead the terrain flattens. The western entrance to GCNP
is at the end of the eastern arm of Lake Mead (180 m msl).
       This  terrain  controls the mesoscale,  but not the synoptic scale flow

       Transport Regimes

       Several modeling and measurement studies have been conducted in the
vicinity of the MPP  over the past 20 years (Freeman and Egami, 1988; Yamada,
1988; Koracin et al, 1989; White et al, 1989).  Results from  these studies
provide a conceptual  model of pathways by which MPP emissions can reach
GCNP.   Figure 2 illustrates  the three synoptic flow  patterns of greatest
importance;  (1)  summertime dry southwesterly flow (flow from the southwest
toward the northeast), (2) summertime monsoons, and (3) winter storms.
       Both mesoscale and synoptic scale meteorological conditions influence the
movement of the MPP plume. The relative influence from each of these transport
and  transformation  scales differs from summer (June, July,  August) to winter
(December,  January, February). Southerly to southwesterly flows are needed to
transport MPP emissions to the GCNP.   Spring and fall are transitional periods
that  contain mixtures  of the summer and winter regimes  and  are  not as  well-
differentiated from each other.  Figure 3 illustrates the dominant air flow for each
quarter of 1990 as derived from the Dri Mountain wind data.
       During the summer, southwesterly,  westerly  and southerly winds are
common in the vicinity of the MPP.  There are two distinct cases; one with dry
airmasses and a second with moist monsoon airmasses, respectively.  During the
winter, the  most common situation is northerly winds  associated with a high
pressure  ridge over the Pacific Coast.   However, infrequent  frontal passages
result in  westerly and southwesterly flows on the order of 10% of winter days.
The  latter conditions can transport MPP emissions toward  GCNP.

       Dry,  Southwesterly Flow from  Southern California  and the Pacific

       The most common occurrence is dry, southwesterly synoptic flow caused
by heating of the Mojave Desert which creates a lower pressure with respect to
incoming air  masses.   These  air masses  traverse  the Mojave Desert after
entraining pollutants emitted from urban-southern California.  These include
pollutants flowing through Tehachapi, Cajon and Banning passes.
       This  scenario  has a  high frequency of occurrence during the summer
months.  The regimes change daily from decoupled flow during the night with
localized circulation patterns within the Mohave Valley and along the slopes of

                             Summer Monsoonal Flow
                high drifts west
                                 Winter Storms
Figure 2.     Synoptic flow patterns of concern, a) Dry summer southwesterly flow.
              b) Summer monsoonal flow,  c) Winter storms.


!/:590    3/3:  .990
    J   =jCO
                                                                     3==ED CUSSES I-DS,
                                                                   23   1 9<  -S
                                                                   1  —   3 3<  -5
                                                                     =   5 j<  »s <  95

                                                                   5 •  I I 10 3<  -s
   OH I  -OUf^T4 IN
7' 1/199O   9/30/199O
      0  2300
                                                   33 ; MQu^'i I '<
                                                    : 350   :5 • 3 :
                                                      0   5300
                10X     20X
          Figure 3:     Dri Mountain quarterly wind roses  from 1990.

the constraining mountains, to coupled flow that is dominated by the synoptic
winds aloft.

       Summer Monsoons

       During July and August, moist air is frequently transported from the Gulf
of California and/or the Gulf of Mexico in  southeasterly to southerly  flows.
Synoptic wind speeds vary from 6 to 20 m/s at 6000 m agl.  These air masses
traverse northern Mexico, the southern part of Texas, New Mexico, and most of
the state of Arizona. Pollutants emitted from the smelters in Arizona and Mexico
as well as those from Phoenix and Tucson can be entrained into this airmass.
This synoptic flow is driven by a large-scale low over the western part of the
U.S. created by strong surface heating.
       Differential heating causes updraft motions on the slopes of mountain on
both sides  of the valley,  resulting in chains  of  clouds developing  along  the
ridgetops.  These clouds may offer a mechanism for rapid oxidation of SO2 to
SO4 if the plume is entrained in them and if oxidants such as H2O2 are present' in
sufficient amounts.
       The reacted and unreacted emissions could then be carried through  the
Mohave Valley by the southerly component of the wind toward Lake Mead, after
which they might be transported toward GCNP by locally channeled flow or
caught up in the monsoonal flow and transported across the plateau to the GCNP.
Summer monsoon episodes are usually of 3 to  5 days in duration.

       Winter Storms

       In general, the synoptic weather patterns are not as favorable for transport
from MPP towards GCNP in winter. The Great Basin and the Colorado Plateau
are frequently dominated  by high  pressure cells creating  a flow that  is  not
conducive for transport from MPP to GCNP.   Southwesterly to westerly flow
occurs mainly during the movement of frontal systems, developing over  the
Pacific Ocean from west to east.  These  storms  in general exhibit a minimal
warm  frontal activity.  As a consequence the southwesterly to  westerly flow
needed for transport from MPP to GCNP will occur as the cold  front with its
associated trough approaches the Mohave Valley.   This weather type can last
from a day to three days,  be wet or dry, and  usually there are about ten cases
during the winter period.


       The Winter Haze Intensive Tracer Experiment (WHITEX), conducted by
the National Park Service, was designed to evaluate the feasibility of attributing
single point source  emissions to visibility impairment in selected geographical
regions.  WHITEX  was  conducted  during a  six week period in January and


February 1987. During this time, an artificial tracer, deuterated methane (CD4),
was released from the NGS. Aerosol, optical, tracer and other properties were
measured  at Hopi  Point, which is in  GCNP, and  other locations.  Synoptic
weather maps indicated a high frequency of high pressure over the area, which
resulted in transport of  the NGS  plume from  the  northeast toward GCNP.
Trajectory analysis and deterministic modeling indicated transport from the area
of NGS to.Hopi Point during the period with  highest sulfate concentrations.
       The extinction budget at Hopi Point indicated that sulfate aerosol (and
associated water) contributed two-thirds of the non-Rayleigh light extinction
during WHTTEX. Attribution analysis used the Tracer Mass Balance Regression
(TMBR) receptor model and the Differential Mass Balance (DMB) hybrid model.
According to the NFS analyses, NGS contributed substantially to sulfate and light
extinction at Hopi Point.
       The WHITEX data analysis  methodology, results, and use of the results
were cause for considerable controversy.  The Committee on Haze  in National
Parks  and Wilderness Areas evaluated WHTTEX (National Research Council,
1990). The Committee neither fully supported or fully discredited the WHITEX
report. Based on evaluations of meteorological, photographic, chemical and other
physical evidence,  the Committee concluded  "at some times  during the study
period, NGS  contributed significantly to haze in  GCNP."   However,  the
committee also concluded that "WHITEX did not quantitatively determine the
fraction of S04= aerosol and resultant haze in GCNP  that is attributable to NGS
       A key uncertainty identified by the Committee is the use of  TMBR and
DMB to apportion secondary species such as SO4=. Limitations of the regression
analyses noted by the committee are: "(1) satisfactory tracers were not available
for all major sources; (2) the interpretation did not adequately account for the
possible covariance between NGS contributions and  those from other coal-fired
power plants in the region; and (3)  both models employ inadequate treatment of
sulfur  conversion,  which is an important controlling factor in the formation of
haze at GCNP."  Another limitation noted by the Committee was  the lack of
measurements  within the canyon (beneath the  rim).  A more complete review of
the National Research Council WHITEX evaluation is provided in Appendix 8.

       SRP Study

       The Navajo Generating Station Visibility Study was conducted for the
SRP,  the  operators of NGS,  from January 10 through March 31,  1990.   Its
purpose was to address visibility impairment in GCNP during the winter months
and the level of improvement that might be achieved if SO2 emissions from NGS
were reduced.  The  study was  performed to provide input to the  rulemaking
process of the  EPA regarding NGS SO2 controls  (Richards et al, 1990).
       Perfluorocarbon tracers were released from each of the three stacks of
NGS.  Surface and upper air meteorology, particle and gaseous components, and

tracer measurements were made at many sites.  Deterministic modeling was done
to estimate the contribution of NGS and other sources to sulfate levels for two 6
day periods with poor visibility.  Various data analysis techniques were used to
examine the relationships among NGS emissions, meteorology, air quality, and
visibility during both episode and non-episode conditions.
      The SRP study  concluded that NGS  emittants were absent from  the
vicinity of Hopi Point most of the time .  The study estimated that the average
contribution of NGS to-fine sulfur at Hopi Point was small, although NGS sulfur
dominated during one 4-hour period. However, it was noted that the frequency
of wind directions transporting the plume toward GCNP were lower than normal
during this  time period.

      MPP Emission Modulation Studies

      The MPP  was inoperable  for the seven month period June through
December 1985. Using data from the period of shutdown and during operation
of MPP, a  study was done by SCE (Murray et al, 1990) to assess the effect of
MPP upon particulate sulfur levels at Spirit  Mountain,  Meadview, and  Hopi
Point. Spirit Mtn. is 20 km northwest of MPP, Meadview  110  km north-
northeast of MPP and Hopi Point 240 km northeast of MPP.  Meadview,  5  km
west of the boundary of GCNP, is expected to have the highest particulate sulfur
impact from  MPP among the three  sites.  The study  found  no statistically
significant  difference in sulfate levels at the three sites between operation' and
shutdown of MPP. It was suggested that the substantial year to year variability
of sulfate was responsible for not detecting a statistically significant difference.
The 95% confidence bounds  for the MPP impact was from less than 11.6% to
less than 21% at Meadview and less than 3.3% to less than 7.8%  at Hopi Point
during favorable transport conditions.  The upper  limit on  average sulfate at
Meadview  was  estimated to  be  15%, which is the level of uncertainty in  the
statistical analysis.
       From  data presented by Murray, it can be seen that sulfate levels at  Spirit
Mountain,  generally  not affected by MPP,  were greater during the outage
compared to non-outage periods, indicating higher background levels during  the
outage.  However, at Meadview, average sulfate levels were lower during  the
outage.  Thus,  levels at Meadview were lower during the outage even though-
regional levels were higher.  While suggestive, the number of samples was  not
sufficient to prove an impact from MPP.  This comparison, done  as part of the
scoping process, appears in Appendix 4.
       A more sophisticated study  of the outage will be conducted under  the
Desert and Intermountain Air Transport program at DRI, sponsored  by  SCE.
Chemical  and physical analysis  of filters for the SCENES program, used in
Murray's  study,  were  analyzed  only  every third  day.   Samples  for  the
intermediate  days were  archived.  The  new study will  analyze all  samples,
including  those previously analyzed.   A  more sophisticated  meteorological


classification  scheme  will also be  done.   The sulfate levels for the  same
meteorological regimes can then be compared for the outage and non-outage
conditions. Other emission modulations of shorter duration (i.e. periods where
only one  of  the two units  at MPP  was operating)  will also be analyzed.
Deterministic wind field and transport modeling will be done for each of the
meteorological regimes.  The modeling will account  for variations within each
regime.  A detailed compilation of regional SO2 emissions for the control and
outage periods will be done.  A draft version of the outage study plan appears in
Appendix 5.

3.   General Field Study Design

      The duration of the field study  will be one year.  It was considered
important in evaluating the overall impact of a source to consider a complete
annual cycle. By monitoring for an entire year, all the seasons may be studied.
For practical reasons, the year was divided into "intensive" and "non-intensive"
periods. During the intensive periods tracer will be emitted from the MPP stack
and tracer and particulate data will be collected continuously at over 30 sites.
During the non-intensive periods  tracer will not be released, the number of
particulate monitoring sites will be scaled back considerably and sampling will be
done only two days per week.  Meteorological and optical monitoring will be
done continuously.

       Selection of the Intensive Periods

       In selecting the intensive study periods, it was desired to select periods in
which the MPP may be most likely to contribute to haze in GCNP. It is expected
that  secondary  sulfates formed from  oxidation of MPP SO2 emissions  is the
largest portion of the MPP contribution  to haze in GCNP. Primary particulate
emissions from MPP contribute to haze nearer to the power plant, but  at the
distance of the GCNP, secondary sulfates are expected to dominate. Dry phase
oxidation of SO2 is much slower than aqueous phase oxidation.  Thus, cloudy
periods can cause much more rapid conversion of SO2 to sulfate. Aqueous phase
oxidation is on the order of 50-100% per  hour if oxidants are present in sufficient
quantity (Lee,  1986).
       Cloudy periods with wind directions transporting the MPP plume toward
GCNP are the periods when impacts to visibility at GCNP due to MPP would be
most likely  to occur.   As discussed in  Section 2, these conditions may occur
during the summer monsoon and  certain winter periods.  Calculations of the
potential impact of MPP to haze at GCNP  under highly simplified conditions
were done for  dry southwesterly and monsoonal summer conditions, and pre-
frontal winter conditions.  These calculations indicated a potential for perceptible
visibility impairment at GCNP from  MPP  emissions for all three cases (see
Appendix 4).
       Moisture parameters calculated from long term National Weather Service
data from Las Vegas are shown in Figure 4.  Specific humidity, which gives the
amount of water vapor in the air, is highest in August, with July having slightly
less moisture.  Average monthly dew point temperature for the years 1982-1990
(Figure 5) at Dri Mountain also showed a peak in August, with slightly lower
values in July.  Relative humidity peaks in December and January.  Also note that
August has higher relative humidity than July.  December and January are the
cloudiest months,  with February and March only slightly  less cloudy.   A
secondary, peak in cloudiness occurs in July, with somewhat less cloudiness in

      Relative Humidity - Las Vegas
Specific Humidity - Las Vegas



j 9
i i 5
1 M




~Ss 6-
f 5-
Specific Humid
D -* K) CO .&•

. s 5 I
1 1
! .1. .1 i .1
! ^
! !
j i
i 5
i ^
         Jan Feb Mar Apr May Jun Jul AugSep Oct NovDec
  Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Cloud Cover Climatology  Las Vegas       Precipitation Climatology  Las Vegas
3U . 	
45" «
40 -j



; t
: 1


i i



i 1






i *


^ ^
; i

i t


! !
; t








r fc
^ t
^ J
« t
5 *
^ k



       Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
                                                                              1921 -80

    Figure 4.    Las Vegas moisture climatology

                (a)   Average specific humidity by month: 1951-1980.

                (b)   Average relative humidity by month: 1951-1980.

                (c)   Average percent daytime cloud cover by month.

                (d)   Average precipitation by month.



          -10 -
              L]    2    3    4   5    6    7    8    9   10   iT   12

             Mean  Monthly DewPoint (C) -  1982  to 1990
Figure 5.  Mean monthly dew point temperature at Dri Mountain: 1982-1990.
360 -

315 -

270 -

225 -

180 -

135 -

 90 -
            0 -
                                     6   7
                                            —i	1—
                                             10   11
         Mean  Monthly Wind Direction  (deg)  - 1976 to  1990
Figure 6.  Mean (resultant) wind direction by month at Dri Mountain: 1976-1990.


August. The precipitation data show two distinct peaks; one is July and August,
the other December through February.  Of interest is the substantial difference
in average precipitation in November and December between the 1921-1950 and
1951-1980 data. This suggests that year to year variability is large.
       The climatology described above suggests a summer intensive period
covering portions of July and August and a winter period that could be any time
between December and  February.  January  and December showed the highest
values for the moisture related parameters, with January's precipitation data being
more consistent than December's.  January was chosen for the winter intensive.
August has slightly higher relative and specific humidity than July.  Thus,  the
summer intensive will be centered on early August.
       Even though we  are attempting to optimize  the study periods  for  the
specific conditions described above, meteorological conditions are highly variable
from year to year.  Most frequent winter flow  at MPP is away from GCNP.
However, winter flow is often toward the Joshua Tree Wilderness, another Class
I visibility protected area. In summer, we are likely to experience dry flow from
the southwest a significant portion of the time in addition to the moist monsoonal
flow.  Thus, information about other common conditions will also be obtained.
       Mean vector (resultant) wind direction each month for the years 1976-1990
at Dri Mountain is shown in Figure 6. Dri Mountain is a pointed hill 150 meters
high and adjacent to the Colorado River a few kilometers north of MPP.  The
instrument level is approximately at  the same elevation (MSL) as the  top of the
MPP stack.  However, the plume centerline is generally 400-700 m, averaging
663  m above stack base, which is 250-550 m above Dri Mountain. The winds
at Dri Mountain would be expected to be influenced more by channeling due to
topographic features than the winds at plume  height, particularly during nighttime
and  morning hours.  Winds at plume height typically have a greater westerly
component (toward GCNP) than Dri  Mountain  winds.   The  winds at Dri
Mountain indicate a predominance of northerly winds during November through
February and  southerly  winds  during April through September.  March and
October are transitional periods. Three January periods during 1976-1990 had
south to southwest resultant winds, indicating more frequent flow toward GCNP
during these years.
       The analysis of humidity, clouds, precipitation, and winds suggest optimal
of January 4-31, 1992 for the winter intensive and July 15 to August 25, 1992 for
the summer intensive.

       Siting of Monitoring Instrumentation

       The aerosol,  tracer and optical monitoring network includes three classes
of sites. These are denoted as (1) receptor, (2) other Class I, and (3) background
sites.  A more detailed description of air quality and meteorological monitoring
is described in sections 5 and 6, respectively. The aerosol and tracer monitoring
was  designed to provide sampling and analysis every day for many sites during

the intensive periods, and sampling and analysis two days a week during the rest
of the study year. The reduction of monitoring for the non-intensive periods is
necessary due to cost considerations.
       The preliminary network of sites is shown in Figure 7. The siting will be
finalized after a monitoring planning meeting to be held in Las Vegas in early
October. A listing  of the sites, approximate elevation, instrumentation, and a
brief reason for selecting each site is given in the table below. The receptor sites
(R1-R4) are either within or in very close proximity to GCNP.  The other Class
I sites (11-16) are in areas  that may be impacted by MPP and/or serve as
background sites. Most of the receptor and other Class I sites had some degree
of existing  or planned monitoring prior to Project MOHAVE.  These sites will
have supplemental monitoring associated with Project MOHAVE and will operate
during the  entire study year.  The background sites  (B1-B21) are intended to
characterize high elevation and low elevation transport into the study area as well
as showing more detailed concentration patterns  within the study area.   The
background sites  will  operate  only during  the intensive  periods.    The
instrumentation to be used is described in Sections 4-7 and references cited in
those sections.
                        SITE IDENTIFICATION TABLE
Id No      Name       Elevation

Rl    Meadview          900
R2    Long Mesa         1830
R3    Hopi Point         2160
R4    Indian Gardens      1220

II     San Gorgonio       1680
12     Joshua Tree         1500
13     Tonto              730
14     Sycamore Canyon   2000
15     Petrified Forest      1680
16     Bryce Canyon.      2600

Bl    Tehachapi Pass      1240
B2    Cajon Pass         1380
B3    Baker              280
B4    Amboy             190
B5    Parker              130
Particle  Optical
& Tracer





B6    Wickenburg         620         3
B7    Las Vegas Wash     370         3
B8    Cottonwood Cove    210         3                  S,U
B9    Yucca              580         3
BIO   Dolan Springs       850         3
fill   Truxton             1370        3                  S,U
B12   Seligman           1620        1
B13   Prescott  (airport)     1620        3
B14   Overton  Beach       370         3
B15   New Harmony       1520        3
B16   Marble Canyon      1220        3
B17   Mt Springs Summit  1680        3
B18   Spirit Mountain      1700        3
B19   Hualapai Mt Park    1980        3
B20   Camp Wood        1980        3
B21   Jacob Lake          2400        3

Explanatory notes:

1 Full IMPROVE samplers. 24-hour samples midnight to midnight, Wednesday
and Saturday during the non-intensive periods.  Twice daily samples of aerosol
and tracer will be taken each day during the intensives.  Specific hours for the
beginning and end of each daily sampling period will be determined at the
monitoring coordination meeting. DRUM samplers with 4 or 6 hour sampling
periods (only selected samples will be analyzed).
2  Full IMPROVE samplers.   24-hour samples (aerosol and tracer) each day
during the intensives.  24-hour samples midnight to midnight Wednesday and
Saturday  during non-intensive periods.

3  IMPROVE channel A and filter pack for SO2. 24-hour samples (aerosol and
tracer) each day during the intensives. No sampling during non-intensives.

4  Long Mesa will only have a DRUM sampler for particle monitoring.

T=  transmissometer,   N=  nephelometer,   P=  photography,  S=  surface
meteorology, U=  upper air meteorology

Surface and  upper air  meteorological data will be collected at additional sites
identified in Section 6.
       Important considerations in selecting sites include the availability of power
and accessibility.  The power requirement imposes strict limitations on siting.

           Points of Reference
           Coal Fired Power Plants

           Receptor Sites
           Other Class I sites
             Background Sites
           LA Basin Pass Sites
           Low Elevation Transport
           High Elevation Transport
                                                 Reid Gardner
                                                                                           Bryce Canyon
                                                                            New Harmony
Tehachapi Summit
    S Angeles
                                          Las Vegas Wa
                                       Sprs. Summit
                          Cottonwood Cove
                             Spirit Mountain
                                                              rton Beach
                                                                                      Marble Canyon
                                                             T        » Truxton
                                                          Oolan Springs
                                             . ,                        T
                                            Mohave  A Hulapai Mt. Park
                                                                                   ndian Gardens
                                                                            Hopi Point
                                                                                                 Petrified Forest
                                                                                      Sycamore Canyon
Cajon Summit

      • San Gorgonio
                                                                                  Camp Wood
                                                                                      Prescott (airport)
                               Joshua Tree
                                                                                                 • Tonto
                                                                                           O Phoenix

Meadview was chosen because it is within about 5 km of GCNP, has existing
monitoring by the Desert Research Institute (DRI), and  is at the west end of
GCNP, thus closer to MPP than other areas of GCNP. Long Mesa is also at the
edge of GCNP and the location of another DRI monitoring site.  Hopi Point and
Indian Gardens are existing NPS monitoring sites within GCNP. Joshua Tree and
Sycamore Canyon are potentially impacted by MPP. The remaining Class I sites
will help characterize  transport into the area.
       The Tehachapi Pass site (Bl) is located in a pass between the San Joaquin
Valley and Mojave Desert and is intended to monitor the exchange of air between
these areas.  The San Joaquin Valley is a large source of  SO2.  The Cajon  Pass
site (B2) is located between the Los Angeles Basin and Mojave Desert and is a
major exit pathway for Los Angeles Basin air.  Sites B3-B6, 12, and 13 are low
elevation southern boundary sites.  These locations form  an arc to characterize
the sulfur flow into the main study  area from the  southwest to southeast.
       Locations B17-B20 form a second arc to the south of GCNP. These sites
are located on terrain  that rises 900 to 1200 meters above the  surrounding area.
The sites should be in the middle of the mixed layer during the summer intensive
and frequently above the mixing layer during the winter intensive.  Measurements
from these sites when tracer is absent, coupled with the nearby low-elevation
southern sites B7-13, should characterize the sulfur flux  from the  southwest
through southeast exclusive of MPP sulfur into the receptor area. At other times,
tracer from MPP may be present  at these sites.   In conjunction with the low
elevation southern sites, these sites will help determine vertical distributions of
sulfur and tracer.
       Sites B7-B13 are located in possible MPP transport corridors between the
southern boundary sites and GCNP. These locations will indicate if the emissions
from MPP are  transported toward GCNP  in a  narrow  cone or more widely
dispersed air mass, in  addition to identifying the most common transport corridor
from MPP to  GCNP.
       The MPP plume  usually travels to the south along  the Colorado River in
the wintjsr.  It is suspected that the plume may sometimes  leave the river area in
an  eastward direction through a gap in the mountains near site  B9.  The high
elevation sites B18-B20 along with B10-B12 should be able  to  verify if MPP
emissions are  being transported from the area of B9 to the east or northeast in a
low-level surface layer or more dispersed in a deeper layer.
       Locations B7, B8, BIO,  and B14  are placed in  an attempt to  isolate
emissions of MPP, the  Reid  Gardner power plant, and Las Vegas as they are
mixed over Lake Mead  on their way to the western end of GCNP (Meadview).
Under stagnation  conditions,  the  high  elevation sites  B17 and B18 should
characterize the cleaner  regional air above the mixing layer.
       The  northern  boundary sites B15,  B16,  B21, and 16  are located  to
characterize flow into the region from the north.  These  sites will help identify
the effects of the Wasatch Front urban and industrial sources, the NGS, and other
coal-fired powerplants to the north and east.  Site  B15  is very close to  Zion

National Park.  Flow from MPP is likely to be transported toward this site often
during the summer. Locations B15 and B16 serve as low elevation sites. 16 and
B21 are high elevation sites.

4.   Tracer


       During the intensive study periods, an artificial tracer will be released
from the stack of the MPP and monitored at the same 31  locations as the air
quality  monitoring.   There are several  reasons  for  releasing  tracer. Tracer
monitoring data will identify the general transport patterns for the MPP plume.
Knowing where the plume goes is critical to begin  to understand the larger
question of the MPP's visibility effects. To fully resolve the plume position and
extent,  a very extensive monitoring network,  including aircraft measurements
would  be required.  This  is  beyond the resources  of the study.   The 31
monitoring locations  should provide  the  approximate location  of the plume,
although its horizontal and vertical extent will be uncertain.
       Different artificial tracers will be released from the Los Angeles Basin and
San Joaquin Valley during one-half or more of the summer intensive.  These
additional tracers  will be released to gain insight into the transport of emissions
from these large source areas into the  Project MOHAVE study area. They will
help identify the interaction between MPP and southern California emissions and
provide dilution ratios for southern California emittants.
       The tracer will be  used to provide a check of deterministic modeling
results.   A  transport  model,  using  wind  fields generated by a  dynamic
meteorological model, will predict  plume locations.  The  tracer data will be
compared to the transport model predictions to evaluate the model performance.
The concentrations of the tracer will be used to evaluate dispersion of the plume
predicted by the models as  well as location.  The dynamic meteorological and
transport models are discussed in section  10.
       The tracer  will also be used for receptor and hybrid  modeling purposes.
The tracer will serve as a unique "tag" for the MPP.  The  receptor and hybrid
modeling is described in section 13.

       Choice of Tracer

       Ideally, a tracer should closely mimic the species of  interest for receptor
modeling and  chemical transformations; in this instance SO2 and its conversion
to SO4, and deposition of the sulfate particles.  This would suggest using isotopes
of sulfur or oxygen. However, the large amounts of these materials that would
be required are not available;  to produce them would  require resources greater
than those available for this study.
       Among the potential tracer materials are deuterated methane (CD4),
various perfluorocarbons (PFT's), Sulfur hexafluoride (SF6), and particulate rare
earth oxides.   CD4 and PFT's  and SF6 are conservative tracers; thus conversion
of SO2 to SO4 and deposition of SO2 and SO4 can not  be  directly simulated.  It
has been  suggested that nonconservative  rare earth  particle tracers be used


because of their potential to mimic sulfate particles.  However, sulfate particles
are not directly emitted in significant quantities; rather they are typically formed
during transport  at rates which vary with meteorologic and other atmospheric
conditions. Thus some variable proportion of the rare earth particles will have
deposited  before the sulfates are formed.  Additionally the deposition of SO2
occurs more rapidly than either sulfate or rare earth particles. A combination of
conservative and paniculate tracers could yield additional insights into the fate of
MPP emissions than that obtained using a single tracer.  However, the additional
expenses  associated  with  using  an  additional class of tracer  is beyond  the
resources  of Project MOHAVE.
       SF6 has been used in many short range experiments.  Although the cost
per  kilogram  is low compared to other conservative  gaseous tracers,  the
background concentration is much higher, which more than offsets the decreased
unit cost.  SF6 is not practical for the spatial scale of the study region.
 !      CD4, used in WHITEX has low background values and  is detectable at
vejry low concentrations, so small amounts of this tracer are sufficient. Though
the cost per unit  mass is high, the total cost of tracer material is less than the cost
of PFT's. However, the sample analysis cost is very high ($800-$1000/sample),
compared to about $20/sample  for PFT's.   The lower cost of PFT analysis
encourages the analysis of many more samples, for the available budget. For CD4
the strategy is to analyze a subset of all possible samples.  Analysis of all samples
allows a more thorough evaluation of the deterministic modeling.  Different PFTs
can be released from other sources of interest and  analyzed from the same sample
for virtually the  same low analytical cost.
       The SRP tracer study, which used PFT's,  apparently had some major
problems  with the tracer portion of the study.  Collocated samplers showed near
zero correlation.  Apparently this was at least partially due to the fact that many
samples were  near the limit of detection.  Other tracer studies have also had
apparent quality control problems, for  example, contaminated  samples.   It is
imperative to have a rigorous quality control program for the tracer components
of the study. The quality control methods to be used' for the Project MOHAVE
tracer study is described later in this section.  There is no fundamental reason
that would prohibit PFT's or other tracers from giving reliable, quantitative
       Project MOHAVE will use perfluorocarbon tracers.  The tracer to be used
to track the MPP plume is ortho-cis perflorodimethylcyclohexane (ocPDCH).
The tracer material to be released is ortho (o) PDCH, 45 % of which is ocPDCH.
Perfluoromethylcyclopentane (PMCP) will be used to tag the Los Angeles Basin.
Perfluorotrimethylcyclohexane (PTCH) will be used  to track emissions from the
San Joaquin Valley. The ambient background of ortho-cis PDCH is very  low,
0.3  parts per quadrillion (ppq) (Dietz,  1987).  The SRP study used  PDCH and
other PFT's but  analyzed for  total PDCH, not  individual  isomers.   The
background of total PDCH is 22 ppq.  PMCP background concentrations are 3.3
ppq; PTCH background is 0.3 ppq. Brookhaven National Laboratory (BNL) will


do the tracer analysis for Project MOHAVE.  In addition to analyzing isomers,
BNL pre-concentrates the sample; thus much greater sensitivity is achieved
compared to the SRP analysis methodology (Dietz, 1991).
       Tracer Release

       Tracer can be released at a constant emission rate or at a constant ratio of
tracer to SO2. Variation of tracer to SO2 ratios was a complicating factor in the
WHTTEX receptor modeling analysis. If released at a constant rate, SO2 emission
rate variations would complicate the receptor modeling, requiring adjustment of
the ratio of tracer to sulfur dioxide concentration.  This requires knowledge of
plume age.  However, for use in deterministic modeling, it is more desirable to
have a constant tracer emission rate, to simplify the dispersion calculations. If
a constant release rate were used, the deterministic model would be used to give
the plume age1 necessary to adjust the tracer to sulfur dioxide emission rates in the
receptor modeling.  The MPP is a base loaded unit. It typically operates at either
full capacity, 1/2 capacity (one unit down) or is down.  Tracer will be released
at a rate proportional to the SO2 emissions if a practical approach to do it can be
devised.  If not,  then the tracer release rate will track the status of the power
generation units with full, one-half or zero tracer emissions,  corresponding to
two, one, or zero units operating. This will more closely preserve the ratio of
tracer  to SO2 emissions than a constant tracer release rate.  Good coordination
between MPP operators and the tracer release personnel will be expected. Tracer
release from the San Joaquin Valley and Los Angeles Basin will be at a constant

Release Equipment

       The perfluorocarbon tracer liquids are very similar in viscosity to silicone
fluids, but are quite dense (densities from 1.7 to 1.8 g/mL liq.).  Large release
rates,  tens of kilograms per hour, have been accomplished with (1)  atomizers
spraying directly  into the air, or (2) by vaporizing a PFT liquid stream, diluting
with air below the PFT dewpoint at the exit, and emitting the diluted stream into
the air or other fluid (such as the flue gases going up a power plant stack).
       For low release rates, tenths of kilograms per hour,  such as will be needed
for Project MOHAVE and as was used in METREX in 1984 (Draxler,  1985), the
tracer  can be released by evaporation using the METREX-designed equipment.
The release unit  has only two moving parts: a squirrel cage fan motor and  a
metering pump.  The tracer flows in a closed circuit from the reservoir through
the peristaltic pump rollers (the tubing is compressed to move the liquid) directly
into the airstream on the surface of a heated disk.  The disk  and heater are
located in a cylindrical mixing chamber.  The heater, adjustable up to 600 W,
maintains the temperature of the disk above the tracer's boiling  point.   The

system's  electronics control the duration of release and the duration that the
system is off. Times for each on-off cycle can be set by tens/whole/tenths of an
hour for  each cycle.  A small strip chart recorder notes when the pump and
heater are on. The pump rate is preset on a calibrated dial. The airflow should
be sufficient to ensure all the vapor is diluted below the saturation mixing ratio
for the expected ambient temperature without blowing the tracer drops off the
heater before they vaporize.
       Three release units were built by the NOAA Air Resources Laboratory in
Silver Spring, MD, and now reside at their laboratory in Idaho Falls. The system
was  designed to handle release  rates of the magnitude  needed for Project
MOHAVE.  However, substantial design changes may be made by NOAA in
consultation with Brookhaven in  order to insure reliable operation (including
accurate release rates and constancy of release).

PFT Programmable Samplers

       Each site will be equipped with a programmable Brookhaven atmospheric
tracer sampler (BATS).  The sampler was initially developed by BNL and was
commercially manufactured by the Gilian Instrument Corporation (West Caldwell,
New Jersey). The unit consists of two sections: the lid, containing the sample
tubes, and the base, containing the power control.  The entire unit is housed in
a weather-resistant 36 cm x 25 cm x 20 cm container and weighs approximately
7 kg. Power is supplied by an internal rechargeable nominal 8-VDC battery for
operation at remote locations, or by a charger where 155-VAC is available. For
Project MOHAVE, each unit must be run on a charger in order to collect the full
twenty-three (23) 36-or 72-L air samples.
       The BATS removable lid  holds 23 stainless steel sampling tubes, each
packed with approximately 150 mg. of Ambersorb adsorbent.  The Ambersorb
adsorbs the tracers from the sample air flowing through the tube. Breakthrough
of the perfluorocarbon tracer gases is less than 0.1 %.  The tracer gases remain
adsorbed until extreme heat is applied to the tube to desorb the tracer at analysis
time. The sample air flow is directed consecutively through the adsorbent tubes
by means of a  multiple port switching valve which is controlled by  the base.
Since the lid is removable and interchangeable, multiple lids can be used on a
single base.
       The BATS base contains a DOE-Environmental Measurement Laboratory
constant  mass flow pumping system  (Latner,  1986) which draws sampler air
through each tube. The flow rate is selected by  setting an internal switch to draw
either 10, 20, 30, 40, or 50 mL/min of air; the switch cpntrols the on-off cycling
rate of the pump over a 1-min period.  A constant flow rate through each sample
tube in the lid is regulated by a pressure sensing circuit located at the outlet side
of the pump.  The circuit is an integrator that supplies a  voltage ramp to the
pump motor, rising or falling as  indicated by  the outlet pressure.  A flashing
light-emitting-diode (LED), mounted on the  BATS base control panel, gives a


visual indication that the pumping system is operating properly.  This pumping
system has proved to be more reliable than the originally installed pump, but
consumes more  power.   Programmable controls are also placed  on the base
control panel which are  used  to  control  the number of samples, the  sample
duration, and to control either single or multiple sample start and stop times of
a 7-day period.  Two liquid crystal displays (LCDs), also mounted on the control
panel, show the clock time, day of the week, and current tube number.  A digital
printer and integrated circuit memory module (Lagomarsino,  1989) record the
start time, the day of the week, and the tube number for each sample. The BATS
base controls are also used to assist in automated analyses when the lid is coupled
to a gas chromatograph (GC).
      For analysis, the perfluorocarbon  tracers, retained on the Ambersorb
adsorbent in the BATS tubes from the sampled air,  are  desorbed by resistance
heating  of the  stainless  steel  tubes to 460°C.  Current from the BNL gas
chromatograph   system  (16.3  Amps AC) is  supplied from a  low  voltage
transformer (—1.55 VAC  at  the lid jacks) through the Canivalve solenoid
assembly.  The  assembly consists of a 24-position rotary  solenoid having two
power decks capable of handling 20 amps.  Twenty-three leads are wired to the
power deck, each connected to the adsorbent  tube floating clamp at one end of
the respective tubes.  The  clamp at  one  end must  float to allow for thermal
expansion of the tube on  heating (~ 0.8mm).  A set  screw  secures  the collar on
the tube within  the clamp; a similar set  screw on  a common  aluminum rail
secures the other end.  Polyurethane rubber tubing (1/8-inch OD by 1/16-inch ID)
is expanded over  the 1/8-inch OD  stainless steel  adsorbent tubes and wire
clamped  to secure;  the other  end  is  attached to  the Scanivalve 1/16-inch

Tracer sample analysis

       Tracer sample analysis will be done with a gas chromatography system.
The gas  chromatograph system is composed  of a gas chromatograph,  of data
handling  devices,   gas  standards, and  a BATS.    The Varian  6000  gas
chromatograph consists of a series of specially  designed traps, catalysts, columns,
and an ECD-electron capture detector. The data handling system consists of an
analogue electronic filter on the ECD  electrometer output connected to a Nelson
Analytical  300  Chromatograph  system comprised of a Model 7653 Intelligent
Interface  and  an  IBM  PC/AT  with an ink jet  printer and  Nelson  2600
Chromatography Software.  Brookhaven has also written extended software for
further data processing and  GC  calibrations.
       Analysis  of a sample occurs when  the sample is automatically thermally
desorbed from the BATS sample tube.  The sample is passed through a precut
column and a Pd catalyst bed before  being reconcentrated in an in-situ  Florisil
trap.  Once the trap is thermally desorbed, the sample again passes through the
same catalyst bed, another Pd catalyst bed, and then through a permeation dryer.

The sample is then passed into the main column where it is separated into the
various  perfluorocarbon  constituents and  then  ultimately into  the  ECD  for
detection.  Further details on the analytical system is given in Dietz (1987).
      Release rates  and  expected  crosswind   average  peak   centerline
concentrations at the Long Mesa and Hopi Point receptor locations are shown in
the table below.
Summary of Expected Tracer Concentration
Tracer Release
San Joaquin Valley
Los Angeles Basin
Rate, kg/h
Long Mesa
Hopi Point
Hopi Point
Expected PFT levels
Conc.d, fL7L
2 ± 1
14 + 5
       fL = femto Liter = 10'15L
       100 kg for 30 days in January 1992 and 170 kg for 50 days in July-August
       70 kg for 21 days in July 1992
       250 kg for 21  days in July 1992
       Crosswind average peak centerline concentration
       At receptor sites, 12-hour tracer samples will be collected, and will sample
36 liters (L) of air. All other sites will sample 72 L over a 24-hour period.  The
table on the next page shows relevant information regarding the amounts of tracer
expected, backgrounds, levels of detection, and signal to background ratios for
the GC analysis. It can be seen that the limit of detection (LOD) and uncertainty
are very small compared to background, except for PTCH, which has a limit of
detection of about 16% of background and uncertainty of 50% of background.
Thus, for the MPP and Los  Angeles Basin tracers  (ocPDCH and PMCP), even
an additional tracer concentration of a fraction of background can be reliably
quantified.   For all 3 tracers, the  uncertainty of the amount of tracer above
background (signal-background) is small for expected crosswind plume centerline
       A sample chromatogram for a 20 L of ambient.(background) air is shown
in Figure  8.  Background levels of  the tracers  used  (ocPDCH, PMCP, and
PTCH) can be clearly distinguished and quantified.  PMCH, which is not being
released, can be used as a reference.

Figure 8.     Chromatogram of 20 L sample of ambient air. (a) Elution time up to 6.5 minutes.
             (b) Elution time 6.5 to 13 minutes.

For 12 hour (36L) samples
Area of Response (counts/fL)
Receptor concentration (fL/L)
Quantity in 36L (sample and
background) fL
Quantity in 36 L background, fL
Limit of Detection, fL
Limit of Detection, counts
Uncertainty = 3 Limits of Detection,
Counts in 36L (sample and
Counts in 36L background
Signal to background
Signal - background, counts
= 0.3 fL'/L
55,800 ± 60
3,888 ± 60
14.35 ± 0.22
(± 1.5%)
51,912 ± 85
(± 0.16%)
= 3.3 fL/L
99,680 ± 60
19,040 ± 60
5.24 ±0.017
(± 0.3%)
80,640 ± 85
(± 0.11%)
« 0.1 fL/L
21,600 ± 500
1080 ± 500
20.00 ± 9.27
(± 46.4%
20,520 ± 707
(± 3.4%)
      fL = femto Liter = 10'1SL

Tracer Quality Control

      Rates of air flow through the sampler are checked before and after the
sampler is sent to the field monitoring site.  This is to determine the total quantity
of air sampled each sampling period.  Adjustments  are made to compensate for
altitude and temperature differences. Three additional PFTs that are not released
are used  as a cross-check of the sampling volume.  The concentration  of these
PFTs is essentially constant, so the quantity of air sampled can also be calculated
from the  amount of these tracers collected.
      The sample analysis is  done at 460°C.  This is 50°C above the
temperature needed to desorb all PFTs.  After analysis, the sample tubes are
"baked out" at 510° C to remove any remaining traces of PFTs. Before sending
the tubes  out to the monitoring sites, every fourth tube is analyzed.  At this time,
the tubes should have zero tracer.  They are analyzed down  to 30-50 counts,

which is about 1 % of background. If a tube has zero signal, then it has not been
sampled, because the ambient background has not been detected.  The samplers
will be programmed to collect 20 or 21 samples;  tubes 22 and 23 should be zero.

5.   Air Quality Monitoring


      Air quality monitoring for Project MOHAVE has many applications.  The
extinction budget analysis requires data for all the major particle components (e.g.
sulfate, organic and elemental carbon, crustal, and liquid water as estimated by
relative humidity) by particle size to be used in conjunction  with optical  data
(scattering and extinction coefficients).  The hybrid and receptor models need
particle  and gaseous sulfur  concentrations and particulate trace elements as
endemic  tracers (such as  arsenic  for smelters) in addition to  measurements of
artificial  tracer. Oxidants, especially hydrogen peroxide, should be monitored to
assess the potential oxidant limitations of SO2 to sulfate  conversion.  The air
quality monitoring network will document the regional distribution of particulate
and SO2  and establish boundary conditions for the study area; used along  with
wind field information, transport of pollutants into the area will be identified.
Eigenvector analysis of the pollutant fields will identify common patterns and may
identify specific sources with the patterns. These data will also provide for a
check of the deterministic modeling results.

      IMPROVE Samplers

      The IMPROVE sampler consists of four independent filter modules and
a common controller, as shown in Figure 9. Each module has its own inlet, PM-
2.5 or PM-10 sizing device, flow rate measurement system, flow controller, and
pump. In the three PM-2.5 modules, the airstream passes through a cyclone that
removes  particles larger than 2.5 pm in diameter.  The airstream then  passes
through a filter, which collects all the fine particles.  In the PM-10 module, the
inlet prevents particles larger than 10 urn from being sampled.
       Channel A collects  fine particles  (<2.5  fj.m)  on a Teflon filter and
provides  total fine mass, elemental analysis  (H and Na-Pb), and absorption.
Particle Induced X-Ray Emission  (PIXE) analysis gives the concentration of the
elements Na-Pb;  Hydrogen is obtained by Proton Elastic  Scattering  Analysis
(PESA).   Absorption is  determined by  the Laser  Integrating  Plate Method
      Channel B uses a fine nylon filter behind  a nitrate denuder for  ion
chromatography analysis  (Cl', NO2', NO3' and SO42-).   Channel C is used to
obtain organic and elemental carbon from a fine quartz filter.  A thermal/optical
carbon analyzer which makes use of the preferential oxidation of organic and
elemental carbon compounds at  different temperatures is used.  Channel D
measures PM-10 total mass on a Teflon filter and SO2 with an impregnated quartz
filter.    More  detailed  descriptions of  the  IMPROVE  samplers,  analysis
methodologies, and protocol appear in Pitchford and Joseph (1990), and Eldred


et al. (1988).  The location of sites and monitoring schedules for IMPROVE
samplers is shown in Section 3.
                                          PM10 teflon
                vacuum hoses
                in conduit
       Figure 9.  Schematic of IMPROVE sampler.
       DRUM Samplers

       DRUM (Davis Rotating-drum Universal-size-cut Monitoring) samplers will
 be used at six locations.  The DRUM particulate samplers partition the aerosol
 into eight size ranges.  This provides critical information to relate aerosol to
 extinction because  of  the  strong relationship  between particle size and light
 scattering.  PIXE and PESA analysis is done to determine the concentration of

each element (H and Na-Pb) by size range. The size distribution, hence the light
scattering efficiencies, for different particulate component species can be inferred,
if sufficient material is collected (e.g. see Cahill et al, 1987).  The DRUM
sampler is described by Raabe et al  (1988).
       Six DRUM samplers will be deployed.  The sampling time will be either
four or six hours. The receptor sites will have DRUM sampling for the entire
study;  the remaining samplers will be placed at other locations of interest yet to
be identified,  Among the,possible sites are Tehachapi Pass, Cajon Pass and Spirit
Mountain.  Analysis of all samples is beyond the resources of Project MOHAVE.
All samples will be archived; analysis of samples will be done for selected
periods of interest.

       High Volume Dichotomous Samplers

       High volume (300 L/minute) dichotomous  samplers  will be used to
improve the trace metal data base for receptor modeling with endemic tracers.
Aerosols in the  size ranges 0.05-2.5 jim and  2.5-20 /im will be collected on
Teflon filters.  Instrumental Neutron Activation Analysis (INAA) and X-Ray
Florescence (XRF) will be done on the samples.  Three samplers will be used.
One sampler will be equipped with a trap to collect semi-volatile organics.  The
locations have not been decided yet; one will characterize background and one
will be near the  mouth of the Grand  Canyon.
       Scanning Electron Microscopy  (SEM) will be used to characterize
individual  particle  morphology and  elemental  composition.    In addition,
Computer-Controlled SEM (CCSEM) analysis  can be used  to  increase  the
numbers of particles analyzed and eliminate possible human operator microscopy
bias.  CCSEM data can then be used in data  analysis approaches that require
quantitative composition as a  function of particle size and shape distributions.
Mie Theory calculations to determine extinction budgets can use this data. Unlike
other data sets, CCSEM data allow for direct analysis of the questions of aerosol
mixture (i.e., the extent to which component species are constant in all particles
in an individual sample).  Receptor models can be based upon an endemic tracer
approach using CCSEM data, or it can use the individual particle characterization
information to aide in resolving  issues raised by other attribution approaches.

       Hydrogen Peroxide Measurements

       Hydrogen peroxide  (K2OJ is likely to have a  significant role in  the
formation of sulfate  in the study region when clouds are present.  Aqueous phase
conversion of SO2 to SO42- is critically dependent upon hydrogen peroxide (HjO^
and ozone (O3) (Penkett et al., 1979; Calvert et al., 1985).  Hydrogen peroxide
is thought to be the leading oxidant of dissolved SO2 in the eastern United States,
where  the pH of atmospheric water is generally below 4.5 (Heikes et al., 1987).
In the  desert southwest, where the pH of atmospheric water is typically higher,


ozone may also be important in the aqueous phase oxidation of SO2.  Saxena and
Seigneur (1986)  also identify  O2 catalyzed by Fe3+ and Mn2+ as an important
aqueous phase oxidant of SQj.  Hydrogen peroxide reaction rates with dissolved
SO2 are typically 50-100% per hour (Lee et al, 1986); thus the presence  of
clouds with sufficient H2O2 present can result in rapid  sulfate formation.  The
amount of hydrogen peroxide  available  for  oxidizing SO2 may  be  limited,
especially during winter, when photochemical generation of H2O2 is low  (Calvert
et al.,  1985; Kleinman, 1986).
       The NAS review of WHTTEX noted that H2O2  measurements were not
made;  the NAS used values measured in Tennessee (about the same latitude as
GCNP) to estimate potential sulfate formation.  Members of the Committee  on
Haze in National Parks and Wilderness Areas suggested that Project MOHAVE
make some measurements of hydrogen peroxide.  If measurements of hydrogen
peroxide show sufficient amounts to convert all the SO2  to sulfate, we can likely
conclude the atmosphere is not oxidant limited.  However, showing that molar
quantities of  hydrogen peroxide less than sulfur dioxide does not necessarily
indicate oxidant limiting conditions.  Ozone effects may be significant if the pH
is adequately high. Heikes et al., found that SO2 concentrations were a factor of
3-5 greater than H2O2 concentrations in the surface layer, but above the surface
layer H2O2 concentrations were twice the SO2 concentrations.  Even with aircraft
vertical profile measurements,  Heikes et al. concluded that the hydrogen peroxide
measurements were ambiguous in determining if the atmosphere was oxidant
limited.   Their near  cloud  observations suggested  that  physical-dynamical
processes may be as  or more important than a simple molar comparison of SO2
to H2O2 at ground or cloud level.
       It is not possible for Project MOHAVE to fully characterize the temporal
and  spatial  distribution  of  atmospheric hydrogen   peroxide  necessary  to
conclusively determine oxidant limitations. However, limited measurements may
provide some insight into the potential for hydrogen peroxide oxidation of S02.
As  in the NAS report,  sulfate  concentrations  may  be compared  to H2O2
concentrations to see if sufficient H2O2 existed to account for the measured sulfate
values.  Project  MOHAVE will make a limited number of hydrogen peroxide
measurements.   The SRP NGS study made  hydrogen  peroxide measurements
during the winter of  1990.  These measurements may be used to estimate H2O2
for the winter intensive.

       Methylchlorofonn Measurements

       Methylchloroform has  been identified as a tracer of weekday emissions
from the Los Angeles Basin (White et al.,  1990). Miller et al. (1990) found that
methylchloroform levels at Spirit Mountain are correlated with particulate light
scattering, with the majority of hazy conditions having elevated methylchloroform
levels. Methylchloroform measurements, in conjunction  with meteorological data
and modeling, can aid in identifying periods when air previously in the Los


Angeles Basin is in the study area.  However, a limitation of methylchloroform
as a Los  Angeles Basin tracer is that the emissions are primarily  weekday
emissions, with weekend emissions being much lower. Thus weekend emissions
from the Los Angeles Basin might not be tracked using this tracer and the
absence of methylchloroform does not necessarily indicate an absence of air from
the Los Angeles Basin.
       Desert Research  Institute will measure methylchloroform at Spirit
Mountain, Meadview, and Long Mesa.  These data  will be investigated for use
in identifying the presence of air previously in the Los Angeles Basin.  During
the summer intensive the release of perfluorocarbon tracers from the Los Angeles
Basin  and San  Joaquin  Valley should provide a check on the  utility  of
methylchloroform as a Los Angeles Basin tracer.

6.   Meteorological Monitoring

      Meteorological monitoring  is  necessary  to  characterize  the  speed,
direction,  and depth of  transport  in the region and for model initiation and
validation.  The existing National Weather Service (NWS) surface and upper air
monitoring sites are insufficient to characterize the complex meteorological setting
of the study area. In addition, NWS upper air measurements (rawinsondes) are
taken only twice per day.  Thus,  they may not capture important small time scale
meteorological  changes  and  because  they   provide  nearly  instantaneous
measurements, they may not be representative of average conditions.
      The Wave Propagation Laboratory (WPL)  of the National Atmospheric
and Oceanic Administration (NOAA) will provide much of the meteorological
measurements data for Project MOHAVE. Air Resource Specialists (ARS), the
optical monitoring contractor, will provide surface meteorological data at the four
receptor sites. WPL has a unique capability of providing continuous wind and
temperature profiles in  the atmospheric boundary layer (ABL)  using wind
profiling radars with Radio Acoustic Sounding Systems (RASS).  The radars
transmit 915 MHz signals and receive back-scattered signals from the atmosphere.
With three antennas, usually two tilted and one vertical, the three  components of
the wind can be measured using the Doppler effect. The best results are obtained
when the winds are averaged over  about one hour. The RASS component uses
the Bragg scatter of radar waves from vertical propagating acoustic waves to
measure the sound speed. Because the sound speed depends upon air temperature,
temperature profiles can be derived.   Usually the instrument is configured to
provide one 5-minute averaged temperature profile each hour.  The backscattered
intensities received by the wind profiler in the form of signal-to-noise ratios can
also  qualitatively  indicate mixing  depths.    The  advantage  of the  wind
profiler/RASS instruments over rawinsondes is that they provide continuous
profiles in time.


  The wind profiler/RASS data consist of wind profiles, nominally to 2.5 km, and
temperature profiles to almost 600  m.  These data are necessary  to characterize
the speed,  direction, and  depth of material transport in the region  and also
necessary  for model initiation  and validation.   The primary objective is to
measure the transport of material from the MPP  to GCNP. Also, it is important
to characterize the flow from major urban areas  in the region (e.g., South Coast
Basin,  Las Vegas, Phoenix/Tucson)  and  to  separate  this flow  from  flow
containing the MPP emissions. There are two other major power plants nearby,
the Reid Gardner Plant near Overton,  Nevada, to the northwest of the Grand
Canyon and the NGS near Page,  Arizona to the northeast. It is also desirable to


determine the frequency of transport from these sources.
       There are several ancillary problems which relate to the potential transport
paths which the MPP plume may take into the Grand Canyon region.  An indirect
path is along the Colorado River to the north and then over Lake Mead. Because
the lake is lower in eieiation than the surrounding terrain and the ABL over the
lake is usually more  stable than that over the surrounding'land  due to the
relatively cool water, the pollution may pool and collect in the Lake Mead Basin.
Also, material from other sources (e.g. Las Vegas/Henderson or Reid Gardner)
may collect in the same basin. A change in wind may transport this material into
the lower portion of the Grand Canyon near Meadview. It is therefore important
to monitor the winds and stability in the Lake Mead area,  both over the lake and
near the western entrance to the Grand Canyon.
       With southwesterly flow near the surface, the material from the MPP may
be  transported more directly toward the  Grand Canyon region over the high
plateaus to the northeast  of the plant.  This path  also requires meteorological
       Surface meteorology will be monitored at all the wind profiler sites so that
the lower gates of the profilers can be compared with surface parameters.  Also,
NOAA/WPL will provide at least four surface  pressure sites.  Gaynor  et al.
(1991) have shown that winds calculated from surface pressure gradients can  be
used as surrogates for transport winds. The pressure array will allow calculations
of mesoscale transport winds which  can  be compared with and be adjunct to
profiler winds.
       Another contribution from NOAA/WPL will be the wind and temperature
data from profiler/RASS  operations performed as  part of the South Coast Air
Basin study beginning in July, 1992, and continuing through the summer intensive
period.  Among the tentative locations  for these instruments are  the  Cajon,
Banning and Tehachapi passes.   These data will be useful adjuncts to Project
MOHAVE by providing upstream information on potential transport from the Los
Angeles Basin and San  Joaquin Valley  into the MPP region.    Starting  in
February, 1992  wind profiler data from sites on the Mogollon  Rim (central
Arizona) will be available.   These wind profiler sites  will help characterize
periods of flow from the southeast into the study area.
       NOAA/WPL will also provide tethersonde and airsonde profiles for short
periods and at critical locations during the winter and summer intensives.  These
profiles will be measured at transport, drainage, or pooling locations that will not
have regular continuous measurements.  Because of the limited height range of
the tethersonde, the preferred locations for these profiles will be in regions with
shallow boundary layers.  One general area of this type is the Lake  Mead basin
where a relatively shallow ABL  compared to the surrounding desert  may persist
well into the morning due to the  cool water surface and to the nocturnal drainage
of cool air into the basin.

       Field Study Plan

       A wind profiler with RASS will operate in close proximity to the MPP
from  September  1991  through  September  1992.    This  location  will  be
supplemented with the DRI operation of an AeroVironment Doppler sodar for a
quality control (QC)  check of the profiler and to supplement the  profiler with
detailed low level winds.   Another wind profiler with RASS will  be located at
Truxton to monitor the possibility of direct southwest to northeast transport from
the plant.  The Truxton site is in open terrain; this allows the data from this site
to reflect  the general flow patterns over  the entire study  area.  It will also
measure the south and southeast summer monsoonal flow from which directions
material may be transported from Phoenix and Tucson or from smelters to  the
south and  southeast.  A doppler sodar will also operate most of the study period
at Mead view.
       In support of the winter intensive study, two additional wind  profilers will
be operated from mid-November 1991 through late-January  1992. The  site
locations will be the following:

1)     South of MPP in the vicinity of Needles, which is usually  downwind of
       MPP during the winter.

2)     At Temple Bar, on the south shore of Lake Mead, about 30 km west of
       GCNP.  This  site will help characterize low level flow over Lake Mead,
       which may vary significantly from the flow at higher levels.

       During the winter  intensive  period, NOAA/WPL  will   intermittently
operate a  tethersonde and/or radiosonde to supplement  the upper air data. The
locations may be at Cottonwood Cove to monitor wind and stability in the upper
Mohave Valley, or near Lake Mead to monitor the meteorology in the Lake Mead
       From July 1992 through  September 1992, a  supplemental profiler will
operate at Cottonwood Cove (Lake Mohave) in support of the summer intensive
experiment. The plume is typically transported past this site, especially during
night and  morning hours, and may exit the Colorado River valley  near this  site
during the late morning  and  afternoon.  An additional wind  profiler will be
located at Meadview. The sodar at Meadview will be moved to Temple Bar to
measure low level flow above Lake Mead.  The combination of doppler sodar at
lake level, combined with a wind profiler at Meadview, 500 meters  above lake
level, will provide a  vertical profile extending to about 3 km above lake level.
NOAA/WPL will likely participate in the South Coast Air Basin  Study which will
occur during the same period as the MOHAVE summer intensive.  WPL will
have six profilers operating in the South Coast Basin.  One or two of those will
be on the east  (desert)  side  of the  Tehachapi,  Cajon,  or Banning  Passes.
Combining data from the South  Coast  profilers with data  from  the  profilers

deployed for the MOHAVE summer intensive will provide a rare opportunity to
continuously monitor the winds from Southern California to the Grand Canyon.

      Data Collection

      All the profilers will provide hourly consensus averaged winds in two
modes ~ a high range resolution mode, usually about 60 to 100 m and a low
range resolution mode, usually 200 to 400 m.  Minimum heights of around 150
m and maximum ranges of about 2.5 km are expected.  During the more moist
summer monsoon period,  much higher ranges may be expected.
      The RASS temperature profiles are measured once per hour representing
5 minute consensus averaged profiles. The minimum range is about 150 m; the
maximum range expected under dry desert conditions is nominally 600 m.  The
Doppler sodar at Overton  or Temple Bar will provide a minimum range of about
50 m and a maximum range of about 600 m with about a 50 m range resolution.
      The surface meteorological data associated with the profilers will probably
represent 5 minute averages of wind, temperature, and relative humidity measured
about 3 m above the ground. The locations measuring surface pressure will also
have temperature  and relative humidity instrumentation.
      Where phone lines are available, all profilers,  including those with RASS,
and the sodar will be interrogated by phone once per day and the ASCII files sent
to a hub  work station located  at  NOAA/WPL in  Boulder,  Colorado.  This
validation level zero data will also be available at each site from printer paper and
on the hard  disks of each controlling PC.  The surface meteorological  data
collected at the profiler sites will also be sent over the same phone lines to the
hub.  The pressure sites, unless co-located with the profilers, may not have phone
line capabilities depending on the feasibility of installing lines.

      Data Quality Assurance

      Wind  profilers and Doppler sodars identical  to those to be deployed for
Project MOHAVE are periodically tested and compared at NOAA's Boulder
Atmospheric  Observatory which  includes a 300 m meteorological tower.  The
RASS derived temperatures are also compared to thermometers on the tower.  All
instrumentation will have been previously tested in  other field studies prior to
deployment.  The collocation of a Doppler sodar at the MPP with a wind profiler
will provide a continuous field quality assurance check on both the profiler and
the sodar.
      All the data that is recorded and printed out at each site and sent over
phone line to the hub in Boulder will be level zero.  The field programs on each
control computer  for the radar/RASS  and sodar provide consensus  averaging
which is equivalent to on-line, real-time sorting of data according to consistency
criteria.   The wind profiler/RASS and sodar data  will be  screened by an
automated editor (Wuertz  and Weber, 1989) after each 24 hour collection period.


This data will in turn be inspected by qualified staff and flagged if required. The
resulting ASCII files of winds and temperatures, along with graphical displays,
will then be available for quick dissemination by  diskette  or by  electronic
       The in situ surface meteorological data will require similar inspection and
will be averaged into one hour blocks.  These data will be available for similar

       Data Processing and Analysis

       The  senior scientific staff at NOAA/WPL will cooperate closely with the
modelers to ensure that level one data are readily available to them in a useful
form.   NOAA/WPL  scientific staff will take  leadership in analyzing wind
profiler/RASS, sodar,  tethersonde, radiosonde, and surface meteorological data
to gain insight  into the often complex transport  processes in the project region.
This effort will require the use of various types of data from project collaborators
outside of NOAA. The surface pressure array may be very critical in extending
the understanding of material transport over a larger area than that covered by
upper  air wind measurements.

7.   Optical  Monitoring


      The optical  monitoring  plan  for  project MOHAVE consists  of two
fundamental aspects:

1)    View Monitoring

      View  monitoring documents the visual impairment of specific unique
      vistas under various air quality conditions.  View monitoring is primarily
      accomplished with 35mm color slide photography and 8mm color  time-
      lapse photography. Color slides provide high resolution documentation of
      the visible effects of uniform and layered hazes on the vista.  Digitization
      of the slides can be done to yield relative radiance fields that can be used
      to calculate  color contrast, average landscape contrast,  visual range,
      modulation depth, equivalent contrast,  and just noticeable change.  In
      addition,  slides of extremely clean days can be used as the basic input to
      present visual air quality scenarios.  8mm time-lapse photography captures
      the important spatial and temporal patterns of visibility events that  allow
      for a more in-depth understanding of visual air quality.

2)    Electro-Optical Monitoring

      Electro-optical monitoring measures the basic electro-optical properties of
      the  ambient atmosphere and aerosols, independent  of  specific  vista
      characteristics.   Monitoring will include measurements of the ambient
      atmospheric  extinction  coefficient (bext), and its scattering (bscj and
      absorption (b,,,,) components.  Primary operational monitoring techniques
      include  the  transmissometer  (bext),  nephelometer  (bK^,  and  filter
      absorption (babs). Temperature and relative humidity measurements,  taken
      simultaneously with electro-optical measurements, are mandatory to infer
      visibility  effects associated  with  chemical  and  physical  interactions
      between water vapor, liquid water, and aerosols.

      Project MOHAVE will  incorporate  current  state-of-the-art monitoring
instrumentation,  operating and quality assurance procedures, and data collection,
reduction, editing,  and reporting  protocols that have been developed for  the
IMPROVE monitoring program (ARS, 1990a;  ARS 1990b).

      View Monitoring


      Automatic 35mm and 8mm camera systems will be an integral part of the
optical monitoring for project MOHAVE. The spatial and temporal variations in
visual air quality captured by these systems will be used to:

      •     Document how vistas appear under varied  conditions;
      •     Qualitatively record the frequency that various conditions occur;
             e.g.  incidence  of uniform  haze,  layered haze,  plumes,  and
      •     Provide a quality assurance reference for collocated electro-optical
       •     Serve as a backup method to estimate the electro-optical properties
             of the atmosphere (if appropriate teleradiometric targets are in
       •     Support the calculation of advanced visibility indices;
       •     Support computer imaging studies;
       •     Provide  quality  media for  visually presenting program goals,
             objectives, and results to study participants, decision makers, and
             the public.

Systems based on the following cameras will be used:

       •     35 mm  cameras:     Olympus OM series
                                 Contax 136 and  167
                                 Cannon EOS series

       •     8mm time-lapse:     Minolta 601  series

      Standard operating procedures developed  for the IMPROVE monitoring
program will be followed (ARS,  1990a).

      Monitoring  Locations and Sampling Frequency

      The 35mm camera systems will be  located at all receptor sites and other
Class I Area sites.  The 8mm time-lapse systems will be located at Meadview and
various  scenic view points along the south rim of the Grand Canyon.   During
non-intensive periods, only 35mm cameras will operate,  taking three exposures
daily at  0900, 1200, and 1500 hrs.
      During intensive monitoring periods, 35mm cameras at the receptor sites
and at GCNP will take nine exposures  daily from 0800-1600 hrs.  Time-lapse
photography will take  1 frame per minute from 0800-1600 hrs daily.  Additional


view monitoring locations will be added as the study progresses.

       Electro-optical Monitoring

       Extinction Measurements

       The Optec,  Inc. LPV-2 long path transmissometer will be the primary
instrument used to measure b^ for project MOHAVE.  The transmissometer
incorporates a light detector (receiver) at one end of a specific atmospheric sight
path.  The receiver directly measures the illuminance of a constant output light
source (transmitter) located at the opposite end of the path.  Calibration of the
transmissometer accurately determines the inherent output of the transmitter.  The
transmission of the sight path can then be calculated:
       T      = transmission of sight path r
       Ir      = illuminance measured by receiver at distance r
       !„,     = calibration illuminance of transmitter

       By measuring the exact length of the sight path the average atmospheric
extinction coefficient of the path can be calculated:
       bext    = average extinction coefficient of sight path r
       T      = transmission of sight path r
       r      = length of sight path r

       During the past ten years, transmissometers have been developed, tested,
and  deployed "in the IMPROVE  monitoring network, National Park Service
IMPROVE protocol sites, and various other monitoring programs.  They have
become the accepted method for reliably making continuous precise, accurate, bext
measurements.  Standard operating and data reporting procedures developed for
the IMPROVE program  will be followed (ARS, 1990b).

      Scattering Measurements

      Integrating nephelometers will be used to measure b^.,.  The integrating
nephelometer measures b,^ by directly measuring the light scattered by aerosols
and gases in an enclosed sample volume.  The scattered radiation is integrated
over a large range of scattering angles.  Since the total light scattered out of a
sight path is the  same as the  reduction of light along the  sight path due to
scattering,  a  properly  calibrated  integrating  nephelometer  gives a  direct
measurement of b,,
       Nephelometer measurements  are involved in considerable controversy
because of the modification of the  ambient  aerosol as it passes through  the
sampling train and optical chamber.  The instrument heats the air thus lowering
the  relative  humidity  environment  of  the  aerosols.    This  leads  to  an
underestimation of ambient bscat.   Extreme efforts  have been made to operate
nephelometers as close to ambient temperatures as possible. The best results have
been a heating of approximately 1.5' C.  This is approximately a 10% change
in relative humidity,  which can lead  to  underestimation of  ambient  bsc4t
measurement.  In addition, nephelometers underestimate the scattering by coarse
particles (>  2.5jim in diameter).  As with the  transmissometers,  standard
protocols developed for the IMPROVE program will be followed (ARS,  1990b).

       Absorption Measurements

       Where collocated transmissometers and nephelometers are collecting data,
b^ will be estimated by subtracting b^^ from bext.  The term, babs, will also be
estimated by  absorption measurements  from channel A filters collected by  the
aerosol monitoring network.  These babs measurements will be average values for
the collection period of each filter.  Data from these  measurements  will be
available only for periods when aerosols measurements are taken.

       Temperature and Relative Humidity Measurements

       Accurate air temperature and relative humidity data are critical to establish
the relationship between ambient aerosols and visibility effects.  Small changes
in relative humidity, especially above 70%, can dramatically affect aerosol size
and optical characteristics.  Rotronic Instrument Corporation Model MP-100F
sensors will be used in Project MOHAVE.  The MP-100F combines a 100 ohm
platinum temperature sensor with an enhanced hygroscopic polymer film humidity
sensor to provide an integrated air temperature/relative humidity device that will
maintain a 2% relative humidity measurement accuracy over the range of 0-100%
relative humidity. These sensors will be operated with every transmissometer and
nephelometer in the Project MOHAVE network.

      Monitoring Locations and Sampling Frequency

      Transmissometers and nephelometers will operate continuously through the
year of Project MOHAVE at various locations.    Data from instruments
specifically installed for Project MOHAVE as well as data from other existing
networks will be collected for inclusion in the Project MOHAVE data base.
Measurements from the following sites, with sponsoring networks, are listed in
the table below.  Data will be collected and archived as hourly averaged values
for the entire monitoring year.
Transmissometer and Nephelometer Monitoring Locations in the Southwest
Bandelier NM
Big Bend NP
Bryce Canyon NP
Canyonlands NP
Chirichahua NM
Grand Canyon NP
south rim
Long Mesa
Mesa Verde NP
Page, Arizona
Petrified Forest NP
San Gorgonio W
Spirit Mt. , Nevada
Tonto NM
Guadalupe Mts. NP
Sponsoring Network


1 |






8.   Emission Inventory and  Characterization


      Emission inventory and  source characterization are necessary for  the
deterministic   and  receptor  modeling.     Receptor  models  need  source
characterization for the main sources of interest.  This involves compiling a ratio
of elements that uniquely identifies a source and can be monitored at the receptor
sites.   The emission inventory is used  to  supply  input to the deterministic
modeling.  Emission inventory consists  of quantifying  the emission rates of
substances of interest from all sources that may be reasonably expected to impact
the study area.  For Project  MOHAVE, sulfur dioxide  emissions are of the
greatest interest.   The SO2 emissions from MPP  will  be modeled  with  the
transport and chemical models described in section 10. The level of modeling of
other sources is still being investigated.   Project MOHAVE intends to include
transport and  first-order chemical modeling of other significant  sources of SO2,
including the  southern  San  Joaquin Valley, the Los Angeles  Basin,  other
powerplants,  and copper smelters  within the  domain of the meteorological
modeling area.   The source profiling  will also detail  the primary particle
emissions in order to assess whether primary particles contribute significantly to

      Review of Existing Data and Inventories

      The emission inventory used in the SRP NGS study (Systems Applications
International,  1991) will be reviewed.    State  air pollution agencies will be
consulted about emission data, especially regarding any changes for  the main
sources  of SO2. The power output of the MPP will be used to determine the
emissions from the MPP. The operational status of other large SO2 sources will
also be  checked and emission rates adjusted if necessary before the  modeling

      MPP Stack Sampling

      Stack sampling will be done to determine the composition and quantity of
MPP emissions, which are needed for the  receptor, hybrid and deterministic
modeling analyses.  This component of the study has not yet been planned. The
study plan will be updated when details of the stack  sampling are known.

9.   Centralized  Data Management  and Validation


      EPA/EMSL in Las  Vegas  will be the data managers  for Project
MOHAVE. Information will be obtained from the following sources:

      SOURCE                       DATA TYPE

NOAA/WPL                     Surface and Upper Air Meteorology

Brookhaven National Laboratory    Tracer concentrations

UC-Davis                       Aerosol  and SO2

Air Resource  Specialists           Optical and Surface Meteorology

EPA-RTP/AREAL               Aerosol

National Meteorological Center    Surface and Upper Air Meteorology

Colorado State University         Meteorological Modeling

CAPITA- Washington  University   Monte Carlo Modeling

The data to be collected  are  described in more detail in Sections 3-7.  A data
management  and  validation plan  will be developed by the data  management
coordination committee.  A sketch of the expected elements of the plan to be
developed is presented in the remainder of this section.
       Two levels of validation (Levels 1 and 2) will be systematically applied.
Level 1 (univariate) validation involves checking the data for outliers, rates of
change, proper indication of time and location  of data,  etc.   In  Level  2
(multivariate) validation, consistencies among variables and  the appropriateness
of  spatial and  temporal patterns  are investigated.   For  example,  the  light
scattering (b^,)  measured by a  nephelometer  should be  less than  the  total
extinction (bext) obtained by a transmissometer.  Level 3 validation occurs during
the data analysis.  If data inconsistencies are found, the documentation  regarding
the questionable observation is examined for correctable errors (e.g. transcription
errors). Uncorrectable, suspect data are flagged, but not removed from the data
set.  Data known to be incorrect and not recoverable are removed from the data
       Each group responsible for collecting data will perform at least Level 1
validation. UC-Davis will do a partial Level 2 validation of the aerosol  data.
The data managers at EPA/EMSL Las Vegas are responsible for the Level 2


validation.  Systematic procedures and protocol for the Level 2 validation will be
developed and fully documented prior to releasing the data.  Level 1 protocols
utilized  by organizations  responsible  for each  data subset  will  also  be
documented.  A computerized listing of the data will be prepared.  Level 2 data
will be distributed to data analysts and other interested parties.  At the end of the
study, all  data will be assembled and documented.   A brief discussion of
validation  conducted by some  of the  participants appears in  the  following

       Aerosol Sampling (UC-Davis)

       A number of the measured or derived parameters are interrelated.  This
allows data intercomparisons as a method to evaluate  system performance and
check for outliers. The intercomparisons made are listed below:

       (1)    Fine sulfur vs. fine sulfate
       (2)    Fine sulfur vs. PM-10 sulfur
       (3)    Fine hydrogen vs. fine mass
       (4)    PM-10 hydrogen vs. PM-10 mass
       (5)    Sum of fine components vs. fine mass
       (6)    Sum of PM-10 components  vs. PM-10 mass
       (7)    Elemental carbon vs. optical absorption
       (8)    Organic carbon vs. nonsulfate hydrogen
       (9)    Fine mass vs. extinction
       (10)  PM-10 mass vs.  extinction
       (11)  Fine mass components vs. extinction
       (12)  PM-10 mass components vs. extinction

Details of  the quality assurance arid data validation are given in Pitchford and
Joseph (1990).

       Transmissometer Data

       The transmissometer data is subjected to three levels of validation. In the
first level, validity codes reflecting transmissometer instrument operation are
added to the raw transmissometer data files.  In the second level, data and
validity codes are checked  for inconsistencies using a screening program.  The
bext data are  adjusted for lamp drift of 2% per 500 hours of  lamp-on time.
Validity codes are added to all  data.  The  third level, consists of 2 steps,

       (1)  Calculation of uncertainty values for all data; and

       (2)  Identification of bext values affected by weather.

Validity codes for be;ct include:

       0 = valid
       1 = Invalid:  Site operator error
       2 = Invalid:  System malfunctioned or removed
       3 = Valid:   Data reduced from alternate logger
       4 = Weather:  Relative Humidity >  90%
       5 =          bext >  maximum threshold
       6 =          Abext > delta threshold
       7 =          bext uncertainty > threshold
       8 = Missing: Data acquisition error
       9 = Invalid:  bext below Rayleigh
       A = Invalid: misalignment
       L = Invalid: Defective lamp
       S  = Invalid:  Suspect data
       W =  Invalid:  Unclean optics

       Radar wind  profilers and RASS

       Real-time processing consists of a Doppler spectra peak picking routine
which searches for spectra peaks beginning from the highest level of good signals
to the lowest gate.  As the routine searches for peaks in a downward direction,
it requires consistency from gate-to-gate.   If a peak shifts beyond a given
threshold between gates,  that peak is rejected.  To help eliminate ground clutter,
the algorithm also rejects peaks near zero  velocity if a secondary peak away from
zero is available.  After  the peaks, or first moments, for each individual radial
are chosen in this way,  a consensus averaging is performed.  This technique
requires at least 50% of the points on each gate of each radial for a 55 minute
period (5 minutes for RASS  temperature) to fall within a bin of 2 m/s in width
before the individual points in the bin are averaged.  If less than 50% of the
points fall within the bin, the radial component is  flagged as bad and is not
available for that period.  A similar technique is used for the RASS derived
temperatures with a  bin threshold of 1 ° C.
       The normal  post-processing  quality  assurance procedures  consist  of
applying a time/height editor, normally referred to as the Weber/Wuertz editor
(Weber and Wuertz, 1989), to each 24 hour period of one hour averaged profiler
wind data or 5 minute averaged temperature data (one 5 minute average provided
each hour).  The editor assesses the  neighborhood of each point for consistency
in both speed and direction (or temperature), allowing for a larger tolerance for
direction differences at lighter wind speeds.  The tolerances  are  adjustable and
depend on the  prevailing meteorology  during a particular  experiment.   The
neighborhood size is also adjustable, but usually the eight adjacent points are
chosen, if available.  This editor has proven to be very powerful in eliminating
outlier points.  The results of this processing provide the Level 1 data.

       NOAA/WPL is experimenting with applying a more sophisticated form of
this editor on the radial moments before performing an hourly average.  The test
data are from the 1990 San Joaquin Valley Air Quality Study.  The technique
requires considerable processing.  The decision to use  this technique for Project
MOHAVE depends on the quality of the data, which in turn depends on site
       The post-processed, Level 1 data will be compared with optically tracked
rawinsonde (airsonde) wind and temperature profiles measured at each location.
Several rawinsonde profiles will be available at each of the wind profiler locations
representing different stability and meteorological conditions at each site.

10.  Descriptive Data Analysis and Interpretation


      A large quantity of data will be collected in support of Project MOHAVE.
The descriptive  data  analysis and  interpretation component of  the  study is
intended to summarize the main features of the data as well  as especially
interesting cases,-and offer physical explanations whenever possible.  In contrast
to the attribution analyses described in Section 11, this section will organize the
data in a manner that will allow inference of effects from different sources, but
will not generally be quantitative sufficiently to permit source apportionment.

      Descriptive Statistics

      Descriptive  statistics will  include  calculation of  means,  standard
deviations, skewness,  and extreme values of the  variables.  In addition,  time
series of the data will be  presented.  These will include time  series  of the
extinction coefficient  (bext), tracer,  sulfate, nitrate,  organics,  light-absorbing!
carbon,  fine soil, and various trace  elements and  meteorological variables, for
example.  Correlations between variables will also be calculated.

      Extinction Budget

      Light extinction is caused by scattering and absorption by particles and
gases.   In  general,  particle  scattering is the primary component of extinction,
although in remote areas of the  southwest, scattering by gases that compose the
atmosphere (Rayleigh scattering) is  a significant fraction on the clearest days.
Black carbon (from diesel  engines,  forest fires, etc.)  is the principal  agent of
particle absorption,  and is occasionally an important contributor to haze in the
study region.  NO2 is the  only common gaseous pollutant that absorbs in, the
visible portion of the spectrum and is not likely to be a significant contributor to i
haze in  GCNP.                                                         !
      The extinction budget analysis involves determining the contribution tp
extinction by all the major aerosol components.   There are two fundamentally
different approaches to estimate the extinction budget. A statistical approach uses
multivariate analysis to explain the optical parameter (bext or b,^ by a linear
combination of the components. These components are the concentrations of the
pollutant species (e.g., crustal,  sulfate, nitrates, elemental and organic carbon,
etc.) multiplied by best-fit determined coefficients interpreted  as extinction
efficiencies. The hygroscopic particle species (e.g., sulfate and nitrate) include
a function of relative humidity to  incorporate the effects of water  upon the
extinction efficiencies of these species.
       An externally  mixed aerosol  (i.e.  separate aerosol components are not
contained within the same particles;  for example, sulfate-coated crustal particles


would not constitute an external aerosol mixture) is implicitly assumed by the
statistical approach for extinction budget  analysis.  The extent to which this
assumption is true, and the implication of it being violated, are hard to estimate
in any individual  situation.  In general,  the  greatest impact of non-external
mixtures is thought to be associated with an interpretation of how changes in
aerosol composition would affect atmospheric  optics.  In other words, there is
increased uncertainty associated with the prediction of how visibility will respond
to changes in emission caused by .violation of this assumption.
       In  addition  to the concern  about implied assumptions, any use of
multivariate statistics carries with it the concerns caused by use of possibly highly
covariant independent parameters, and the use of measured parameters with large
differences in relative measurement uncertainty.  Both of these concerns can
result in biased results.  However, there are standard approaches to detect and
minimize the impacts of these concerns.
       The other approach to estimating  extinction efficiencies for the various
aerosol components  is by first principle calculations (Mie  Theory).   These
calculations  require  as input,  certain paniculate characteristics such as  the
distribution of particle size, shape, and indices of refraction. Generally, the size
distribution can be estimated from size segregating sampler measurements.  For
Project MOHAVE,  this will be done with the DRUM  impactors  for  some
components,  such as sulfur and crustal components, but not for others, such as
organic carbon and nitrate species.   A functional relationship between water and
the hygroscopic particles must be assumed to estimate its effects on particle  size.
Particle shape is generally assumed to be spherical, and the refractive indices are
assumed to be the same as the bulk indices for the various  measured particle
chemical components.  Assumptions must also be made concerning the nature of
the aerosol mixture (i.e., external,  internal, or some  combination) in order to
calculate the extinction efficiencies of the components.  The extent to which these
deficiencies and assumptions affect the calculated extinction is unknown.
       In spite of the  uncertainties  discussed above, extinction budget analysis
done by the two approaches generally results in similar extinction efficiencies.
Since Project MOHAVE will use both Mie Theory and  statistical approaches, the
results can be intercompared for  consistency, and reconciled with  extinction
efficiency values from the literature to arrive at best estimates of the  extinction

       Empirical Orthogonal Function Analysis

Empirical, orthogonal function  (EOF) and possibly other  types of eigenvector
analysis will be done to help summarize the data and gain insights into possible
physical mechanisms at work. When working  with large amounts of data,  EOF
analysis is especially useful by effectively reducing the  dimensionality of the data
set.   A large  number  of observations at  many locations can be reduced into  a
reasonable number of spatial patterns (eigenvectors), with a time series associated

with each eigenvector showing the time variability of each pattern.
       The EOF analysis is purely a statistical technique that attempts to account
for most of the variability in a data set by a few eigenvectors.  Although no
physics is explicitly included in  the analysis, the data  set represented by the
eigenvectors  is  certainly affected by physical.processes.   EOF  analysis in
conjunction with sound physical reasoning, including knowledge of meteorological
conditions, location of emission  sources,  etc. can  help  in  the  formation of
hypotheses and provide a qualitative check of receptor and deterministic modeling
       The variables for which EOF analysis is likely to be done are sulfate, SO2,
tracer, elemental carbon, organic carbon, fine soil, and certain trace elements.
EOF analysis of the vector wind field may be done as  well. EOF analyses of the
modeled output wind and concentration fields will be investigated as a method to
help organize the large amount of model output. Additional EOF analysis using
two (or  more) parameters such as  sulfate  and tracer can be used to identify
common jointly occurring patterns of more  than one parameter.

       Meteorological Classification

       A meteorological classification scheme will be developed and applied to
the study years and several previous years.   The scheme will classify days into
types on the basis of similarity of meteorological parameters.  There are several
reasons for doing a classification.  One reason is to  compare the frequency of
each weather pattern during the study year with other years  to determine how
representative the  study year  is.   Each pattern is likely  to  have transport of
visibility  affecting  pollutants from  different areas;  the relative  frequency of
patterns  for the study year compared to long-term averages can help put the
impacts during the study year into perspective. It also provides a logical method
of stratifying the data of the study year into a manageable number of patterns.
Averaged spatial patterns of sulfate, etc. along with the variation within each
pattern can reveal the main pathways for transport of both hazy and clear air into
the study area.  Contributions from individual source  areas may also be inferred
from the concentration fields associated with each pattern.
       The meteorological classification scheme can  aid in the interpretation of
the EOF analyses.   The time series of the EOF  analyses indicate  the times a
particular  eigenvector  is  significant.   By  determining the  corresponding
meteorological pattern most commonly  associated with each eigenvector,  it is
easier to interpret the physical factors associated with the eigenvectors.
       The classification scheme  will also be used to study the MPP outage of
June-December 1985. Sulfate concentration levels and spatial patterns associated
with each weather pattern will be compared for the outage year and other years
with SCENES  data.  This will help put bounds on the contribution  of MPP to
regional sulfate levels.
       Of critical importance in the classification scheme are surface and upper


air wind speed and direction, atmospheric moisture and thermal stratification.
The wind data are necessary to account for the transport and dispersion properties
of the flow. Moisture is necessary to determine the potential for aqueous phase
oxidation of SO2 to sulfate and washout. Thermal stratification is needed to know
if pollutant emissions are likely to remain trapped in basins or are mixed through
a deeper layer of the atmosphere.

11.  Attribution


      The attribution analysis will be done using a variety of analytical tools;
particularly deterministic, hybrid, and receptor models; and the MPP emission
modulation study.   Deterministic modeling is an  approach  that  attempts to
explicitly account for physical and chemical processes transporting and acting on
emissions from a source.  Receptor models use measurements made at the area
of concern (receptors) along with characterization of the emissions from sources
potentially  affecting the receptors.   The contribution  of  each  source  to
concentrations at the receptors is  determined statistically through multivariate
analysis techniques which link the  sources to  the measured concentrations.
Hybrid  models use a combination of deterministic and  receptor  modeling
techniques. Apportionment implies determining the concentration of sulfate at the
receptor areas resulting from MPP and other  sources.  The apportionment of
secondary aerosols  such as sulfate is a complex problem as noted by the National
Research Council  (1990) and others.  Transport,  dispersion,  deposition, and
transformation must be accounted for.
      Results from the extinction budget analysis will be used in conjunction
with the sulfate attribution to determine the fractional contribution of MPP sulfate
to the extinction  coefficient.   The  effect of primary  particles will also  be
considered.  The next step is to evaluate the perceptibility of the contribution of
MPP to the extinction coefficient. Finally, the question of the effect of reducing
emissions  from MPP upon visibility will be addressed.  Each of the main study
components used in the attribution analysis is described in the following sub-
      The results  from the various models and analyses will be  compared and
reconciled. Reconciliation is a critical component of the analysis. If results from
a particular model or analysis  cannot be reconciled  with other analyses, the
results will not be  used.  The range of uncertainty  in each calculation  and the
reconciled results or consensus will be estimated.
      A major area of concern that has been expressed by some is the
possibility for misuse of tracer data for source  apportionment.  Specifically,
two issues have been expressed: (1) that any tracer  level above background
measured at the receptor sites will be interpreted as attribution of visibility
impairment by tracer sources; (2) that tracer data will be incorporated into
analyses inappropriately by  repeated regression analysis  with whatever
parameters  and  formulisms  are  needed until  a  statistically significant
relationship is found, though no physical relationship is evident.
      Project MOHAVE planners do not interpret the appearance of any
tracer  above  background as  sufficient  criteria  to  indicate  visibility
impairment and will contest any who make  such a  claim.  To demonstrate
good faith and concern for appropriate scientific methods, Project MOHAVE


will arrange for tracer data to be withheld from all who have any role in
attribution analysis until such tune as physically meaningful empirical source
attribution  formulisms have  been developed based upon other  Project
MOHAVE data.  This will be done to promote the development of physically
reasonable models prior to the availability of tracer data, and to avoid the
appearance of forcing the models to fit preconceived notions.

      Deterministic Meteorological Modeling

        Deterministic meteorological modeling is based on fundamental physical
conservation relationships. These relationships include conservation equations for
momentum, temperature, mass, and the three phases of water.  Meteorological
modeling for Project MOHAVE will be done by Colorado State University using
the Regional Atmospheric  Modeling System (RAMS).  A brief overview of
RAMS is given in Appendix 6.  Additional information about RAMS appears in
Pielke et al., (1990).  A dedicated super workstation (IBM RISC) will be used
for the modeling. The model will provide detailed wind and turbulence fields and
a prediction of cloud  height and location.  Cloud predictions will  be checked
against satellite photographs.
      The meteorological domain for the simulations will cover the southwestern
United States.  To obtain better terrain resolution near MPP, a telescoping nested
grid will be used. In a nested grid approach,  the larger scale results  provide the
boundary conditions for input into a finer scale modeling domain. The entire one
year study period will be modeled. For selected cases from the intensive study
periods, modeling with much finer resolution will be done. The preliminary grids
to be used for the analysis are shown in Figure 10.  Grids  1 and 2 will be used
for the year-long study; the case studies will also use grids  3, 4, and 5.   The
horizontal and vertical number of grid points,  and the horizontal grid spacing for
each grid are shown below.
       Grid     # of grid points    Spacing (km)

       Grid 1   x=100  y= 60 z=44    32

       Grid 2   x=104  y= 72 z=44    8

       Grid 3   x=144  y=144 z=44    2

       Grid 4   x= 80  y=80 z=44    0.5

       Grid 5   x= 80  y=80 z=44    0.5

                                        Figure 10.  Meteorological modeling grid.

       The model is initialized every 12 hours using the analysis field supplied
the National Meteorological Center's Nested Grid Model.   The Nested Grid
Model uses surface and upper air data to generate initial fields of variables such
as pressure, temperature, moisture and wind.  The CSU RAMS modeling takes
this initial field and generates mesoscale fields for the next 12 hours before being
       The use of data from  the wind profilers will be investigated for  use- in
four-dimensional data assimilation. In this mode, the measured data are used to
adjust, or "nudge" the model results.  Data from the profiler site at Truxton are
the most likely to be used for nudging.  The Truxton  site is in relatively open
terrain and more likely to be representative of general flow in the  study area man
the other sites which are expected to be considerably influenced by local terrain
features. Data from the radar wind profilers not used in the nudging process will
be used to evaluate the performance of the model. The evaluation will be done
for every hour having modeled and wind profiler data. A quantitative comparison
between the predicted and observed  winds will results  in "figures of merit"  for
model predictions as a function of  meteorological conditions.   The  wind and
turbulence fields obtained  from  the deterministic meteorological  model will
provide the necessary input to calculate the transport and dispersion of MPP and
other emissions of interest.

       Transport, Chemical and Deposition Modeling

       Once the wind fields have been determined, a model is needed to account
for transport, chemical transformation and deposition.  The transport model to be
used is a version of the CAPITA Monte Carlo  model currently being developed
for EPA under a cooperative agreement with Washington University in St. Louis.
A copy of the cooperative agreement proposal, which describes the modeling
approach in more detail, appears as Appendix 7.  The model is being developed
specifically for visibility related studies.  Evaluation and calibration of the model
is being done using data sets  such as IMPROVE, SCENES and NESCAUM.
Modifications to the model to fully utilize the wind, turbulence, and moisture
field supplied by the meteorological  model may be necessary.
       In the modeling  approach,  simulated  pollutant  quanta  (particles) are
"emitted" from each source.  These quanta are moved in fixed time increments
using wind fields supplied  by  the meteorological model.  During transport  the
pollutant  quanta  are  subject  to chemical transformation and  removal.  The
dispersion is achieved by imposing a randomized perturbation to the trajectory at
each  time step.   Transformation and removal  are also imposed as  stochastic
events at  each time step.  The result of the Monte Carlo simulation is  a large
number (lOMO6) of pollutant "particles"  dispersed geographically for every time
step of the simulation.   The  model is considered  a  Monte Carlo simulation
because of the probabilistic treatment of transport, transformation, and removal.

      The model  will make use  of the  turbulence field  generated by the
meteorological model to perturb the trajectory. The moisture fields given from
the meteorological model will be used to select between wet (heterogeneous) and
dry (homogeneous) conversion  rates  in  the SO2 to sulfate  transformation
parameterization. Details of the modeling methodology are still to be determined.
The model is based on that described by Patterson et al (1981).

      Hybrid and Receptor Modeling

      Measurements of endemic and artificial tracers will be used to estimate the
transport and dispersion of MPP and other sources. The transport can be verified
by  checking  with  trajectories  given by  the  meteorological  model. .  The
transformation and deposition of SO2 and sulfate are also necessary;  these may
be parameterized based  upon such information as  solar  radiation, moisture
(especially clouds), oxidant availability, and vertical mixing.  The hybrid models
will  use tracer measurements  to  account  for  transport and  dispersion,  and
parameterizations to account for deposition and transformation.
        Versions of the chemical mass balance (CMB), differential mass balance
(DMB), and possibly the tracer mass balance regression (TMBR) models will be
used.  CMB uses the relative ratios of natural or man-made tracers at the sources
and receptor locations to apportion primary species for each measurement period.
CMB will be  used to  apportion  primary  species using  the high  volume
dichotomous sampler data.  DMB, a hybrid model, uses trace material to establish
dispersion factors and calculates the effects of deposition and oxidation. TMBR
uses the variation of trace  material over time to estimate primary or secondary
aerosol  contributions from each source.   In its original formulation, TMBR
requires a constant tracer to SO2 emission  ratio.  However, Project MOHAVE
will use a tracer emission rate that will be only approximately proportional to SQ
emissions.  TMBR may be used in an exploratory mode to investigate the effect
of departure from model assumptions.  Any use of the results must acknowledge
and quantify the effects of  departure from the model assumptions.
      In DMB and TMBR it is assumed that each source has a uniquely emitted
tracer associated with it. The best available emissions and source characterization
data will be used to identify unique tracers for each significant source.  If unique
tracers are not available,  CMB may  be applied first to partition the ambient
species  concentrations into components attributable  to the various  groups of
        The NAS review of WHTTEX noted a number of concerns about the use
of DMB and TMBR. A summary of the NAS WHITEX comments  and the steps
that will be taken by Project MOHAVE to  help resolve these issues  appears as
Appendix  8.  The  DMB and TMBR models will be modified to help  alleviate
concerns in the NAS review of WHITEX by taking advantage  of the  more
detailed meteorological fields generated by the meteorological modeling that is a
part of Project MOHAVE.  For example, the effect of moisture upon conversion


rates for both DMB and TMBR will utilize the moisture fields generated by the
meteorological model.
         As mentioned above,  the implementation  of DMB will  differ  from
WHTTEX by incorporating physical processes in a more robust manner.  In
Project MOHAVE, the effect of moisture on sulfate formation will be treated
more rigorously than in  the WHTTEX study.  In addition to surface moisture
measurements, the deterministic meteorological model will give calculations of
moisture at many vertical levels.  This information will include prediction of
clouds, which can be compared to  satellite photos and  surface observations.
Rather  than scaling linearly with surface  relative humidity,  the probability of
plume-cloud interaction will be estimated  and  used  to assign an SO2-to-sulfate
conversion rate.  Rates of sulfate formation occur rapidly in clouds, and much
slower  without clouds.  Different conversion rates based upon whether or not
clouds are present should more appropriately account for the effect of moisture
on conversion rates.
       In DMB, as currently formulated, deposition and conversion rates are
constant; the  assumed rates  are multiplied  by plume age  to give sulfate
concentrations and deposition loss. Trajectory calculations are used to give plume
age.    Dispersion  is accounted for  by  ratioing  ambient  trace  material
concentrations attributable to  a source by known  trace  material release rates.
In Project MOHAVE variable conversion  rates will be used, such as described
above.   With the  deterministic  meteorological modeling wind fields,  reliable
plume age calculations should be possible.
       The equations for, and assumptions used in, CMB, DMB, and TMBR as
presently formulated are  given in Appendix 9.

       Extrapolation of  Intensive Study Periods to the Long-Term

       To determine longer term impacts to visibility at GCNP, it is necessary
to extrapolate from results of the intensive study periods.  This will be a two-step
process; the first step will relate the entire 12 month study period to the intensive
period, while the second will extrapolate from the 12  month period to a multi-
year period. The first step involves application of source-oriented and hybrid
models, which will be developed and evaluated with intensive period data, to the
meteorology and air quality data for the entire study period.  In the second step
the relative frequency of long-term meteorological patterns will be compared with
those of the study period.
       Deterministic models  will be evaluated and calibrated using the  more
complete data of the intensive periods.  The resulting models  will then be run
with data from the entire study period. For all modeling analyses a portion of the
data may be withheld in order to independently test the models.
       During the intensive studj periods,  hybrid modeling will use the artificial
tracer results  for MPP  and  any other sources  tagged with artificial  tracers.
Hybrid models will also use endemic tracers for the remaining significant sources.


Results from hybrid models based on endemic tracers will be compared to results
of the same models  using artificial tracers to evaluate  the utility of endemic
tracers.  If successful, models using endemic tracers will then be applied to the
entire study period (covering a complete annual cycle) and used in conjunction
with the deterministic modeling analysis.
       The representativeness of the study year to longer term average conditions
will be studied. It should be acknowledged that significant year to year variability
in meteorological conditions occurs and that the likelihood of any given year
being "typical" is not high. The frequency of occurrence of each meteorological
regime identified in the meteorological classification process described in Section
11  will be compared for  the study year and other years for which  data are
available.  Where they exist, optical and air quality measurements from previous
years will  be  compared to the study  year  measurements.  The frequency of
occurrence of each pattern for the study period and longer term average  can then
be compared to put the study year into perspective.

       MPP Emission Modulation Study

       The MPP  was  inoperative  for a  seven month period from  July to
December 1985. This presents a unique opportunity for investigating the effects
of MPP.   The MPP emission modulation  study, discussed  in Section 2  and
Appendix 5 is  a potentially powerful receptor approach to estimate the extent of
MPP contributions to downwind sulfate levels.   The analysis will be conducted
by  using  a meteorological classification scheme to control  for year-to-year
variations  in meteorology, and comparing measured  sulfate at Spirit Mountain,
Meadview and Hopi Point for periods of varying MPP and other SO2 emissions.
The study will include the  following elements:

       Independent statistical analysis of the experiment.

       Chemical analysis  of  all filters.  (Quality  assurance will be evaluated
       through comparison of current results to past data through regression and
       time series analysis).

       Classification   of  the  synoptic  and  mesoscale  weather   patterns
       (meteorological regimes) affecting transport of the MPP plume.

       Deterministic  wind  field, transport, and dispersion modeling for each of
       the meteorological regimes.

       A detailed compilation of regional SO2 emissions data for the control and
       outage  period to allow  an accounting for variation in SO2 emission

       All data manipulation will be performed in the "blind" to avoid charges
of bias or data selection.  Results of the study will be used along with the
modeling and other analyses to estimate the effect of MPP on visibility at GCNP.

       Framework for Interpreting Results

       In a complex program such as this, a sound plan for compilation of results
is as important as the collection of high quality and representative data and the
performance of appropriate interpretive analysis.  Development of an approach
to  organize the results from this program helps to focus attention and resources
on critical steps for the entire program and communicate those ideas to others.
       Just as it is inappropriate .for worst case  results  to receive  primary
attention, it is also inappropriate to  dwell on average  or typical conditions,
especially for an  instantaneous effect  such as visibility.  The 12  month study
period  with  hourly deterministic model results requires some  method for
summarizing  the results of the study that avoids these pitfalls.  A preliminary
conceptual framework for summarizing the results of Project MOHAVE is shown
in the table on the following page.   The key idea is the stratification of time
periods based upon
the locations with  respect to  GCNP of MPP  emissions  and those of other
significant sources, such as  from southern California.  These would be based
upon the modeling  studies.  Another  stratification is whether the  plumes have
undergone wet or dry chemistry (based upon modeling results and observations).
If useful, other stratifications could  be developed.   The frequency  of each
condition, the average and standard deviation of the percent sulfate from MPP,
the percent of extinction from MPP and a measure  of the perceptibility  of the
MPP impact is estimated for the  study period.
       Extrapolation to a  long-term average  will be done through the use of a
meteorological classification scheme  as  previously described.   This  type of
approach provides an efficient manner of presenting the magnitude and frequency
of estimated MPP emissions on GCNP over a long-term period that could be used
to evaluate the significance of existing impairment.

   Conceptual Framework for Summarizing Project MOHAVE Results
GCNP Impact &
MPP & Other
Sources Dry
MPP & Other
Sources Wet
MPP Alone Dry
MPP Alone Wet
Other Appropriate

% Sulfate

% Extinction

Measure of

SCA refers to the urban and industrial areas of southern California.

12.   Overall  Quality Assurance


       An independent quality assurance audit will be done by ENSR. The major
emphasis of independent quality assurance in  Project MOHAVE will be upon
verifying the adequacy of the participants' measurement procedures and quality
control procedures, and  upon identifying problems and making them known to
project management.  Although  routine audits  will play a role, major emphasis
will be placed upon  the efforts  of senior scientists  in examining methods and
procedures in  depth.  This approach  will be followed because fatal flaws in
experiments emerge not from incorrect application of procedures by  operators at
individual  sites  or  laboratories,  but rather from  incomplete  procedures,
inadequately tested methods, deficient quality control tests, or insufficient follow-
up of problems.

       System Audits -  Study Planning and Preparation

       Senior  auditors will review study design documents to ensure that all
measurements  are  being planned  to produce  data with  known precision and
accuracy.  The auditors will verify that adequate communications exist between
measurement and data analysis groups to ensure that measurements will meet data
analysis requirements for precision, accuracy, detection limits,  and  temporal
resolution. Quality control components of the measurements will include:

       Determination of  baseline  or background  concentrations and  their

       Tests for sampler contamination.

       Adequate and precise measurement of aerosol and tracer sampler volume
       and time.

       Blank, replicate,  and collocated samples.

       Assessment of lower quantifiable limits (LQL),  and determination of
       measurement  uncertainty at or near the LQL.

       Regular calibrations and calibration checks, traceable to standard reference

       Procedures  for collecting QC test data  and for calculating and reporting
       precision and  accuracy.

      Periodic QC summary reports by each participant.

      Documented data validation procedures.

      Verification  of  comparability  among  groups  performing  similar

      A senior auditor will visit each measurement group, laboratory and data
management and analysis group prior to the intensive field studies to verify that
adequate progress is being made toward beginning measurements on schedule and
within acceptable quality limits.   A thorough review of written procedures will
be part of this evaluation, including a review of all standard operating procedures.
Issues to be addressed include:

      Availability of equipment and supplies.

      Manpower availability.

      Readiness of written procedures and data collection protocols.

      Adequate sample ID and sample tracking system.

      Thoroughness of method evaluation tests.

      Understanding of QC procedures and adequacy of protocols for collecting
      QC test data.

      Testing of software used for data management, data validation, and data

      Measurement System and Performance  Audits

      Audits of the field sites, the laboratories, and the data management and
analysis center will be conducted once during the study, probably at the beginning
of the winter intensive measurement period.  System audits will verify that the
items described  in the system audits  section are being applied.  Performance
audits will include:

      Field  sites -  Instrument calibration checks, leak-checks  on aerosol and
      tracer samplers, and on the tracer injection system.

      Laboratories  - Relabeling of existing samples by ENSR and reanalysis by
      the study laboratories to verify precision and reproducibility.  Submittal
      of prepared  samples  of known concentration, where needed.  If the


       laboratory already participates in a regular intercomparison program  or
       if it uses standards directly traceable to NIST, then a system audit will
       verify this, and no additional samples will be prepared.

       Data management  -  Manual calculation  of  derived concentrations and

       Data analysis - Manual data traceability  tests to verify pre-analysis

       Based on audit results  and discussions with project  management,  the
auditors will identify problems which have the potential to jeopardize data quality.
They will provide immediate feedback to operational personnel and will provide
letter reports  following  the audits.  Corrective action  request  forms,  to  be
completed by operational personnel and returned to the auditor, will verify that
problems  have been addressed.  Throughout the study, the auditors will review
the participants' QC summary reports.


       ARS,  1990a: Visibility monitoring and data analysis using automatic
camera systems: standard operating and quality control procedures document. Air
Resource Specialists, Inc., Fort Collins, CO.

       ARS,  1990b: Standard operating  procedures  for  monitoring ambient
atmospheric extinction and scattering coefficients. Air Resource Specialists, Inc.,
Fort Collins,  CO.

       Cahill, T.A.,  P.J.  Feeney,  R.A. Eldred  and W.A.  Malm,  1987:
Size/time/composition data at Grand Canyon National Park and the role of
ultrafine sulfur particles. Transactions TR-10; Visibility Protection: Research and
Policy Aspects (P.S. Bhardwaja,  ed.).   Air pollution Control  Association,
Pittsburg, PA, pp. 657-667.

       Calvert, J.G., A.L. Lazrus, G.L.  Kok, E.G. Heikes, J.G. Walega, J.
Lind and C.A. Cantrell, 1985: Chemical mechanisms of acid generation in the
troposphere. Nature, 317, 27-35.

       Dietz, R.N., 1987: Perfluorocarbon tracer technology. From "Regional
and long-range transport of air pollution",  Lectures of a course held at the Joint
Research Center, Ispra, Italy, September 15-19, 1986, S. Sandroni, ed., pp. 215-
247, Elsevier Science Publishers, Amsterdam.

       Dietz, R.N., 1991: Personal communication.

       Draxler, R.R.,  1985:  One year of tracer dispersion experiments over
Washington, D.C., Atmos. Environ., 21, 69-77.

       Eldred,  R.A.,  T.A.  Cahill,  M.  Pitchford and W.C. Malm, 1988:
IMPROVE-a new remote area paniculate monitoring system for visibility studies.
Proceedings of the 81st annual meeting of APCA, June 19-24, Dallas, TX,  88-

       Freeman,  D. and  R.  Egami,  1988:  Dispersion modeling at Mohave
Generating Station. Report no. DRI-8525-F1.0 prepared for Southern California
Edison Co., Rosemead  CA, February 1988.

       Gaynor, J.E., D.E. Wolfe and Y. Mori, 1991:  The effects of horizontal
pressure gradients and terrain in the transport of pollution in the Grand Canyon


      Heikes, E.G., G.L. Kok, J.G. Walega and A.L. Lazrus, 1987: H2O2, O3
and SO2 Measurements in the Lower Troposphere Over the Eastern United States
During Fall. J. Geophys. Res., 92, 915-931.

      Koracin,  D., T. Yamada, B. Grisogono, T.E. Hoffer, D.P. Rogers and
J. Lukas, 1989:  Atmospheric  boundary layer in Mohave Valley.  Presented at
AWMA/EPA  specialty conference "Visibility and fine particles", October 15-19,
1989, Estes Park CO.

      Lagomarsino, R.J., TJ. Weber, N. Latner, M.  Polito, N. Chiu and I.
Haskel,  1989: Ground-level air sampling systems. In "Across North America
Tracer Experiment (ANATEX)", Vol. 1, R.R. Draxler and J.L. Heffter, ed.,
NOAA Tech.  Mem. ERL ARL-167, Silver Springs MD, January 1989, pp.  13-

      Latner, N., 1986: Tethered Air Pump System, Report EML-456, U.S.
Dept. of Energy Environmental Measurements  Laboratory, New York NY.

      Lee, Y.-N., J. Shen, P.J. Klotz, S.E. Schwartz and L. Newman, 1986:
Kinetics of hydrogen  peroxide- sulfur (iv) reaction in rainwater collected at a
northeastern U.S.  site. /. Geophys. Res., 91, 13264-13274.

      Malm, W., K. Gebhart, D. Latimer, T.  Cahill, R. Pielke and J. Watson,
1989:  National Park Service  report on the  winter haze  intensive tracer

      Murray,  L.C., R.J. Farber, M. Zeldin and W.H. White, 1990: Using
statistical analysis  to evaluate modulation in SO2 emissions. In Visibility and Fine
Particles, C.V.  Matthai, ed. AWMA, Pittsburg, PA pp. 923-934.

      National  Research  Council,  1990: Haze in the Grand Canyon -  An
evaluation of  the Winter Haze Intensive Tracer  Experiment.  Prepared by the
Committee on Haze in National Parks and Wilderness areas.  National Academy
Press, Washington D.C.

      Nelson, L. R., 1991: Personal communication.  May 8, 1991.

      Patterson,  D.E., R. B. Husar, W.E. Wilson and L.F. Smith, (1981):
Monte Carlo  simulation of daily regional sulfur  distribution - comparison with
SURE data and visibility observations during August 1977.  J. Appl Meteor., 20

      Penkett,  S.A., B.M.R. Jones, K.A. Brice,  and A.E.J. Eggleton, 1979:
The importance  of atmospheric ozone and hydrogen peroxide in oxidizing sulfur


dioxide in cloud and rain water. Atmos. Environ., 13, 1615-1632.

      Pielke, R.A., W.A. Lyons, R.T. McNider, M.D. Moran, D.A. Moon,
R.A.  Stacker,  R.L. Walko, and M.  Uliasz, 1990: Regional and mesoscale
meteorological modeling  as  applied to  air  quality studies.   Proc.  of 18th
NATO/CCMS Int. Tech. Meeting on Air Pollution Dispersion Modeling and Its
Application, 13-17 May 1990, Vancouver, British Columbia.

      Pitchford, M. and D. Joseph, 1990: IMPROVE Progress Report. Report
EPA-450/4-90-008, U.S. Environmental Protection Agency, Office of Air Quality
Planning and Standards, Research Triangle Park, NC, May 1990.

      Raabe, O.G., D.A. Braaten, R.L. Axelbaum, S.V. Teague and  T.A.
Cahill, 1988: Calibration studies of the DRUM impactor.  J. Aerosol ScL, 19,

      Richards, L.W.,  C.L.  Blanchard,  D.L. Blumenthal,  1991. Navajo
Generating Station Visibility Study: Executive  Summary (Draft  number 2).
Sonoma  Technology  Inc.  report  STI-90200-1124-FRD2, April 16,  1991.
Prepared for Salt River Project, Phoenix, AZ.

      Systems Applications International, 1991.  Deterministic modeling in the
Navajo Generating  Station Visibility Study (Draft Final Report).  Prepared  by
Systems Applications International, San Rafael, CA, January 17, 1991. Report
SYSAPP-91/004b.  Prepared for Salt River Project,  Phoenix, AZ.

      Saxena, P. and Seigneur, C., 1986: On the oxidation of SO2 to sulfate in
atmospheric aerosols.  Atmos. Environ., 21, 807-812.

      White, W., D.P. Rogers, T.E. Hoffer and J. Lukas, 1989: 1986 Mohave
Generating Station plume intensive  study.  Final report prepared for Southern
California Edison Co., February,  1990.

      Yamada,  T.,  1988:  Preliminary  simulations of wind, turbulence and
tetroon trajectories. Interim report prepared for Desert Research Institute, Reno,
NV, December  1988.

      Wuertz,   D.B.,  and B.L. Weber,  1989:  Editing  wind  profiler
measurements.  NOAA technical report ERL 438-WPL 62, U.S.  Government
Printing Office, 78 pp.

                                              Appendix 1
                          Project MOHAVE Update Summary - September 16, 1991
Description of Study Component
Responsible Party
Overall -     12 months starting in Sept. '91
Intensives -   Two 4 to 6 week intensives,  (1) January '92, (2) July and
             August '92.
Conduct review of available emission data and compile inventory.

Do source profiling of MPP during a portion of each intensive period.

Continuous source sampling at MPP: SO2, NOX and paniculate
concentrations (plus frequent panicle composition). *
Bruce Polkowsky,
            Italics used to indicate unfunded study components.

                                                 1-  1

 Description of Study Component
Responsible Party
 Deterministic meteorological modeling (wind, turbulence and moisture
 fields) for each day of the 12 month period with domain and resolution as

 Apply a Lagrangian Monte Carlo transport model with some chemistry
 included to transport, disperse and chemically transform the plume using
 the output from the meteorological model.

 Full chemistry deterministic modeling  (RADM, with enhanced particle
 treatment, in-cloud processes and optics) for about 20 to 30 selected days.
 Expected input includes ammonia gas, paniculate ammonium ion, and
 hydrogen peroxide measurements at a few locations during intensives.
 Would employ the output from Pielke 's modeling.
Roger Pielke, CSU
                                                                                       William Wilson, RTP
                                                                                       Jason Ching, NOAA-
Continuous in-stack release of perfluorocarbon tracer during each of two
intensive periods.  Sampling at 31  sites on 12-hour (at 4 receptor sites)
and 24-hour (all other sites) sampling schedule (see attached map for
sampling locations).  The Hopi Point and Meadview sites will each have
two additional collocated samplers. A 21 day pre-release sampling pre-
test at all sites to establish background levels and for QA.

Additional perfluorocarbon tracers released to tag the Los Angeles Basin
and San Joaquin Valley. *

Increased time resolution of tracer data: 6 hours at receptor sites,  12
hours at other sites.
Russell Dietz
(overall), Brookhaven
Nat'l Lab and
Ray Dickson (tracer
release), NOAA Idaho

Dietz & Dickson
                                                                                       Dietz & Dickson
             Italics used to indicate unfunded study components.


Description of Study Component
Responsible Party
12 month period: Continuous vertical wind profiling over the 12 month
period at MPP plant site and Truxton using radar wind profilers.  A Radio
Acoustic Sounding System (RASS) will also be deployed at the plant site
to give boundary layer vertical temperature structure. A doppler sodar
will be deployed at  Meadview most of the study period to measure wind
profiles.  Additional instrumentation to include at least 4 surface
meteorology stations with pressure sensors (temperature & relative
humidity also) to examine response of locally channeled flow to larger
scale pressure gradients (November 1991-August 1992).

Surface  meteorological stations at the 4 receptor sites measuring  wind
speed, wind direction,  temperature, relative humidity and solar radiation.

Intensives:  Two additional  radar wind profilers will be operated during
the winter intensive near Needles and Temple Bar.  Two additional
profilers during the  summer intensive will be located at Meadview and
Cottonwood Cove.  The Meadview wind profiler will replace the sodar,
which will be moved to Temple Bar for the summer intensive. Radar
wind profiler data from Los Angeles Basin and western Mojave Desert
sites will also be available for the summer intensive. Surface
meteorological stations at radar wind profiler sites measuring winds,
temperature and  relative humidity.  Some special studies  using
tethersondes and/or  radiosondes may be done at locations of interest.
John Gaynor, NOAA
                                                                                      John Molenar, Air
                                                                                      Resource Specialists

                                                                                      John Gaynor
                                                  1- 3

Description of Study Component
Responsible Party
Air quality
 Particle Monitoring: Full IMPROVE samplers at the 4 receptor, 4
 IMPROVE sites and 2 Improve protocol sites.* IMPROVE channel A at
 all 21 remaining sites.  See attached map for site names and locations
 (total of 31 sites).  Drum sampling in 8 size ranges at the receptor and 2
 additional sites, with 4 or 6 hour resolution. Selected drum sampler filters
 to be analyzed.

 Intensives: two 12 hour samples per day, every day at receptor sites; 24
 hour samples every day at all other sites.  Tracer and SO2 sampling at all
 sites following the particle sampling schedule.  H2O2, NH4 and NH3
 monitoring periodically during  intensives.

 Hi vol dichotomous samplers and annular denuder samplers at 3 sites for
 high sensitivity particulate analysis necessary for CMB modeling.  More
 details will be available soon.

 Remainder of study year: 24 hour particle and SO2 sampling with
 IMPROVE samplers every Wednesday and Saturday at receptor,
 IMPROVE and IMPROVE protocol sites.

Increase number of particle monitoring sites for non-intensive periods.**
Bob Eldred, UC-Davis
      *      The IMPROVE sampler has 4 channels.  Channel A collects fine panicles (<2.5fjLm) on a teflon
     filter and provides total fine mass, elemental analysis (H and Na-Pb), organic and elemental carbon and
      absorption.  Channel B uses a fine nylon filter for ions (Cl~, NO.,', NO3' and SO42~).  Channel C is used to
      obtain organic and elemental carbon from a fine quartz filter.  Channel D measures PM-10 total mass on
      a teflon filter and SO2 with an impregnated quartz filter.

Description of Study Component
Responsible Party
Continuous monitoring for the entire period. Nephelometers at all receptor
sites and a transmissometer added at Meadview, in addition to ones
already at IMPROVE sites.

Airborne lidar aerosol mapping several weeks during the intensive
periods. *
John Molenar, Air
Resource Specialists
                                                                                      Jim McElroy, EMSL-
                                                                                      Las Vegas
Analysis of historic meteorologic data to optimize timing of intensive
periods. Analysis of MPP emission modulation (1985 shut-down).

Eigenvector Analysis

DMB modeling

CMB modeling with high  sensitivity particulate data.

Extinction Budget

Reconciliation of results from receptor & deterministic modeling and
eigenvector analysis, extinction budget, trajectory analyses, etc.  Overall
summary of results.
Mark Green, DRI
                                                                                      Mark Green
                                                                                      Bill Malm, NPS

                                                                                      Robert Stevens, RTF

                                                                                      Marc Pitchford,
                                                                                      EMSL- Las Vegas
                                                                                      Marc Pitchford
             Italics used to indicate unfunded study components.

             Italics used to indicate unfunded study components.


Description of Study Component
Responsible Party
Quality         Each component of the study is responsible for QA on its' portion of the
Assurance      study.

               Overall QA audit covering all portions of the study to be done by
               independent reviewer.
                                                                   Charles McDade
                                                1  -6

                             Appendix 2

                    Project MOHAVE1 Conceptual Plan

       This plan documents  the  thoughts  and intentions  of those who are
preparing to determine the contributions by the Mohave Power Project (MPP) to
haze in Grand Canyon  National Park  (GCNP).  Its purpose is to provide  a
vehicle to obtain review and comment by various interested parties at an  early
point in the planning process when adjustments are more easily accommodated.
This conceptual plan is designed  to provide overall guidance to the technical
experts who are responsible for developing the more detailed  study plan.

       The first part of this paper contains information on the study background,
objectives, and an overview of the approach.  This is followed, in the second
part, by an expanded discussion of the approach which contains information on
the visibility attribution process, use of artificial tracers, ambient monitoring, and
data interpretation and models.


       The 1991  fiscal year budget  for  the United  States  Environmental
Protection  Agency (EPA) includes a Congressional "add-on" at the level of $2.5
million for a 2-year effort  titled  "Pollution tracer study  at  the Mohave
Powerplant".  Discussion has revealed that congressional intent was to have EPA
perform a study to assess MPP's contribution to visibility impairment in GCNP.
Members of congress have demonstrated an interest in visibility impairment in
the Federal Class I Areas (i.e., national parks and wilderness areas meeting
certain requirements); and in particular an interest in GCNP impairment by large
point sources of SO2. For many this interest was intensified by the results  of the
1987 Winter Haze Intensive Tracer Experiment (WHITEX)  conducted by the
National Park Service (NPS).
         1  While  Mohave is  the  name of a coal-fired power plant  in
           Nevada,   Project  MOHAVE   contains  an   acronym   for
           Measurement Of Haze and Visual Effects.


      WHITEX involved a six-week long intensive monitoring study during
which an  artificial tracer was released from the Navajo  Generating  Station
(NGS)2. NFS analysis of optical,, air quality, and meteorological data indicated
that a significant fraction of the winter hazy periods in GCNP were due in large
part to sulfates  resulting from NGS emissions.  EPA used  these results as the
basis for proposing additional emission controls at NGS.  The  WHITEX data
analysis methodology, results, and use of the results were cause for considerable

      In  an attempt to resolve the  technical issues raised by  WHITEX, the
National Research Council of the National Academy of Sciences  (NAS) was
requested  to consider the relative importance of human derived and  natural
emissions  that contribute to visibility reduction.  The Council established  a
Committee on Haze in National Parks and Wilderness Areas.  One task of the
committee was to  evaluate WHITEX.  Their report neither wholly endorsed nor
discredited the  NPS WHITEX findings, though it did provide an illuminating
discussion of the technical issues. In an effort to avoid some of the controversy
of WHITEX and to take advantage of the expertise assembled by NAS,  Project
MOHAVE has requested the opportunity to discuss this conceptual plan with the
committee.  The  committee is scheduled  to be briefed  on  this effort in early
Spring 1991.

       Salt River Project (SRP), the operators of the NGS,  in an  attempt to
resolve  their doubts concerning WHITEX, supported a more extensive tracer
study in the winter of 1990. Though only preliminary results of this study are
now available, it appears to also indicate NGS emissions in GCNP during haze,
though at  a lower frequency of occurrence.

       It is the  goal of the planners of Project MOHAVE to take advantage of
the best and most successful aspects of the WHITEX and SRP studies,  and to
address the issues raised by the NAS WHITEX review to the maximum extent
possible, and to use and  extend information previously  obtained by numerous

       Previous air quality studies in the region containing the desert southwest
(including SCENES, VIEW, VISTA, WRAQ and RESOLVE) provide a great
deal of background information useful to the planning of this project,  Prevailing
           NGS is a 2250 Mw(e) coal-fired powerplant located near Page,
           Arizona, approximately 25 Km northeast of GCNP.

southwest winds, especially in the summer, carry MPP emissions toward GCNP.
They also  carry emissions from the southern California urban/industrial area
towards GCNP. There is considerable evidence that southern California is the
dominant source area of pollutant haze for GCNP.  A major technical challenge
for Project MOHAVE is to separate the influence of MPP from that of southern
California and other regional influences.

       The most important man-made pollutant species responsible for GCNP
haze are particulate sulfates.  These are generally formed in the atmosphere by
chemical conversion  of gaseous S02, which is emitted by combustion  of fuel
containing sulfur.   Other particulate components  important  to  GCNP haze,
organics and crustal species, are from natural and man-made  sources.  GCNP
visibility levels are often so good that light scattering by air molecules (Rayleigh
scattering) is also  a significant contributor to the extinction coefficient.

       MPP's  most  significant potential contribution to  GCNP haze  is by
emissions of SO2  that are converted to sulfates. Other sources contributing to
particulate sulfate  are southern California (primarily by oil refineries), other coal-
fired power plants (e.g., Reid Gardner north of Las Vegas, NV and  NGS near
Page, AZ), copper smelters in southern Arizona, New Mexico, northern Mexico
and Utah and oil refineries in Texas and the Monterrey area of Mexico.  Other
sources which may influence GCNP visibility are  large urban areas  (e.g.,  Las
Vegas, NV, Phoenix/Tucson, AZ and the Wasatch Front in Utah) and wildfires.
These sources are expected  to  be more dominated by organic and elemental
carbon pollutants  than sulfate.

       This conceptual plan considers two related objectives: (1) to determine the
 MPP contribution to GCNP haze and (2) to determine the relative contributions
 of the major pollution emission sources (including MPP) affecting GCNP haze.
 For both objectives,  determining  the  contribution  to  GCNP haze implies a
 quantitative evaluation of intensity, spatial extent, frequency, and duration.  The
 intensity of  haze contributed by a source  includes  both an absolute physical
 measure  of haze  (e.g.,. contribution  to  the extinction  coefficient)  and its
-perceptibility (e.g., scenic  element contrast change, or change in modulation
 transfer function).  A  part of both objectives is an assessment of the changes in
 visibility at GCNP that would be expected if MPP emissions were changed.

       The first objective implies determining the contributions to GCNP haze
by two source categories: MPP and a composite of all non-MPP sources.  The
second objective expands upon the first objective.  Instead of concentrating on
one source's impact, it calls for simultaneous assessment of all the important
sources of haze for GCNP.  There is no doubt that a study designed to meet the
first objective would also address other sources to some extent. However, this
would be incidental to the first objective, unlike the second objective where it is
the primary focus.

  A program designed to meet the second  objective is beyond the resources
presently available for this effort.  Unless additional support becomes available
Project MOHAVE will be designed to meet the first  objective and to prepare a
foundation for further investigation of the impact of regional haze in GCNP.
Approach Overview

    1   The EPA  Office of Air Quality Planning and Standards (OAQPS) has
overall management responsibility for Project MOHAVE.  Robert Bauman, the
OAQPS Project Leader, has selected a Project Steering Committee to advise him
on the overall direction of the study.  Several technical advisory panels are being
constituted to  provide  recommendations at a greater level of detail.  Experts
selected for the technical advisory panels will  provide the primary means for
Project MOHAVE to incorporate insights gleaned from earlier investigations.
Figure 1 indicates the program's management/advisory structure.

       Project  MOHAVE  will  use sophisticated deterministic and  receptor
modeling  to identify the MPP influence. During two intensive study periods
(four to six weeks each), a unique tracer material will be Combined with the MPP
emissions in concentrations sufficient to be detected hundreds of kilometers away.
The tracer will provide a check of the deterministic modeling results and provides
a unique signature for the MPP plume, for use in receptor  modeling.  The main
emphasis  will be  on deterministic modeling, with secondary emphasis given to
receptor modeling.   It is not prudent  to implicitly  trust  the results  of either
modeling  approach alone; thus both approaches will be tried.  If results of the
two approaches are  substantially different, an in-depth investigation into the
reasons for the differences and an evaluation of the results will be done before"
any conclusions are reached regarding MPP's impact.

       The  intensive  periods will  be  selected  to  optimize  the chances  of
establishing  the  maximum  contribution  to  GCNP  impairment by MPP.


           PROJECT  MOHAVE
                         PROJECT MANAGER
                           R. Bauman, EPA
                          STEERING COMMITTEE

Responsible for technical d

esign and study oversight.

 Tracer selection
 Release mechanisms
 In-stack monitoring
 Release protocol
- Equipment selection
- Site selection
- Ambient monitoring
- Optical monitoring
- Meteorology
- Data processing
- Filter analysis
- Dispersion modeling
- Trajectory analysis
- Receptor modeling
- Extinction budget
- Attribution analysis
- Perception & effects

Tentatively these would be the summer monsoon season and the mid-winter storm
season.  Both periods have the possibility  of transporting MPP emissions to
GCNP decoupled from southern California emissions, and sufficient moisture for
possible liquid phase conversion of S02 to sulfate (much faster conversion than
the alternative gas  phase reactions).  These  conditions are  intermittent even
during the periods of their greatest frequency.  Thus  the intensive periods will
also include more typical summer and winter conditions where MPP influence in
GCNP is not expected to be as great.

       To further ensure that data from .the intensive periods can be interpreted
in terms of longer term typical conditions, the overall  study period will be 12 to
15  months.  During the non-intensive periods of the study, air quality and
meteorology measurements will be made at numerous locations throughout the
study area.  Intensive study period data will be used to evaluate source-oriented
deterministic models using augmented upper air meteorological data and receptor
models based upon endemic tracers.   The deterministic models will then be
applied to the entire study period.  If successful, receptor models using endemic
tracers will be also" be applied using data collected for the entire study period.
Finally,  the study  period  results  will be  extrapolated to the long-term by
comparison with and if necessary adjustment to climatological characteristics of

       A tentative schedule for Project MOHAVE calls for field measurements
to start in July 1991 and continue until  September of 1992. The winter intensive
period will be in January 1992, with the summer intensive period from mid-July
to late August 1992. Data interpretation and report preparation is anticipated to
continue for approximately 18 months after  the  end of the  field monitoring

       The attribution of impacts from MPP and other sources will ultimately be
derived from an extinction budget by air pollutant species.  The majority of the
MPP impact is expected to be from secondary sulfate particles.  Measurements
of the particle components such as sulfate, nitrates, carbon and crustal species .are
related to optical measurements  by statistical and first principle  approaches to
produce the extinction budget.

       It is expected that the contribution of sulfate particles from MPP and other
sources will be  estimated primarily from  deterministic  modeling.  Receptor


modeling will also be  done  for  this purpose,  providing a  check of  the
deterministic modeling analyses. The results for the two types of models will be
compared for consistency (model reconciliation). Eigenvector analysis will also
be done to support results from the modeling studies.  During the two intensive
periods,  an artificial tracer will be injected into the MPP plume.   The tracer
provides a check of the transport and dispersion calculated by the deterministic
modeling.   It also  provides a unique signature of the MPP plume for  use in
receptor modeling.  To estimate impacts for the remainder of the study period,
deterministic modeling will be performed and receptor modeling using endemic
tracers will be investigated.

       Substantial monitoring will be required to  support  the extinction budget
and modeling studies. This will include ground based and airborne meteorologic,
air quality and optical measurements and remote sensing  of vertical wind and
temperature profiles. Expanded descriptions of the main components of the study

       During the  intensive study periods an artificial tracer will be released
continuously either through the stack at MPP or by balloon  at plume height in the
immediate  vicinity of the power plant.   A stack release would  give  more
confidence that the plume and tracer are well mixed and is the preferred method.
However balloon release of tracers has been routinely done (NOAA, Idaho Falls)
and is a feasible alternative. For objective 2, different artificial tracers would be
released at  other sources or source areas to tag their  emissions more precisely
than through the use of endemic tracers.  Other sources to tag may include the
San Joaquin Valley (Tehachapi Pass), the Los Angeles Basin (Cajon Pass), Las
Vegas, Reid Gardner Powerplant, Navajo Powerplant and copper smelters.

       Tracer can be released at a constant emission rate  or at a constant ratio
of tracer to SO2. Variation of tracer to S02 ratios was a complicating factor in
the WHITEX receptor modeling analysis.  If released at a constant rate, S02
emission rate variations would complicate the receptor modeling, requiring
adjustment  of the ratio of tracer to  sulfur dioxide concentration. This requires
knowledge  of plume age. However, for use in deterministic modeling, it is more
desirable to have a constant  tracer emission rate, to simplify the dispersion
calculations. Also, the deterministic model can give the plume age necessary to
adjust the tracer to sulfur dioxide emission rates in the receptor modeling.

       Ideally, a tracer should closely mimic the species of interest for receptor
modeling and chemical transformations; in this instance SO2 and its conversion


to SO4 and deposition of the sulfate particles.  This would suggest using isotopes
of sulfur or oxygen.  However, the large amounts of tracer necessary may not
be available and to produce them would require  resources greater than those
available  for this  study.  For studying transport  and dispersion  patterns,  a
conservative tracer is desirable.

       Among  the potential tracer  materials are  deuterated methane (CD4),
various perflorocarbons (PFT's) and particulate rare  earth oxides.   CD4 and
PFT's are conservative tracers; thus conversion of SO2 to S04 and deposition of
S02 and S04 must be accounted for.  It has been suggested that non-conservative
rare earth particle tracers be used because of their potential to mimic sulfate
particles.  However,  sulfate particles are not directly emitted in significant
quantities; rather they are typically  formed after  considerable transport time
which varies with meteorologic conditions. Thus some variable proportion of the
rare  earth particles  will  have  deposited before  the  sulfates are  formed.
Additionally the deposition of SO2 occurs more rapidly than either sulfate or rare
earth particles. Issues such as  these must be further investigated before any
decision regarding the use of rare earth tracers is made.

       CD4  has  low background  values  and  is  detectable at  very  low
concentrations, so  small amounts  of this tracer are sufficient.  Though the cost
per unit mass is high,  the total cost of tracer material is expected to be much less
than the cost of PFT's.  However, the sample analysis cost is very high ($800-
$1000/sample), compared to about $20/sample for PFT's.  Thus, it may not be
feasible to analyze all samples. If CD4 were used, samples would be selected for
analysis based on air quality and meteorologic data.

       The lower analysis costs for PFT's makes it possible to analyze many, if
not all of the samples. More information canlbe obtained regarding the plume
position and spatial extent. This would allow a more thorough evaluation of the
deterministic modeling. In addition, regression analyses with the receptor models
and other statistical analyses would be based on a larger number of samples than
if CD4 were used. With the availability of various PFTs,  release times can be
staggered such that the age of the samples can be estimated from the  sample  as
well as from trajectory analyses.  Alternately, different PFTs could be released
from different sources,  as previously discussed  and the deterministic model
results used to estimate plume age.

       The SRP tracer study, which used PFT's, apparently had some major
problems with the  tracer portion of the study. Collocated samplers showed near
zero correlation. Four different PFT's were used. The analyses for the first two


PFT's were apparently of better quality than for the third and fourth. The South
Coast Air  Quality Study (SCAQS) is said to have  shown high variability of
collocated  samples near  the detection limits while at the higher concentrations
variations of a factor of two were common. There is no theoretical reason that
prohibits the use of PFT's  or  other  tracer materials to give quantitative,
consistent data.  However the  pitfalls  associated  with past experiments demand
careful attention'and a quality assurance program that  monitors the tracer data
during the  collection process.  These issues must be resolved before selection of
a tracer approach.  A quality  assurance plan for tracer release and monitoring,
including collection/analysis  of background and  collocated samples  will  be

       Project MOHAVE field measurements are designed to meet the data
requirements discussed in the Data Interpretation and Modeling Section, below.
The extinction  budget analysis  requires  data for all of the major  particle
components (e.g.,  sulfates, organic and elemental carbon,  crustal, and  liquid
water as estimated from relative humidity) by particle size and concurrent optical
parameters (e.g., extinction and scattering coefficient). The attribution analysis
requires data  for tracer, particle and gaseous  sulfur concentrations, particulate
trace elements as endemic tracers (e.g., arsenic for smelters and selenium for
coal burning), and meteorology (e.g., surface and upper air winds, temperature,
and  humidity).     Additional  monitoring  of  endemic   tracers   (e.g.,
methylchloroform  for southern  California)  for  non-MPP sources will  be
conducted to the extent that the resources  will allow.  Table 1  summarizes the
measurements that  are anticipated for this program.

       To aid in the presentation, monitoring locations Have been categorized into
several  types.   Receptor  sites are in  or  near (representative  of)  GCNP.
Monitoring at receptor sites must be capable of supporting extinction budget and
attribution analysis.  Gradient  sites are designed to produce data for attribution
analysis. They include sites between sources of interest and GCNP, upwind and
background monitoring locations.  Upper air meteorological monitoring locations
are selected to improve the spatial resolution of the National Weather Service
network and to provide vertical wind and temperature profiles in critical areas for
input to the deterministic models.  Finally aircraft are needed to make tracer and
pollutant measurements ranging from near the  source to the most distant areas of
the study region, and to evaluate vertical distributions.

Table 1.  List of the optical variables, aerosol species, meteorological variables and
         measurement methodologies proposed for the monitoring sites.
Measurement Type
b«- '
Particulate Matter
Fine Particles
Elemental & Organic Carbon
Trace Elements (includes sulfur)
Size Segregating Trace Elements
Large Particles
Meteorological - Surface
Wind Speed & Direction
Temperature, Relative Humidity
Meteorological - Upper Air
Wind Speed & Direction
Cloud Heigfit and Vertical Pollutant


A, B, C
A, B, C

A, B, C
A, B, CT
A, B, C

A, B

Pierce's Ferry
Pierce's Ferry
Pierce's Ferry
- 	 - 	 "1 1
K2CO3 Impregnated Filter


i I
RADAR Profiler
Ceilometer Or Upward
Looking LIDAR


12 Hours
12 Hours
12 Hours
12 Hours
1 12 Hours
1 12 Hours
12 Hours
12 Hours
12 Hours
12 Hours

A = Receptor sites, B = Gradient sites, C = Aircraft, " = method and frequency may be different for

       The selection of monitoring locations has an influence on the utility of the
data.  Table 2 and figure 2 indicate a preliminary list of monitoring locations
appropriate  for meeting  the first .objective.  An expanded investigation of the
impact of all sources of visibility impairment would require additional gradient
sites and perhaps additional artificial and endemic tracer measurement capabilities
at all  sites.   Final site  selection by  the  appropriate advisory panel  will be
influenced by  results  of simple  trajectory analyses run on two  years  of data
(anticipated in March 1991), and practical considerations (i.e., available power,
access, security, etc.).

       To the maximum degree possible,  existing monitoring sites within the
study area will be incorporated into the monitoring program.  In some cases this
would involve providing supplemental equipment or modifying procedures to
make data collected at these sites consistent with the other sites in the program.
Meteorology data from existing sources (i.e., National Weather Service surface,
upper air, and satellite measurements)  will be incorporated into the project data
base.  To the extent that they  exist, records of wildfires and prescribed burning
and other intermittent source activities will be documented.

Data  Interpretation and Modeling

Extinction Budget:
       Light extinction is  caused by scattering and absorption by particles and
gases.  In general particle scattering is the principal component of extinction,
though in the remote Southwest, scattering by gases that make up the atmosphere
(also known as Rayleigh  scattering) is a significant fraction on the best air quality
days.  Black carbon (from diesel engines, forest fires, etc.) is the primary agent
of particle absorption, and  is occasionally an important cause of haze in the study
area.   NO2 is the only common gaseous pollutant that absorbs in the visible
portion of the spectrum.  It is not expected to play a significant role in Project

       The extinction budget analysis involves determining the contribution to
extinction by  all of the major  contributing components.  This  can be  done
statistically using multivariate analysis to explain  the optical parameter (bext or
bscat)  by a linear  combination of the components.  These components are the
concentrations  of the  pollutant  species multiplied by best-fit  determined
coefficients  interpreted  as extinction  efficiencies.   The  hygroscopic particle
species (e.g., sulfate and nitrate) include a function  of relative humidity to
incorporate the effects of water upon the extinction efficiencies of these species.
Alternately, first principle calculations  (Mie Theory) of the extinction coefficients


Table 2.  Possible monitoring sites for intensive and entire study periods by site type.
Site Types
Upper Air
Entire Study
Pierce's Ferry, Meadview*,
Hopi Point*, Indian Gardens*,
Phantom Ranch*, Long Mesa*
W. Lake Mead, Cotton wood
Cove*, Spirit Mtn.*, Overton,
Needles, Mojave Desert
Laughlin, Pierce's Ferry

Intensive Only
Peach Springs
Additional sites along Colorado
River, southern California, and
northern Arizona
Peach Springs, Mojave Desert
Near stack, upwind, along plume,
'across plume, vertical distribution
* existing NFS and SCE air quality monitoring sites

•    Existing  Moniloring Sites

«    Polculiiil Monilofing Silos

•    Points of Reference
   Teh«ch«pi P»s>
              Cajon l>OSj
      Los Angeles
                                                                piril Mountain
                                                          x     \
                              Euslcrn Mojavc Dosed       MPP  • I
                                  Silci (1-3)                   |
                                       a                       "
Ccnlril Arizmit
  Silci (1-3)
                                                                      ,o P«rkcr Dam
                   Tigure  2.     Existing  and   potential  -monitoring  sites.

can be  done, if sufficient  paniculate characteristics  are known  (e.g., size
distributions).  This program will use both procedures and will  reconcile the
results with literature values of extinction efficiencies.
Attribution analysis:
       Attribution analyses will be done using both source-oriented deterministic
models and receptor models.  Source-oriented deterministic models  transport
emissions from  the source and  can  account for physical processes en route,
including chemical transformation, dispersion and deposition. Receptor models
use  measurements  made  at  the area  of concern  (receptors)  along  with
characterization of the emissions  from sources potentially affecting the  receptor.
The contribution of each source to concentrations at the receptors is determined
statistically through multivariate analysis techniques which link the sources to the
measured concentrations.

       Deterministic meteorological and dispersion modeling provides  a source-
receptor pollution apportionment  procedure which is based  on  fundamental
physical conservation relationships.   These relationships include conservation
equations for velocity,  temperature, mass and the three phases of water.  The
model will provide detailed wind and turbulence fields  and a prediction of cloud
height  and  location.   Cloud  predictions  will be checked  against satellite
photographs and ceilometer measurements, where available.

       For Project MOHAVE, deterministic models will be run in an analysis
mode using assimilation of the observed data for the entire period of the project.
Data assimilation means  that  measured  data  will be incorporated  into the
modeling.  The models' utility is to fill in areas between  data locations making
use  of the fundamental physical relationships governing  the atmosphere.  The
incorporation of data assimilation into the deterministic model offers an effective
methodology to  achieve  the  best estimate of meteorological transport  and

       The meteorological domain for the simulations will cover the southwestern
United States with horizontal grid intervals on the order, of  10 km.  To obtain
better terrain resolution near MPP, a telescoping nested grid will be used. In a
nested, grid approach,  the larger scale results provide the boundary conditions
for input into a finer scale modeling domain. In Figure 2, the rectangle bounded
by dashed lines  "nested"  within the  larger area demonstrates the  concept of a
nested grid approach.  Horizontal  grid intervals in the smallest domain may be
                                 2- 10

approximately 500 m.  For this smaller grid interval, non-hydrostatic models are
generally more appropriate than hydrostatic models.

       The. wind  and  turbulence  fields obtained  from  the  deterministic
meteorological model provide the necessary input to calculate the transport and
dispersion of the MPP plume.  Using this input, a Lagrangian model will be used
to transport, disperse and chemically transform the plume. The first step is to
transport and disperse the plume; the model results will be compared to the tracer
data to evaluate the model.  The next step is to incorporate simple chemistry to
calculate sulfate concentrations and any other species of interest. The predicted
location of emissions from other major sources within the study area will also be

       Using complex chemical modeling (e.g. RADM) and explicit inclusion of
all the major .pollutant sources is very resource intensive and is beyond the scope
of objective 1.  These analyses may be done as part of objective 2, depending on
the level of additional  resources.

       The use of a deterministic model can also assist in the design of the field
program by indicating where instrumentation should be sited and aircraft cross-
sections  flown  so as  to  optimize  the spatial  representativeness  of the
measurements.   Also,  since  the model is based on  fundamental concepts, it
provides a scientific framework to interpret the data.  The same meteorological
model simulations can also be used with a wide range of emission inventories in
order to assess potential emission control  scenarios.

       The use of receptor models  in apportioning primary particles has been
done routinely; however using receptor models to apportion secondary  aerosol,
as in WHITEX,  is more controversial.  As in WHITEX, the receptor models
used will include the tracer mass balance regression (TMBR)  and differential
mass balance  (DMB)  models.  These models were  designed  to estimate the
portion of sulfate due to the MPP  and the other  sources.  The results  of these
receptor modeling will be evaluated in light of the deterministic modeling results.
Project MOHAVE will address concerns raised by WHITEX review by making
additional  measurements and more complete source characterizations.  More
information on particle size distribution, use of endemic or artificial tracers for
other sources, and upper air humidity and cloud height measurements can reduce
the uncertainties involved with the use of these  models to apportion secondary
aerosol.   In TMBR and DMB  it is assumed that each source has a uniquely
emitted tracer  associated with it.  If not,  other methods such as chemical mass
balance  (CMB)  may  be first  applied  to partition  the ambient  species


concentrations into  components attributable to the various groups of sources.
TMBR, DMB and CMB are described in detail in the WHITEX report.

       Eigenvector analysis, e.g. empirical orthogonal function (EOF) analysis,
principal components analysis and factor analysis, will also be done to investigate
impacts by specific sources or source areas. The eigenvector analysis results can
be used to qualitatively check the deterministic and receptor modeling analyses.
Eigenvector  analysis  shows  commonly occurring  spatial patterns and  their
variation in time. The main patterns may be associated with specific sources or
source areas. By examining the time series (time variation)  of each eigenvector,
it  can be determined which times  a particular source  area contributes to
concentrations at each site. Meteorologic information, such as wind speed and
direction and humidity and its temporal patterns along with source information
provides physical information to  help  interpret and support results of the
eigenvector analyses.
       Thp  Modeling  and Data Analysis technical panel will make specific
recommendations concerning the modeling and data analysis approaches  to be
used.    !

Extrapolation of intensives study periods to the long-term:
       To determine longer term impacts to visibility at GCNP, it is necessary
to extrapolate from results of the intensive study periods.  This will be a two-step
process; the first step will  relate the entire 12-15 month study period  to  the
intensive period, while the second will extrapolate from the 12-15 month period
to a multi-year period.  The first step involves application of source-oriented and
receptor models, which are developed and evaluated with intensive period data,
to the meteorology and air quality data for the entire study period. In the second
step |the relative frequency of long-term meteorological patterns will be compared
with those during the study period and qualitative adjustments made if necessary.

       Source-oriented models  will be evaluated and calibrated using the more
complete data of the intensive periods. The resulting models will then be run on
data from the entire study period. For all modeling analyses a portion of the data
may be withheld in order to independently test the models.

       During  the intensive study periods,  receptor  modeling will use  the
artificial tracer results to  apportion sulfate due to MPP and any other sources
tagged with artificial tracers. Receptor models will also use endemic tracers to
apportion remaining significant sources.  Results from receptor models based on


endemic tracers will be compared to results of the same models using artificial
tracers to evaluate the utility of endemic tracers.  If successful, endemic tracer
models will then be applied to the entire study period to apportion sources over
a complete annual cycle and  used in conjunction with the deterministic modeling

       The representativeness of the study year to longer term average conditions
will be studied. It should be acknowledged that significant year to year variability
in meteorological conditions occurs and that the likelihood of any given year
being "typical" is not high. The frequency of occurrence of conditions associated
with impacts from each source such as wind speed and direction, humidity, etc.
can be compared for the study year and other years for which data are available.
Where they exist, optical and air quality measurements from previous years will
be  compared to the  study year measurements.  A meteorological classification
scheme that uses criteria affecting visibility may be developed. The frequency
of occurrence of each pattern for the study period and longer term average can
then be compared to put the study year into perspective.
 Framework for Summarizing Results:
       In a complex program such as this, a sound plan for compilation of results
 is as important as the collection of high quality and representative data and the
 performance of appropriate interpretive analysis.  Development of an approach
 to organize the results from this program helps to focus attention and resources
 on critical steps for the entire program and communicate those ideas to others.
       Just as it  is inappropriate  for worst  case  results  to  receive primary
 attention,  it is also inappropriate to dwell on average or  typical conditions,
 especially for an instantaneous 'effect such as visibility. The 12-15 month study
 period  with hourly  deterministic  model  results  requires  some method for
 summarizing the results of the study that avoids these pitfalls.  A preliminary
 conceptual framework for summarizing the results of Project MOHAVE is shown
 in Table  3.  The  key idea  is the stratification of time periods based upon the
 locations  with respect to GCNP of MPP emissions and those of other significant
 sources,  such as  from  southern California.   These would  be based upon the
 modeling studies.  Another stratification is whether the plume(s) has undergone
 wet or dry chemistry (based upon modeling results  and observations).  If useful,
 other stratifications could be developed. The frequency of  each  condition, the
 average and standard deviation of the % sulfate from MPP,  the % of extinction

Table 3 - Conceptual Framework for Summarizing Project MOHAVE Results
GCNP Impact &
MPP & Other
Sources Dry
MPP & Other
Sources Wet
MPP Alone Dry
MPP Alone Wet
Other Appropriate

% Sulfate

% Extinction

Measure of

   SCA refers to the urban and industrial areas of southern California.

from MPP and a measure of the perceptibility of the MPP impact is estimated for
the study period.

       Stratification of conditions is expected to not only aid in summarization,
but to  reduce the uncertainty levels for some  of the receptor model results by
restricting the variation of parameters assumed to be constant (e.g, chemical
conversion rate). Extrapolation to a long-term average may be done through the
use of a meteorological classification scheme as previously described.  This type
of  approach  provides  an efficient  manner of  presenting the  magnitude and
frequency of estimated MPP emissions on GCNP  over a long-term period that
could be used to evaluate the significance of existing impairment.

                                    Appendix 3

                                 PARTICIPANT LIST

                        Project MOHAVE Planning Workshop

                               April 29 - May 2,  1991
                                  Denver, Colorado
Robert Bauman
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
(919) 541-5629

Donald Blumenthal
Sonoma Technology
5510 Skylane  Drive, Suite 101
Santa Rosa, CA 95403
(707) 527-9372

Jason Ching
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
(919) 541-4801

Russell Dietz
Brookhaven National Lab
Building 426
Upton,  NY  11973
(516) 282-3059
FAX (516) 282-2887

Ray Dickson
1750 Foote Drive
Idaho Falls, ID  83402
(208) 526-2328

David L. Dietrich
Air Resource Specialists, Inc.
1901 Sharp Point Drive, Suite E
Fort Collins, CO 80525
(303) 484-7941

Robert Eldred
Crocker Nuclear Laboratory
University of California at Davis
Davis, CA  95616
(916) 752-1120
FAX (916) 752-1124
Rob Farber
Southern California Edison Co.
2244 Walnut Grove Road
Rosemead, CA  91770
(818) 302-9693

John Gaynor
325 Broadway R/E WP7
Boulder, CO 80303
(303) 497-6436

Mark Green
Desert Research Institute (EMSL-LV)
P.O. Box 93478
Las Vegas, NV  89193-3478
(702) 798-2182
FAX (702) 798-2692

Ronald  Henry
Civil &  Environmental Engineering
University of Southern California
KAP 224E
3620 S. Vermont  Avenue
Los Angeles, CA  90089-2231
(213) 740-0596

Thomas Hoffer
Desert Research Institute
P.O. Box 60220
Reno, NV  60220
(702) 677-3193

Hari Iyer
Colorado State University
Foothills Campus
Fort Collins, CO 80523

Jonathan Kahl
Department of Geosciences
University of Wisconsin-Milwaukee
P.O. Box 413
Milwaukee, Wl  53201
(414) 229-4561

Darko Koracin
Desert Research Institute
P.O. Box 60220
Reno, NV  60220
(702) 677-3193

Doug Latimer
Latimer & Associates
2769 Iris Avenue, Suite 117
Boulder, CO 80304
(303) 440-3332

William Malm
National Park Service
Colorado State University
Fort Collins, CO  80523
(303) 491-8292
FAX (303) 491-8598

Stan Marsh
Southern California Edison Co.
2244 Walnut Grove Road
Rosemead,  CA  91770
(818) 302-9711

Sharon McCarthy
Sigma Research  Group
234 Littleton Road, Suite 2E
Westford, (MA 01886
(508) 692-0330

Charles McDade
ENSR Consulting & Engineering
1220 Avenda Acaso
Carmarillo, CA  93012
(805) 388-3775
FAX (805) 388-3577

Janet Metsa
JCM Environmental
5 Pine Circle
Houghton, Ml 49931
(906) 482-5665
Vince Mirabella
Southern California Edison Co.
2244 Walnut Grove Road
Rosemead, CA  91770
(818) 302-9748
John Molenar
Air Research Specialists
1901 Sharp Point Dr., Suite E
Fort Collins, CO  80525
(303) 484-7941

Gene Mroz
Los Alamos National Laboratory
P.O.  Box 1663,  MSJ-514
Los Alamos, NM 87545
(505) 667-7758
FAX  (505) 665-5688

Peter Mueller
P.O.  Box 10412
Palo Alto, CA 94303
(415) 855-2586

William Neff
325 Broadway R/E WP7
Boulder, CO  80303
(303) 497-6265

John Ondov
Department of Chemistry
University of Maryland
College Park, MD 20742
(301) 405-1859
FAX  (301) 314-9121

Roger Pielke
Colorado State University
Department of Atmospheric Science
Foothills Campus
Fort Collins, CO  80523
(303) 491-8293

Marc Pitchford
P.O.  Box 93478
Las Vegas, NV  89193-3478
(702) 798-2363
FAX  (702) 798-2692

Bruce Polkowsky
U.S. Environmental Protection Agency
Research Triangle Park, NC  27711
(919) 541-5532
 Pradeep Saxena
 P.O. Box 10412
 Palo Alto, CA  94303
 (415) 855-2591

 Nelson Seaman
 Pennsylvania State University
 503 Walker Building
 University Park, PA  16802
 (814) 863-1583

 Chris Shaver
 National  Park Service
 Air Quality Division
 P.O. Box 25287
 Denver, CO 80225
 (303) 969-2075

 Jim Sisler
 Colorado State University
 CIRA Foothills Campus
 Fort Collins, CO  80523
 (303) 491-8406

 Jim Southerland
 U.S. Environmental Protection Agency
' Research Triangle Park, NC  27711
 (919) 541-5523

 Gene Start
 1750 Foote Drive
 Idaho Falls, ID 83402
 (208) 526-2328

 Bob Stevens
 U.S. Environmental Protection Agency
 AREAL (MD-47)
 Research Triangle Park, NC  27711
 (919) 541-3156
Ivar Tombach
Aero Vironment
222 E. Huntington Drive
P.O. Box 5013
Monrovia, CA  91017
(818)  357-9983

John Vimont
National Park Service
Air Quality Division
P.O. Box 25287
Denver, CO 80225
(303)  969-2077

                                Appendix 4



1)    Concentration of S02 at long distances from source is approximately:

                                SO  -    Q
Q = SO2 source strength = 150 tons per day =  1.58 kg s"1  = 1.58 X 109 ug s"1
u = average wind speed in the mixed layer
h = depth of mixed layer
D = distance from source
8 = lateral plume dispersion in degrees
D tan 6 = width of plume at distance D
6 typically 5-15°
distance to  GCNP is 120 km
Thus plume width at 120 km = 21 km for 6 =  10°, 32 km  for 6  = 15°
sulfate is (NH4)2S04

2)    Determine incremental sulfate concentration that is noticeable:

Assume change in bext of 10% is noticeable
Scattering efficiency of (NH4)2S04 is 5 m2g~l

For Rayleigh conditions, bext = 11 Mm"1,  noticeable change is:
For average conditions, bext = 25 Mm"1, noticeable change is:

                                       = 0.50Mm
                                             ' °
                           2-5 XlO""-'            «

CASE 1: prefrontal winter conditions, cloudy

u = 20 m s"1   (conservatively high)
h = 3000 m
6 = 10°
                       (2 x 101 m .r J) (3 x 1 03 m) (2. 1 x 1 04 m)
                               -3 S02

                     =2.6/:m-3 (NH4)2SO4 at 100% conversion

                     = 1.3/im'3 (NH4)2S04 at 50% conversion

CASE 2: Typical summer afternoon conditions,  a) cloudy,  b)!dry

u = 6 m s"1 August average over 18 years at China Lake, 10000 feet MSL
h = 4000 m
6 = 15°

               = 2.1/igm'3 S02 =4.2fjigm~3 sulfate at 100% conversion

              a) cloudy:if - S02 contacts cloud,  sulfate =2.

              b) dry-.assume 3.5%/zr"1 conversion
             transport time=	120A77Z
                                    3 g hnhr'1
                          =5.6 hours =19%  conversion =0.

CASE 3: weak pre-frontal winter conditions, cloudy

u =  6 ms"1
h =  1500 m
6 =  15°
                                                        cc    _3
                  2	  = 5.5 agm 3

    sulfate = ].l.3fjigm'3 with 100% conversion = 5.7/igm~3 with 50% conversion
 For the conditions considered, the plume would range from marginally noticeable to quite
 noticeable  using this  very simple  methodology. The results  indicate  that  further
 consideration is justified; the potential for impact cannot be dismissed without additional


      The difference between average particulate sulfur concentrations at Spirit Mountain and
Meadview were compared for outage and non-outage conditions. The gradients were compared
for all wind directions and for wind directions transporting the plume toward the site. Data is
from Murray, et al, 1989.

                               All wind directions

Site               MPP status         Sjxgnr3           n

Meadview          off                0.363             36
                  on                0.383             454

Spirit Mtn          off                0.410             54
                  on                0.385             442

                     Wind direction < ±90° from MPP to site

Site               MPP status         SuenT3           n
Meadview         off                0.363             36
                  on                0.379             358

Spirit Mtn         off                0.508             25
                  on                0.439             242

For all wind directions :    Spirit^ - Meadview^ = 0.002 ^ignr3
                        Spiritoff - Meadviewoff = 0.047 /igm'3

Average difference in gradient= 0.045 /ignr3

For wind direction < ±90° :Spiritm -  Meadview^ = 0.060 /ignr3
                        Spiritoff -  Meadviewoff = 0.145 jignr3

Average difference in gradient= 0.085 ^ignv3

It can be seen  that the average gradient in particulate sulfur between  Spirit Mountain and
Meadview is greater when MPP is not operating, particularly for wind directions favorable for
transport  from  MPP.  It is hypothesized that the gradient is small  when MPP is operating
because increased  dilution of the southern California sulfur  between  Spirit Mountain and
Meadview is balanced by formation of particulate sulfur in the MPP plume.  During outage

conditions, this does not happen; thus the gradient between Spirit Mountain and Mead view is
       The difference in gradient of 0.085 pgrn'3 particulate sulfur corresponds to about
0.34 figm"3 sulfate as ammonium bisulfate.  Assuming a mass scattering efficiency of 5 m2g'1,
this would add an average of 1.7 Mm"1 to the extinction  coefficient.  This increase would be
expected to be marginally perceptible for very clear days (bsp < 17 Mm"1) and imperceptible for
other days.  However, these estimates are  for  concentrations  averaged many days, while
visibility is likely to vary significantly over the course of a day, and between days.
It should be emphasized that the data base used is very limited and no conclusions can be made
regarding the impact of MPP using this limited data set.  However, the difference in gradients
when MPP is off compared to on suggests the hypothesis put forth above may be correct.

                                  Appendix 5
                                                                     August 26, 1991

                                   Prepard by:

                  Desert and Intermountain  Air Transport Program
The  appropriation by  Congress of 2.5 million  dollars to EPA  to  conduct  a source
apportionment study on  the  Mohave  Power  Project (MPP) has generated widespread
interest in the analysis  of -the data obtained during a seven month period the  MPP was
inoperative in 1985. This outage period represents the ultimate in experiments.  The plant
was turned off and the  effects can be examined.  It serves as a baseline for assessing the
impact of the MPP on visibility degradation in the Grand Canyon National Park (GCNP).
The data base can also be used in other statistical analyses that utilize the fluctuating power
plant load as a sort parameter.

The initial study was published by Murray et al. (1989) which showed  that the sulfate
concentration at Meadview was not significantly lower than that observed during similar
periods in other years.  The study set an upper bound to the sulfate due to MPP observed
at Meadview as less than 15%.  The power of this analysis was limited by its rudimentary
treatment  of inter-annual meteorological differences. The paper's impact is limited by the
fact that the authors did not consider the daily power plant load during the control periods.

The SCENES data base was utilized in the analysis of the  outage period with  respect to
similar periods in other years. That program was designed and implemented to acquire high
quality data for studies  of visibility degradation.

The  outage study used  the  24-hour  average particulate  samples at three  sites, one
background and two receptor, with respect to impact from  MPP.  Chemical and physical
analyses of the filters were carried out only on every third day.  The samples for the two
intermediate days were archived.

A re-examination of the outage and other periods of reduced power plant output  compared
to periods when the plant operates at or near capacity during the SCENES program is
envisioned. The new study will incorporate the data used in  the original analysis  but would
also embody the following elements:

       Independent statistical analysis of the experiment

       Chemical analysis  of all the filters. (Quality assurance will be evaluated through
       comparison of  current  results  to past data through regression  and time  series

      Classification of the synoptic weather patterns  affecting transport from MPP to

      Deterministic  modelling of the wind  flow patterns associated with each of the
      meteorological regimes.

      A detailed compilation of regional SO2 emissions data for the  control and outage
      periods. (Changes  in emission patterns must be included  in the  final analysis.)

These elements will be described in the following sections.

Statistical Analyses-Dr. Paul Switzer of Stanford University will serve  as the independent
statistician. He has a history of involvement in physical measurement processes. He will:

      Be responsible for  the overall experiment design after consultation with the principal
      scientists  (Hoffer,  White and Koracin), the  participants  in the original study and
      familiarization with the existing data base. (The written experiment design document
      would be a cooperative effort.)

      Be responsible for the specification of the techniques used to handle the data and
      the statistical tests that will be applied. All data manipulations can be performed in
      the "blind". This  procedure has been used  hi the past within  the  meteorological
      community to evaluate the  results  of weather modification experiments and has
      proven effective hi eliminating cries of bias and data selection.

      Be responsible for sample handling procedures (if the samples are assigned random
      numbers), data stratification, application of statistical tests and reporting of the

      Participate in the redesign of the meteorological classification scheme, assisting with
      the number of synoptic categories needed  for  stratification and in defining the
      variability limits within categories when the wind field is applied  in the deterministic
      modelling effort. The statistician in consultation with the principal scientists, will set
      the limits on the meteorological data stratification.

      Participate with the individuals who have contributed substantially to the project in
      the preparation and submission of a research paper to a  peer reviewed journal.

Chemical Analysis-All the filters including those already analyzed will be analyzed using
XRF. The contractor will perform the analysis and report the information after the sample
date has been replaced  by a random number supplied by the statistician  or his agent.
Sample  random  numbers would be attached to the filters by the following  procedure:

      A list of dates versus random numbers would be prepared by the statistician.

      The sample ID numbers corresponding to the dates would be used to generate a list
      of sample ID versus random number.  The sample ID would be replaced by the
      random number using the following procedure:

            Two individuals not associated with the project would travel to Oregon (NEA)
            to handle the samples within the contractor's facility.

            The first individual would place the random number associated with the ID
            number on the sample container.

            The second individual would check to be certain that  the two numbers were
            correct before removing the ID number.

The primary element of interest to this study is sulfur.  Selenium, arsenic and the other trace
elements are of secondary  importance.  These  elements are  stable, so the quality of the
sample should not have degraded with time. The filter analyses will be performed through
the external contractor who performed the original analysis.

      Sulfur The sulfate concentration will be determined by measuring elemental sulfur
      using XRF at an intermediate protocol, Protocol 5.

      Arsenic, Selenium and Other Trace Elements The trace element concentrations will
      be obtained from the XRF data. If at the end of the experiment it becomes essential
      to use additional elements as tracers, arsenic and selenium could be determined
      using neutron activation.  Arsenic has a short half-life and will be counted by the
      contractors.  The long half-life of selenium as well as other long half-life elements
      will be counted at DRI to lower the overall costs of sample analysis.

Meteorological Classification-The sampling period was  a 24-hr day,  from midnight to
midnight,  starting in June  1985.  Some adjustments in data handling will be made for the
data taken on 8 and 16 hour • increments prior to 1985. The  meteorological  conditions
prevailing during each sampling  period will  be  classified using the meteorological
classification scheme developed by Farber et al.  (1989) with some modification to
incorporate  more surface data and a probability of the occurrence  of cloud, based  upon
surface observations and upper  air observations.  All sample days will be included in the
computer  calculations of the classification  probabilities.  As a part of the classification, a
parameter quantifying the strength of the synoptic flow, such as geostrophic wind, height
gradient or vorticity, will be tabulated. Statistical analysis of the  strength parameter will be
used to define limits on the wind  speed and direction parameters used in deterministic

Deterministic Modelling—A minimum of two meteorological models with the appropriate
grid  spacing (telescoping  grid starting at 1  km)  will  be  exercised  for  each of the
meteorological classifications. -  The  strength of the synoptic flow  determined from the
classification analysis will serve as  an input to the model.  The wind speed and direction
parameters and their variance will be fixed prior to running the models.  At the present


time, the addition of a chemical module to the meteorological model is not contemplated.
However, should a good chemical module become available it would be exercised along with
the meteorological model.

The transport modelling will be used to assign nominal MPP impacts at Meadview.  A
potential dosage (concentration x time) will be calculated from the simulated dispersion and
duration, of the plume at Meadview. Calculations will be performed for  all classifications
and synoptic  strengths, yielding a  nominal MPP impact corresponding to each sampling
interval.  These daily nominal impacts will serve as input variables to the statistical analysis,
inputs that incorporate all relevant meteorological information in a physically correct way.

SO2 Emissions—A subcontract will be awarded  to  an outside contractor specializing in
emissions inventory following a competitive solicitation. The firm will  inventory the regional
SO2 emissions  and report the results by  month  and subregion.  The inventory will be
compiled for all types of sources for the period of the study, and will be used as a guide to
regional changes in the background SO2/SO4 concentrations.

Project Personnel--The project would be undertaken by the Desert and Intermountain Air
Transport Program (DMAT) under the sponsorship of Southern California Edison Company
(SCE).  The  project manager will be Dr. Thomas Hoffer,  the coordinating scientist Dr.
Warren White, the statistician Dr. Paul Switzer and the deterministic modellers Drs. Leif
Enger, Darko Koracin and David Rogers. The synoptic classification  will be undertaken by
a team comprised of Dr. David Rogers, Dr. Mark Green, Dr. Rob Farber and  Sara Pryor.

Summary-A  reanalysis of data collected during the MPP outage is proposed to refine and
strengthen the bound  on  MPP's contribution to haze  in the GCNP. The experiment, as
proposed, will strive to eliminate bias in the application of data stratification and statistical

The project schedule calls for an immediate start with a spring 1992 completion date.

                             Appendix 6

                             THE CSU  RAMS
   The numerical atmospheric models developed independently under .the direction of
William R.  Cotton and Roger A.  Pielke have recently been combined  into the CSU Re-
gional Atmospheric Modelling System (RAMS). Development of many of the physical mod-
ules  has been accomplished over  the past  15  years and has involved over 50  man  years
of effort.  RAMS is a general and  flexible modelling system rather than a single purpose
model.  For example, current  research using RAMS  includes atmospheric scales ranging
from large eddy simulations (Az & 100 m)  to mesoscale simulations of convective systems
(Az » 100  km). This paper will discuss the options available in RAMS, the engineering
aspects of the system and how the flexibility is attained.


   RAMS is a merging of basically three models  that were designed to simulate different
atmospheric circulations.  These were a non-hydrostatic cloud model (Tripoli and Cotton,
1982) and two hydrostatic mesoscale models (Tremback et ai, 1985 and Mahrer and Pielke,
1977). The  capability of RAMS was recently augmented with the implementation of 2-way
interactive grid nesting.  Because  of this, the modelling system contains many options for
various physical and numerical processes. These options are listed below.
   The following options are currently available in configuring a model:

   1.  Basic  equations:

      Option 1 Non-hydrostatic time-split compressible (Tripoli and Cotton, 1980)
      Option 2 Hydrostatic incompressible or compressible (Tremback et ai, 1985)

   2.  Dimensionality: 1, 2, or 3 spatial dimensions

   3.  Vertical coordinate:

      Option 1 Standard cartesian
      Option 2 Sigma-z

   4.  Horizontal coordinate:

      Option 1 Standard cartesian
      Option 2 Polar stereographic

   5.  Grid Structure:

        • Arakawa-C grid stagger
        • Unlimited  number of nested grids
        • Unlimited  number of levels of nesting
        t Ability to add and subtract  nests
        • Moveable nests


 6.  Finite differencing:

    Option 1 leapfrog on long timestep, forward-backward on small timestep, 2nd or 4th
        order flux conservative advection.
    Option 2 forward-backward time split, 2nd or 6th order flux conservative advection
        (Tremback et ai, 1987)

 7.  Turbulence closure:

    Option 1 Smagorinsky-type eddy viscosity with Rt dependence
    Option 2 Level 2.5 type closure using eddy viscosity as a function of a prognostic
        turbulent kinetic energy
    Option 3 O'Brien profile function in a convective boundary layer (Mahrer and Pielke,
        1977); local exchange coefficient in a stable boundary layer (McNider, 1981).

 8.  Condensation

    Option 1 Grid points fully saturated or unsaturated
    Option 2 No condensation

 9.  Cloud microphysics

    Option 1 Warm rain conversion and accretion of cloud water (rc) to raindrops (rr),
        evaporation and sedimentation (Tripoli and Cotton, 1980)
    Option 2 Option 1 plus specified nucleation of ice crystals (r,-), conversion nucleation
        and accretion of  graupel (ra), growth of ice crystals (r,-), evaporation, melting
        and sedimentation (see Cotton et ai,  1982)
    Option 3 Option 1 plus option 2 plus predicted nucleation and sink of crystal con-
        centration (./Vi), conversion and growth of aggregates (rfl), melting, evaporation
        and sedimentation.  The nucleation model includes: sorption/deposition, contact
        nucleation by Brownian collision plus  thermophoresis  plus diffusiophoresis, sec-
        ondary ice crystal production by rime-splinter mechanism (Cotton  et a/.,  1986).
    Option 4 No precipitation processes

10.  Radiation:

    Option 1 Shortwave  radiation model including molecular scattering,  absorption of
        clear air (Yamamoto,  1962), ozone absorption (Lacis and Hansen, 1974) and
        reflectance, transmittance and  absorptance of a cloud layer (Stephens,  1978),
        clear-cloudy mixed layer  approach  (Stephens, 1977).  (See Chen  and Cotton
        1983, 1987.)  '
    Option 2 Shortwave  radiation model described by Mahrer and Pielke  (1977) which
        includes  the effects of forward Rayleigh scattering (Atwater and Brown,  1974),
        absorption by water vapor (McDonald, 1960), and terrain slope (Kondrat'yev,
    Option 3 Longwave radiation model including emissivity of a clear atmosphere (Rodgers,
        1967), emissivity of cloud layer (Stephens, 1978),  and emissivity of "clear and
        cloudy" mixed layer (Herman and Goody, 1976)


    Option 4 Longwave radiation model described by Mahrer and Pielke (1977) includ-
        ing emissivities of water vapor (Jacobs  et ai,  1974) and carbon dioxide (Kon-
        drat'yev, 1969) and the computationally efficient technique of Sasamori (1972).
    Option 5 No radiation

11.  Transport and diffusion modules:

    Option 1 Semi-stochastic particle model for point and line sources of pollution (Mc-
        Nider, 1981)

12.  Lower boundary:

    Option 1 Surface layer similarity theory based on Louis (1979) as a function of spec-
        ified surface roughness over land and predicted sea  surface roughness  based on
        Garratt and Brost (1981).
    Option 2 Surface layer temperature and moisture fluxes are diagnosed as a function
        of the ground surface temperature derived from a surface energy balance (Mahrer
        and Pielke, 1977). The energy balance includes longwave and shortwave radiative
        fluxes, latent  and sensible  heat fluxes, and conduction from below the surface.
        To include the latter effect, a multi-level prognostic soil temperature model is
    Option 3 Modified form of Option 2 with prognostic surface  equations ^Tremback
        and Kessler, 1985)
    Option 4 Same as Option 2, except vegetation parameterizations are included (Mc-
        Cumber and Pielke, 1981; McCumber, 1980)

13.  Upper boundary conditions:

    Option 1 Rigid lid (non-hydrostatic only)
    Option 2 Rayleigh Friction layer plus Option 1-4
    Option 3 Prognostic surface pressure (hydrostatic only)
    Option 4 Material surface top. (hydrostatic only)  (Mahrer and Pielke, 1977)
    Option 5 Gravity wave radiation condition  (Klemp and Durran, 1983)

14.  Lateral boundary conditions:

    Option 1 Klemp and Wilhelmson (1978a,b) radiative boundary conditions
    Option 2 Orlanski (1976) radiative boundary conditions
    Option 3 Klemp and Lilly (1978) radiative  boundary condition
    Option 4 Option  1,  2 or 3 coupled with Mesoscale Compensation Region (MCR')
        described by Tripoli and Cotton (1982) with fixed conditions at MCR boundary
    Option 5 The sponge boundary condition of Perkey and  Kreitzberg (1976) when
        large scale data is available from objectively analyzed data fields or a larger scale
        model run. This  condition  includes  a viscous region and the introduction of the
        large scale fields into the model computations near the lateral boundaries.

15.  Initialization


     Option 1 Horizontally homogeneous.
     Option 2 Option  1 plus variations to  force cloud initiation.
     Option 3 NMC data and/or soundings objectively analyzed on isentropic surface
          and interpolated to the model grid.
     Option 4 NMC data interpolated to the model grid.

   As one can see, RAMS is quite a versatile modelling system. RAMS has been applied
to the simulation of the following weather  phenomena.

   1. Towering cumuli and their modification

   2. Mature tropical and mid-latitude cumulonimbi

   3. Dry mountain slope and valley circulations

   4. Orographic cloud formation

   5. Marine stratocumulus clouds

   6. Sea breeze circulations

   7. Mountain wave flow

   8. Large eddy simulation of power plant plume dispersal

   9. Large eddy simulation of convective  boundary layer

  10. Urban circulations

  11. Lake effect storms

  12. Tropical and mid-latitude convective systems


   Because of the large  number of options in RAMS, the structuring of the code needs to
be carefully considered.  This section  will  discuss various aspects of the code structure of
the system.

   Pre-processor The code of RAMS is written in as close to the FORTRAN 77 standard
as possible. However, with a program as large as this, the FORTRAN standard is lacking in
several features such as global PARAMETER and COMMON statements and conditional
compilation.  To remedy these insufficiencies, the RAMS code takes advantage of a pre-
processor written as part of the  RAMS  package. This pre-processor itself is written in the
77 standard so that the package as a whole is highly portable. It takes full advantage of the
character features of FORTRAN and has executed successfully on a number of machines
including a VAX, CRAY-1,  CRAY-X-MP,  and CYBER 205 without modification. Some of
the features of the pre-processor are described below:

  1) By including a character in the first column of a line of code, that line can be "acti-
     vated" or "eliminated" from the compile file. This allows for  conditional compilation
     of single lines or entire sections of code.

  2) A pre-processor variable can be  set to a value.  This variable can then be used in
     other expressions including a pre-processor IF or block IF  to conditionally set other
     pre-processor variables.  These variables  also can be  converted to FORTRAN  PA-
     RAMETER statements which can be inserted anywhere in the rest of the code.

  3) A group of statements  can be delineated as a "global" which  then  can  be inserted
     anywhere in the code. This is very useful for groups of COMMON and PARAMETER

  4) DO loops  can be constructed  in a DO/ENDDO syntax,  eliminating the need for
     statement labels on the DO loops.

   Two-way interactive grid nesting The use of grid nesting allows a wider range of  mo-
tion  scales to  be modeled simultaneously and interactively.  It can greatly ease the limi-
tations of unnested simulations in which a compromise must be  reached between covering
an adequately large spatial domain and obtaining sufficient resolution of a particular local
phenomenon.  With nesting, RAMS can now feasibly model mesoscale circulations in a
large domain where low resolution is adequate, and at the same time resolve the largo eddy
structure within a  cumulus cloud in a subdomain of the simulation.

   Nesting in RAMS is set up such that the same model code for each physical process
such as advection is used for each grid.  This makes it easy for any  desired  number of grids
to be used without having to duplicate code for each one. Also, it is easy to add or remove
a nested grid  in time, and to change its size or location.  There  is  still the flexibility of
choosing many model options independently for different grids.

   RAMS has adopted the two-way interactive nesting procedure described in Clark  and
Farley  (1984). This algorithm is the means by which the different nested grids communicate
with each other. The process of advancing coarse grid A and fine nested grid B forward in
time one step begins with advancing grid A alone as if it contained no nest within.  The
computed fields from A are then interpolated tri-quadratically to the boundary points of B.
The  interior of B is then updated under the influence of its  interpolated boundary values.
Finally, the field values of A in the region where B exists are replaced by local averages from
the fields of B. An  increase in efficiency over the Clark and Farley method was implemented
by allowing a coarse grid to be run at a longer timestep than a fine grid.

   The following options are available with nesting in RAMS:

  1) There is no imposed limit (only a practical one) to the number of nested grids which
     can be used.

  2) When two grids B and C are nested within grid A,  they may be either independent
     (occupying different space)  or C may be nested within B.

  3) The increase in spatial resolution of a nested grid may be any integer multiple of its
     parent grid resolution. Moreover, this multiple may be specified independently for the
     three coordinate directions.

  4) A nested grid may, but need not, start from the ground and extend to the model
     domain top.

  5) A nested grid may be added or  removed at any time during a simulation.

  6) A nested grid can travel horizontally at a prescribed velocity.

   I/O structure For those  machines with limited central  memory and a "non-virtual"
operating system or for efficiency on virtual systems, RAMS is constructed with a disk I/O
scheme. When the scheme is operating, a subset of the model's three-dimensional variables
will reside in central memory at any one time.  Computations then can be performed with
this subset.  When these computations are finished,  a new subset of three-dimensional
variables -are requested and  computations performed with these.  The RAMS  structure,
thus, is dependent on this I/O scheme and consists of a series of calls to the  I/O  scheme
and to the routines which do the calculations.

   Modularity For flexibility, RAMS is written as modular as possible.  Each individual
physical parameterization or numerical process is put in a separate subroutine so that the
routines can easily be replaced for different options or with new developments.

   Computational routines The  routines that  do the actual computations for the model
are written so that the implementor of a new or replacement routine does not need to
be concerned  with most of the details  of the rest of the model computations.  All three-
dimensional variables are "passed" to the subroutines through the call statement with other
variables passed through COMMON.  The implementor then  has the flexibility  to structure
his routine in  whatever manner he wishes to produce  the desired result.  This concept will
also make the implementation of routines from other models and programs easier with less
modification required.

   Analysis routines A set of subroutines has been developed for analyzing and  plotting a
variety of quantities from fields output from RAMS. This greatly facilitates the interpreta-
tion and understanding of modeled atmospheric phenomena. The quantities diagnosed by
these routines include vorticity, divergence, streamfunction, energy, momentum flux, most
variances  and  covariances, and layer averaged quantities.

Atwater, M.A. and P.S. Brown, Jr., 1974: Numerical calculation of the latitudinal variation
      of solar radiation for an atmosphere of varying opacity. J. Appl. Meteor., 13, 289-297.

Chen,  C.  and W.R. Cotton, 1983:  A one-dimensional simulation of the stratocumulus-
      capped mixed layer. Bound.-Layer Meteor., 25, 289-321.

Chen, C. and W.R. Cotton, 1987: The physics of the marine stratocumulus- capped mixed
      layer: J. Atmos. Sci., 44,  2951-2977.


Clark, T.L., and R.D.  Farley,  1984: Severe downslope windstorm calculations in two and
       three spatial dimensions using anelastic interactive grid nesting: A possible mecha-
       nism for gustiness. J. Atmos. Sci., 41, 329-350.

Cotton, W.R.,  M.A. Stephens, T.  Nehrkorn,  and  G.J. Tripoli, 1982:  The Colorado State
       University three-dimensional cloud/mesoscale  model  1982. Part II: An  ice phase
       parameterization. J. de Rech. Atmos.,  16, 295-320.

Cotton, W.R., G.J. Tripoli, R.M. Rauber, and E.A. Mulvihill, 1986: Numerical simulation
       of the effects of varying ice crystal nucleation rates and aggregation  processes on
       orographic snowfall. J.  Climate Appl.  Meteor., 25, 1658-1680.

Garratt, J.R., and R.A. Brost, 1981: Radiative cooling effects within the above the noctur-
       nal boundary layer. J. Atmos. Sci.. 38, 2730-2746.

Herman, G. and R. Goody, 1976: Formation and persistence of summertime arctic stratus
       clouds. J. Atmos. Sci.,  33, 1537-1553.

Jacobs,C.A., J.P Pandolfo and M.A. Atwater, 1974: A description of a general three dimen-
       sional numerical simulation  model of a coupled air-water and/or air-land boundary
       layer. IFYGL final report, CEM Report No. 5131-509a.

Klemp, J.B. and D.R. Durran, 1983:  An upper  boundary condition permitting internal
       gravity wave radiation  in numerical mesoscale models. Mori. Wea. Rev., Ill, 430-

Klemp, J.B. and D.K.  Lilly, 1978: Numerical simulation of hydrostatic mountain waves. J.
       Atmos. Sci., 35, 78-107.
Klemp, J.B. and R.B.  Wilhelmson, 1978a: The simulation of three-dimensional convective
       storm dynamics. J. Atmos. Sci., 35, 1070-1096.

Klemp, J.B. and R.B. Wilhelmson, 1978b:  Simulations of right- and left-moving storms
       produced through storm splitting. J. Atmos. Sci., 35, 1097-1110.

Kondrat'yev, J., 1969:  Radiation in the Atmosphere.  Academic Press, New York, 912 pp.

Lacis, A.A., and J.  Hansen, 1974: A parameterization for the absorption of solar radiation
       in earth's atmosphere. J. Atmos. Sci.,  31,  118-133.

Louis, J.F., 1979: A parametric model of vertical eddy fluxes in  the atmosphere. Bound.-
       Layer Meteor., 17, 187-202.

Mahrer, Y. and R.A. Pielke, 1977: A numerical study of the airflow over irregular terrain.
       Beitrage zur Physik der Atmosphare, 50, 98-113.

McCumber, M.D., 1980: A numerical-simulation of the influence of heat and moisture fluxes
      upon mesoscale circulation. Ph.D. dissertation, Dept. of Environmental Science, Uni-
      versity of Virginia.

McCumber, M.C. and R.A. Pielke, 1981: Simulation of the effects of surface fluxes of heat
      and moisture in a mesoscale numerical model. Part I: Soil layer. J. Geophys.  Res.,
      86, 9929-9938.

McDonald, J.E., 1960:  Direct absorption of solar radiation by atmospheric water vapor. J.
      Meteor., 17, 319-328.

McNider,  R.T., 1981: Investigation of the impact of topographic circulations on the trans-
      port and dispersion of air pollutants. Ph.D. dissertation, University of Virginia, Char-
      lottesville, VA 22903.

Orlanski, I., 1976: A simple boundary condition for unbounded hyperbolic flows. J.  Comput.
      Phys., 21, 251-269.

Perkey,  D.J. and C.W. Kreitzberg, 1976:  A time-dependent lateral  boundary scheme for
      limited-area primitive equation models. Man. Wea. Rev.,  104, 744-755.

Rodgers, C.D., 1967: The use of emissivity in atmospheric radiation calculations. Quart. J.
      Roy. Meteor. Soc., 93, 43-54.

Sasamori, T., 1972: A linear harmonic analysis of atmospheric motion with radiative dissi-
      pation. J. Meteor. Soc. Japan, 50, 505-518.

Stephens, G.L., 1977: The transfer or radiation in cloudy atmosphere. Ph.D.  Thesis. Mete-
      orology Department, University of Melbourne.

Stephens, G.L., 1978:  Radiation profiles  in extended water clouds. Webster Theory. J.
      Atmos. Sci., 35, 2111-2122.

Tremback, C.J. and R. Kessler, 1985: A surface temperature and moisture  parameteriza-
      tion for use in mesoscale numerical models.  Preprints,  7th Conference on Numerical
      Weather Prediction, 17-20 June 1985, Montreal, Canada, AMS.

Tremback, C.J., G.J. Tripoli, and W.R. Cotton, 1985: A  regional  scale atmospheric nu-
      merical model including explicit moist physics and a hydrostatic time-split scheme.
      Preprints,  7th Conference on Numerical Weather Prediction, June 17-20, 1985, Mon-
      treal, Quebec, AMS.

Tremback, C.J., J. Powell,  W.R.  Cotton, and R.A. Pielke, 1987:   The  forward in time
      upstream advection scheme: Extension to higher orders. Mon.  Wea. Rev., 115, 540-

Tripoli, G.J. and W.R. Cotton, 1980:-A numerical investigation of several factors contribut-
       ing to the observed variable intensity of deep convection over South Florida. J. Appl.
       Meteor., 19, 1037-1063.

Tripoli, G.J.,  and W.R. Cotton, 1982:  The  Colorado State University  three-dimensional
       cloud/mesoscale model - 1982. Part I: General theoretical framework and sensitivity
       experiments. J. de Rech. Atmos., 16, 185-220.

Yamamoto, G., 1962:  Direct absorption  of  solar radiation by atmospheric water  vapor
       carbon dioxide and molecular oxygen.  J. Atmos. Sci., 19, 182-188.

                               Appendix 7




The 1990 Clear Air Act explicitly recognizes  the existence of long range transport of
air pollution.  Several provisions of this significant new law require regulatory actions
that involve multi-state regions,  dictated by regional-scale air pollution.  For instance,
the Act requires the establishment of Transport Commissions over the next five years.
These commissions will be  charged with the policy developments  for "airshed" on
regional scale involving several neighboring states.

The work of such commissions and many other provisions of the 1990 law has a strong
need for technical input on  the nature and scope  of regional air pollution.  Typical
questions may be: What is the  region  of influence for specific sources; What are the
major source regions contributing  to  -a given receptor;  How  will certain emission
reduction scenarios reflect on ambient pollution levels.

In the past,  the answers to  such questions have been obtained either form intensive
monitoring and measurement campaigns or from prognostic regional models.  Intensive
measurement programs are expensive and generally provide answers applicable only to
measurement domain.   Prognostic models, on the other hand, are  in general rather
unreliable.  Hence, for the effective implementation  of the new law, new approaches
and tools are needed.  The PC diagnostic Monte Carlo model  proposed for this project,
will provide such a policy oriented data analysis/interpretation.

Over the past two decades, much scientific knowledge has  been gathered about the
nature and  scope of regional air pollution.  In fact, it  can be stated that the  main
causes, and  physico-chemical  processes that characterize regional  air pollution are
reasonably well understood.

The CAPITA diagnostic Monte Carlo  model, encapsulates and describes much of the
knowledge about regional sulfur pattern.  It was developed in the early 1980s for the
analysis and interpretation of regional sulfur and visibility  data.  It's application to
other areas are illustrated in section 1.6.

The initial version of the CAPITA model served primarily as a research tool.  In the
1990s there will be a strong need for  operational, easy to use well calibrated models
that can aid the implementation of the complex new Act.   The proposed PC Monte
Carlo model is intended to be a tool to aid policy-related decision making.


The  purpose of this task is to  develop an interactive and physics-based data analysis
tool  for the  analysis and interpretation of visibility related data.  The data analysis tool
is to aid policy oriented and scientific decision making.  The results of the work should
be directly applicable to the implementation of the 1990 Clean Air Act.
1.3  GOALS

The project has the following specific goals:

    a. Implement a personal computer-based version of the CAPITA Monte Carlo
       regional model (PCMC)
    b. Re-examine the calibration of the diagnostic model using more recent high
       quality aerosol, gaseous, precipitation chemistry, and visibility related data sets.
    c. Develop interfaces to meteorological transport data produced by other, more
       elaborate meteorological models.
    d. Present the results suitable for answering policy questions, such as those posed
       by the new Clean Air Transport Commissions.
    e. Develop interactive graphic user interface that will:
         aid  the operation of the model by non-programmers
         facilitate the graphic display  of results
       -  presentation of the physical entities, such  as spatial concentration maps, time
          charts, frequency distributions
    f.  As much as possible,  use off-the-shelf robust software building blocks in the
       creation of the interactive PC Monte Carlo model.

The scope of  this task will  include the porting,  testing, and  re-calibration of the
CAPITA  Monte Carlo  model on a PC platform.   It  also  involves  building  user
interfaces for the input/output  of model data.

In this task, the main scientific/regulatory application of the PCMC  model will  be to
visibility.   This work will not involve significant new research  areas in of atmospheric
processes,  policy analysis or other fields.  Rather, it will use the available knowledge in
these areas and package these into generally usable PC tools.

The results of this task will be an interactive data analysis and presentation package that
will allow the simulation modeling of visibility-related atmospheric processes.  The
model will be packaged as a tool.  As a tool it should be usable by  policy  analysts

within and outside the government as well as by the research community.


The  proposed  interactive data analysis tool will  utilize  the  CAPITA Monte Carlo
regional  atmospheric transport/transformation/removal  model.  The model principles
and some of its applications are described by Patterson et al.  (1981).  The following
description states the concept of the model and illustrates some of its past applications
relevant to this work.

In the Mote  Carlo modelling  approach, simulated pollutant quanta  (particles) are
"emitted" in accordance with an emission inventory.  These quanta are  moved in fixed
time increments using the interpolated measured wind  fields.  During  their transport,
the pollutant  quanta may be  subject to chemical transformations  or  removal.  vThe
    r         ^        J       J                            .     fi>.vo.,f,- • I }•"•*• 5
transport  as well  as the transformations are somewhat randomized/; hence the  name
Monte Carlo method.  The method is also referred to as the Direct Simulation  method
since the physico-chemical processes are  simulated  as  discrete  events rather than
obtained  from the  solution of differential  equations.   The  result of Monte  Carlo
simulations  is  a  large  number  (10^  -   lO^)  of  pollutant  "particles"  dispersed
geographically for every time step of the simulation.

Table 1. Summary of CAPITA regional model
Model Type
Receptor Grid
Grid Resolution
. Model Domain
Model Output
Input requirements
Cloud cover
Mix heights
Mixed Layer
Horizontal Dispersion
Vertical Dispersion
SC>2 Transformation Rate
. Dry Deposition Rate
Wet Removal Rate
Monte Carlo (Lagrangian, and Eulerian in the limit ot large number of quanta)
52 x 60 grids (variable for PCMC)
127 x 127 km at 60 degrees north latitude (one-third of U.S. National Meteorological
Center grid spacing)
North America (variable for PCMC)
Fields of particles each 3-h, each particle representing a mass of emitted pollutant
remaining in the atmosphere as each possible chemical species; converted to fields of
daily SC>2, SO^ concentration and dry and wet depositions at all grid points
Seasonal, surface and tall stack 1 890-i£«0 SC>2 grids.
0000 and 1200 GMT rawinsonde wind profiles at 1 30 sites. In PCMC, externally
generated wind fields accepted
3-hourly observation of precipitation of three intensities at surface synoptic sites.
gridded from 3h surface synoptic sites
gridded from 3h surface synoptic sites
Climatological by season, from work of Holzworth (1972) and Portelli (1977), of
maximum afternoon mixing heights
3-h SC>2 emissions. Released in mixed layer in day, and either 150-450 layer or 0-
150m at night; 1 % primary 804
Inverse-distance squared weighting. Upper air rawinsonde winds are interpolated in
space into 11 layers (0-150, 150-450, 450-750, 750-1050, 1050-1350, 1350-1650,
1650-1950, 1950-2250, 2250-2850, 2850-3450, 3450-5250 m above ground). Wind
for each layer is the vector average, and winds are linearly interpolated -from 1 2h to
3h, using seasonal diurnal interpolation factors at each height to reflect nighttime jets
and midday drag from convective mixing.
3h grids of the space-time average probability of encountering precipitation are used
to scale local wet removal rates as fraction of the maximum rate.
Climatological average, which varies with geographic location and season of year.
Representative peak P.M. values are 800 (winter), 1200 (spring, fall), 1350
(summer). Fixed 150m at night.
Lateral displacement by veering of layers overnight; "eddy diffusion" K = 2000 m^/s
day, 100 m^/s night.
instantaneously mixed throughout mixed layer during day (0900- 1 800 LST); no
vertical miring at night.
Varies seasonally, diurnal and locally. "Dry" part proportional to solar radiation,
function of latitude/season, time of day, and local total sky cover. "Wet" part
proportional to local surface dewpoint.
Zero above local mixed height and above 150m night surface layer. Varies with
stomatal density and opening.
Zero above local mixed height within mixing layer (precipitation probability over a
grid during a specified time) x (wet removal rate constant, 100%/h for 864 and
10%/h for S02).

In the PC implementation,  some of the above model parameters will be changed.  The
changes  will  incorporate  better physico-chemical  knowledge,  better  computational
performance and a more general user interface.   In what follows, the application of the
Monte  Carlo   model  in different  domains is  illustrated.    We consider that  the
illustrations below demonstrate the potential applications of the new PC based model.

Receptor Modeling and Back Trajectory Analysis.   The simplest application of the
model is  for showing back trajectories leading to a specific receptor site.  The approach
is illustrated below as applied to the analysis and interpretation of the measurements in
the VISTTA program (Macias et al.  1981).
 Fig. 1.  (a) Map of the portion of the southwestern U.S. of interest in this study. Some of the major emission sources are
 indicated, (b) (c) and (d)  Intercompanson of calculated air mass histories. Each figure shows two estimates for the history
 of air sampled at Page at 11.00 MST on the indicated date. Single heavy  tracks are  back trajectories derived by
 meteorologists from measurements of upper-air winds (MRI, 1980). Multiple light tracks are back trajectories computed
 by CAPITA Monte Carlo model from adjusted midday surface winds, taking dispersion into account (Patterson ei ai.
 1980). The two estimates, based on independent manipulation of independent data sets, agree satisfactorily for the three
                 differing transport regimes shown here, and for 16 of the 18 days considered.

Multiple Plume Modeling. The model was also applied for the modeling of single and
multiple  plume dispersion.   The  figure below indicates the model usage  for the
visualization of multiple plumes in -the Southwestern U.S. (Macias et al.,  1981).
                                           COPPER SMELTERS
                                        {~~1 FIRES
     Fig. 2. Calculated plumes from potential source areas, from CAPITA Monte
     Carlo  model.  Each figure shows approximate  extent, at  11.00 MST on  the
     indicated date, of material released during the preceding 48 h (A, C) or 24 h  (B).
     The three rows cover the three characteristic time periods identified by Macias el
     al. (1981). (A) 27-30 June: Air which had stagnated over southern California for
     several days moved into the Page area during the  latter half of this period. (B) 3-6
     July: Shifting  southerly winds brought material to the vicinity of Page from
     wildfires north of Phoenix and smellers southeast of Tucson. (C) 8-11 July: A shift
     to more southerly winds during the latter half of this period diminished the impact
                           of southern California on Page.

Regional Transmission Modeling.  The most extensive use of the CAPITA model has
been the regional modeling of sulfates and extinction coefficients over the eastern U.S.
(Patterson et al., 1981).  In that application, the daily  pattern  for sulfate  aerosol and
extinction coefficients were  simulated as shown in the figure below.
                           SURE      MODEL           AND
          ::: 3-4

          = 4-6
                   • >t
= 15-20

= 15-20
I  >2S

                                                                = WW-1022
                   H >W22
           Fie. 3 Daily maps of midday />„, corrected lo 60% RH (first column), 24 h average SURE SO;
         (2nd column), modeled 24 h distribution of emitted sulfur quanta (third column), and unmodified noon
         surface wind field overlaid with the sea level pressure (last column) for 1-6 Aueust 1977.

The daily  pattern  of  measured  $04,  model 804,  visual  range-derived  extinction

coefficient, bext, and air residence time is shown in Figure 4.










2.5 To


                                 3  uT
                               10        15        20

                                       AUGUST 1977
        FIG. I|. Daily spatial averages within Ihe SURE region of SURE sulfale (thin line), fc,.x,

      (thick line) and model sulfate (dashed line). Model sulfale scale assumes 1100 m scale height.

      The dotted trace is proportional to the number of conservative quanta from a uniform emis-

      sion grid remaining within the eastern United States, which defines a regional residence time

      for the airmass.

Global Pollution Modeling. (Patterson and Husar. 1981)
                     Fig. 5.  (a) Emission field of trajectory origins, (b) Sample wind field grid.
         * so    so-co    eo-aoo   aoo-uo   soo-noo   >tooo

Fig. 5  Seasonal maps of vertical burden arising from 1974
850mbar winds and 5-day residence time. Shadings represent
sum  of trajectory endpomts (puff arrivals  per NMC  grid
        square) weighted by decay for exp(-i/r).

                                                                             no-aoo   200400
                                                 Fig. 3. Seasonal maps of vertical burden arising from 1974
                                                 850mbar winds and 10-day residence time. Shadings repre-
                                                 sent sum of trajectory endpoints (puff arrivals per NMC grid
                                                        square) weighted by decay factor cxp(-r/t).

 Retrospective Modeling.  In this application, the model transfer matrices along with
 historical emission trends were used to reconstruct the SO2 concentration trend in New
 York City Central Park for the period 1900-1980 (Husar et al., 1984) (Figure 6).
400 -
     300 -
     100 -
       0 -<—
                   19<>0       1920       1940    '   1&60       i960
                   a    UPPER EST.             r   LOTTER EST,

Estimated  SO2  concentration  trend  for New  York  city,  Central Park.

 This section states the approach and implementation of the proposed goals stated in
 section 1.3.
 1.7.1 PC Based Version of the Monte Carlo Model (PCMC)

 The new  model  implementation will operate  on standard IBM compatible  personal
 computers.  While  the  model kernel  will retain  its  features,  the  model  will  be
 completely wrapped  into a graphic user interface.  It will utilize the readily available
 Microsoft Windows graphic operating environment.

 The. model will be implemented using object oriented programming techniques using
 the C+ +  language.

1.7.2 Re-Calibration of the Model

Following the implementation, the  PCMC model will  be tested and its constants re-
evaluated  using  more recent high quality  aerosol databases.  The  candidate data sets
include  IMPROVE by the  National Park Service; SCENES, a  western U.S.  research
consortium,  and the particle network of NESCAUM (Northeast  States for Coordinated
Air Use Management).  These data sets will allow a more precise evaluation of the
transformation and removal rate constants.
1.7.3 Development of Interfaces for Externally Generated Wind Fields

The  previous model used NWS (National Weather  Service) upper air and  surface
observations  to derive the transport  wind field.   The gridded,  x,y,z dependent  wind
vectors were  generated by the Monte Carlo model itself.

In the PCMC model the above  wind  generation  facilities  will  be  preserved.   In
addition, "hooks"  will be provided to allow the usage of externally generated model
winds. Candidate wind grids include the  NWS 100  km mesh predictive model that is
available operationally. Another wind data source may be the MM4 Mesoscale Model
by NCAR/Penn State.

The  use  of these  external wind field  will not eliminate the  need to  use  the  surface
meteorological observations.   Such input will provide estimates  of  solar radiation,
precipitation  events, relative humidity,  and other variables required by the PCMC.
1.7.4 Present the PCMC Model Output Suitable for Policy Analysis

Unlike the first, research oriented version of the model, the PCMC will be oriented
toward application in regulatory or other decision making.  Hence, the output of the
model will have to be tailored to answer questions relevant to the regulatory function of
EPA or other agencies.

The first such regulatory activity under the 1990 Clean  Air Act, is the formation of
Transport Commissions.  The charge of such commissions is to evaluate the regional
(inter-state) aspects of air pollution. As its first application, The PCMC will provide the
Transport Commissions with a tool to examine the regional air pollution transport from
and to alternative source and receptor areas.
1.7.5  Interactive Graphic User Interface

The PCMC will be Windows-based program.  It's operation is accomplished by menu
selections and point and click queries.  Programming knowledge will not be required.

An example application of the graphic user is outlined below for illustration purposes:

Suppose  a member of the interstate  Transport Commission wishes to evaluate the
potential  impacts of various sources on a given receptor region.   The following user
actions would be required.        -  .

    - Select pollutant of concern.  The emission field for that pollutant is automatically
       displayed on a map.
    - Zoom in on the map and point to a specific location of interest.  The program
       would automatically  display a pie chart for the relative contribution various
       sources to that location.
    - Point to specific sources on  the map.  This would be a query to retrieve all the
       characteristics of that source,  including the emissions.
    - From a menu ask for forward trajectories. The program would automatically
       draw the trajectories for the previously selected receptor location of a year,
       month, day or hour as specified be the user.
    - Select "Show Monitoring Data"  from a menu.  The program would retrieve and
       display the available monitoring data for the previously specified location and
    - Select "Show Model Data" from a menu.  The program would calculate instantly
       the model concentration pattern for the selected location and overlay it to the
       measured date. This would give the commission member a feel for the model

The above illustration is but a small sample of the possible implementation for a user-
friendly data interpretative model.

Since the data will be presented  in physical  units  on maps  and  charts, and  the user
actions will be intuitive, the training and  instruction time  will be small compared to the
use  of the current  modeling and data analysis software.   Hypertext-based  context-
sensitive help will also be available to aid the user.

1.7.6 Use of Robust Software Building Blocks.

The PCMC will  utilize modem, object oriented software building  principles. It will be
object oriented in principle as well as implementation.  It will make use of "Software
1C V (integrated circuits) that are generic, robust, and  suitable for integration  into
larger software applications.

These software  building  blocks  will  include Dynamic   link   Libraries  (DLL's);
Embedded  Objects;  Message  based  communication  among  objects;   Software
Construction  sets,  (such  as ToolBook by Asymetrix  Corporation, and Voyager data
browser by Lantern Corporation.)

A key feature of object oriented approach is that most modules will be reusable. This
will reduce the complexity,  size and  the maintenance cost of the software.


The main deliverable of this project will be a PC based regional model based on the
Monte Carlo principle.  The model will be packaged as policy analysis tool, including
tutorial, as well as on-line and hard copy  documentation.  The PCMC will be made
available and distributable without royalty or other legal constraints.

                          Appendix 8

NAS review of Whitex - Limitations & Suggested Improvements
      The Committee on Haze in National Parks and Wilderness Areas
of the National Research Council, National Academy of Sciences prepared
a report entitled  "Haze in  the Grand Canyon: An Evaluation of the
Winter Haze Intensive Tracer Experiment" (WHITEX).  The WHITEX
experiment studied the effect of the Navajo Generating Station (NGS)
upon visibility in  Grand  Canyon  National  Park  (GCNP).   The
Environmental Protection Agency (EPA) is in the planning stage of a
study (named Project MOHAVE) to determine the effects of the Mohave
Generating Station on visibility in GCNP.
      The NAS report on WHITEX noted a number of limitations in the
study and some suggestions for how the study could have been unproved.
The purpose of this document is to identify how EPA intends to improve
upon the limitations of WHITEX noted by the NAS and to incorporate
the NAS  suggestions  into the Project  MOHAVE study  plan.  NAS
comments (paraphrased)  on the WHITEX study that may be applicable
to Project MOHAVE are listed below. The comments are followed by
responses of how Project  MOHAVE intends to consider these issues.
p.4  (Executive Summary)  Committee identifies problems in the multiple
linear regression analysis (DMB and TMBR):
       1)  Satisfactory tracers are not available for all major sources;
       Project MOHAVE will attempt to identify tracers for all major sources,
source types and source areas.   For example, certain halocarbons may be
used as tracers for the Los Angeles Basin.  Also, sulfur to selenium ratios
may be significantly different for different coal-fired powerplants.  However,
it is acknowledged that all major sources may not have satisfactory tracers

identified.  This lack of complete source profiling often occurs in receptor
modeling and does not necessarily preclude the use of receptor modeling to
obtain quantitative results. However, uncertainties in source profiles needs
to be incorporated into the error analysis.

       2) Interpretation did not account for possible covariance between
          Navajo and other coal-fired powerplants in the area;
       Trajectory  analyses  using   the wind fields  from  the  dynamic
meteorological model will allow determination of times that the MPP plume
and plumes from other sources are jointly present at receptor locations.  This
•will facilitate consideration of covariance  of impacts from MPP and other

       3) Both models treat sulfur  conversion inadequately.
       The exact methodologies of treating sulfur conversion  in the receptor
models has not yet been determined. Rather than scaling  tracer by ambient
surface relative humidity  in  TMBR, as in  WHITEX, other methods  will be
considered.   For example, data may  be  stratified into  "wet" and "dry"
conditions and the model run separately for each subset of data.  Similarly for
the DMB analysis,  instead of assuming constant conversion rates for the
entire data set, subsets of the data may be grouped, w{th Constant rates over
each group.  It is acknowledged that some uncertainty in sulfur conversion is
unavoidable; however,  with  the use of deterministic modeling, checked by
tracer  and  sulfate  data,  along  with  receptor  modeling,   reasonable,
quantitative  estimates of sulfate contributions from each source may be

p.4  WHITEX  did not quantitatively determine the fraction of SO4 aerosol and
resultant haze in  GCNP attributable to NGS.

       As  discussed  above  and  in  response to  other  comments,  with
measurements, deterministic and receptor modeling and model reconciliation,
quantitative apportionment ofsulfate at GCNP can be done, within identified
error bounds.   After sulfate has  been apportioned,  statistical and first
principle approaches  can be used to attribute extinction.

p.4  WHITEX did not adequately quantify the sensitivity of the  analysis to
departure from  model assumptions, nor did it establish an  objective and
quantitative rationale  for selecting among various statistical models.
see response to 2nd comment on page 26.

p. 4   The conceptual framework for  DMB involved physically  unrealistic
simplifications for which  the  effect  on quantitative assessments  was not
       As discussed  elsewhere in the responses, more physically realistic
assumptions will be made wherever possible.  However some simplifications
will  remain,  as  in  all modeling studies.   The  effect  of variations  in
assumptions can be studied to some extent with sensitivity analysis.   Also,
comparison to deterministic models (which also contain simplifications) may
help determine the effect of simplifications  upon quantitative assessments.

p.4  The data base contained weaknesses; especially important was the lack
of measurements below the rim and the paucity of background measurements
(particularly SO4).
        The conceptual plan for Project MOHAVE calls for monitoring  below
the rim of the Grand Canyon and increased background monitoring compared
to  WHITEX,  including S04 and tracer.  It should be  recognized that the
number of feasible monitoring sites is limited due to power requirements and

the inaccessibility of some areas.

p. 4 The background measurements were inadequately incorporated into the
data analyses (in particular, SO4).
       The NAS comments  emphasize that not enough sampling sites were
located in the vicinity ofGCNP (p. 25).  In addition, tracer was not measured
at many locations and only a small subset of tracer data were  analyzed.
Project MOHAVE will have more sampling sites in the vicinity ofGCNP and
operation over a 12-15 month period, compared to  6 weeks for  WHITEX.
This includes more sulfate sample analysis and far more  sample analysis  of
tracer.  However, due to accessibility problems and power requirements, the
number of feasible sampling sites in the location of GCNP is limited.
Thus, the actual number and location of sites may be less than ideal.

p.20  Literature does not demonstrate that MLR  can successfully apportion
secondary species among several source types; therefore  is not advisable  to
rely solely  on such models  for the success of a major field experiment.
       Project MOHAVE is emphasizing the use of deterministic models rather
than MLR for apportionment of secondary  species. The analysis will also use
receptor models and eigenvector analysis as a  check of the  deterministic

p.21  Deterministic met. modeling did not reproduce the diurnal fluctuation
in wind flow observed  at Page.
see response to next comment

p.21  The met. data and deterministic meteorological modeling do not allow
quantification of the contribution  that  NGS  might  have made to haze at
GCNP.  The deterministic modeling cannot pinpoint  the location of'the NGS

plume nor its entrainment into the canyon.  The model uses a grid size of 5
km;  hence it cannot reproduce the complex topography of GCNP nor the
associated small scale meteorological effects, such as gravity flows. Thus the
meteorological studies provide only qualitative evidence of transport.
       Project MOHAVE will more thoroughly model MPP using increased
meteorological data and greater resolution of topography.  Modeling will be
done for the entire 12-15 month study period.   Wind profilers  mil provide a
much increased meteorological data base compared  to the WHITEX study.
Model grid size will be 500 m at key  areas,  allowing greater topographic
resolution and improved representation of small scale flows.   It should be
understood that it is impossible to  exactly model wind fields; of particular
difficulty is flow in highly complex terrain such as the study area.  Monitoring
and modeling  of moisture and chemical transformations  will allow for a
reasonable quantification of MPP impacts to haze at  GCNP.

p. 24  No tracer was used in WHITEX to evaluate urban emissions; therefore
the fraction of haze attributable to these sources is impossible to  calculate.
       Tracers for urban areas will be investigated.   For example, certain
halocarbons have been identified as tracers for the Los Angeles Basin. Other
urban areas,  particularly Las  Vegas will also be investigated for endemic
tracers.  In addition, the deterministic modeling will identify the time periods
when emissions from urban areas are in the Grand Canyon area.

p.  24  The source profile for powerplants was based on  limited aircraft
measurements of NGS  emissions downwind  from the stacks.   The copper
smelter profile was based on old and uncertain data from the literature.
        The planners of Project MOHAVE are aware  of the critical nature of
 accurate source profiles for use in receptor models.  All available data for

powerplant emissions in the region will be used to generate powerplant source
profiles.  The most recent data for smelter emissions will be used.  Resource
limitations preclude significant field efforts to document source profiles of all
important sulfur sources with the potential to impact the GCNP area.

p. 24   Variabilities and uncertainties  in NGS CD4 emission rates led  to
substantial uncertainties in the day to day relationship between CD4 and NGS
sulfur emissions.
       Unlike the WHITEX study,  which used tracer data mainly for receptor
modeling, Project MOHAVE will also use tracer for estimating plume dilution
factors.  For this purpose,  a constant tracer release rate is desirable.  With
variations in MPP load, this will result in variations in tracer to sulfur ratios.
For use  in receptor modeling, as in WHITEX tracer concentrations need to
be scaled to the sulfur emissions, which requires plume age.  Tfie more
sophisticated meteorological modeling to be done for Project MOHAVE will
give a better calculation of plume age than the simple trajectory models used
p.  24   At  Hopi Point,  CD4 concentrations were determined for only 36
samples, an undesirably  small data set for the types and large numbers of
statistical analyses performed on the data.
       The  WHITEX study analyzed a small  number of samples of CD4
because of the very high analysis  costs.  It is expected that perfluorocarbons
will be  used for Project MOHAVE, for which the analysis costs are  not
prohibitive. The tracer sample size will be many times the size for WHITEX,
allowing for a sufficiently large data set for use in statistical analyses.
p. 24  The  ratio of SO2 to CD4 in the stack was not analyzed.
       Project MOHAVE intends to  analyze some stack samples for S02 to
tracer ratio.

p. 24  The report provides little documentation of procedures and quality
assurance for the sampling and analysis of ambient CD4.
       The participants in Project MOHAVE are acutely aware of quality
assurance problems with some  past tracer experiments.   The  skepticism
regarding the quantitative use of tracers  requires not only careful quality
assurance, but  also detailed  documentation of the procedures and quality
assurance  performed.    Project  MOHAVE reports will provide  detailed
documentation of quality assurance for tracer and other data collection.
p. 25   Without data from more stations,  the  effect  of NGS  emissions  is
difficult to differentiate from other sources in the region.
       Project MOHAVE expects to have data from additional stations in the
area  of GCNP  compared to  WHITEX.    Perhaps  more significantly, the
deterministic modeling will help differentiate impacts from MPP and other
p. 26  WHITEX design did not provide the data necessary  to quantify the
effects of departures from the  statistical assumptions made.
       see response to comment #6, page 4.
p. 26   SO4 contribution attributed  to NGS depends strongly on the model
chosen, the tracers included in the model, and the criteria by which the model
is fit to  the  data.   To  establish  a more rational basis  for  quantitative
attribution, more attention must be given to alternative formulations of TMBR
and DMB and the criteria for selecting among them. However, even if these
criteria were adequately considered, the statistical results would  most likely
remain non-robust in  the sense that source attributions generated by  the
various statistical models would probably still  differ substantially from one
another.  One difficulty is that the number of plausible alternative models is
substantial  relative to  the number of samples  for  which CD4  data  are
available.  As the number of models increase, so does  the likelihood that one

of them will test significant merely by chance.
       Model formulations will be done based on theoretical considerations.
Sensitivity analysis of varying model assumptions within reasonable ranges
will be done to determine the bounds of possible results. It is possible that
different receptor (and deterministic) models will yield significantly different
results.   Reconciliation of model results  will be done at this point.  Many
more samples of tracer will be available compared to the WHITEX study, thus
decreasing the likelihood that a model will test significant by chance.
p. 26 WHITEX assumed SO4 yields from NGS and smelter emissions were
proportional  to ambient relative humidity.  This is a simple and  indirect
assumption, which scales intermittent processes along the entire trajectory at
cloud level directly to a continuous variable measured at ground  level.
       In Project MOHA VE, the effect of moisture upon sulfate formation will
be treated more rigorously than done in the  WHITEX report.  In addition to
surface humidity measurements,  the deterministic meteorological model will
give estimates of  humidity at many  vertical levels.   This  information will
include prediction  of clouds, which can be compared to satellite observations.
Rather than scaling linearly with relative humidity, a determination will be
made whether or not the plume  is in contact with clouds.  Rates of sulfate
formation are thought to occur rapidly in clouds, and quite slowly  without,
particularly in winter.   Thus, stratification of data into  "wet" and "dry"
categories seems appropriate.
p.  31   Given the  overriding importance  of the  RH scaling  factor, the
committee believes  that the sensitivity of results to alternative assumptions
should have been explored in formulating the models used for the TMBR and
DMB analyses.  The NFS WHITEX report assumes that the contribution of
background sources,  such as other power plants  and  urban areas,  were
unaffected by RH.  The committee believes the report should have considered
the possibility that yields  from other sources were also affected by RH.

       It is likely that contributions from other sources are affected by relative
humidity.  This will be considered in the analysis.
p. 31 The DMB analyses are dependent on unique "plume ages", the validity
of which is questionable.  Plume ages were estimated only for NGS and not
for other sources.
       Plume ages will be estimated for MPP and a variety of other sources
using wind fields generated by the dynamic meteorological model; this should
provide reasonable estimates of plume ages.
p. 31  DMB  is based upon linear models for the oxidation of SO2 to SO4 and
for the deposition of SO2 and SO4.  In reality, both processes are likely  to
occur at rates that can vary greatly  in space and time.
       see response to comment 3,  page 4.
p. 32  Nonuniformities in conversion and deposition rates lead to variabilities
in the relationship between SO4 concentrations measured at the receptor sites
and tracer concentrations  used  in  the regression  analyses.  Because these
nonuniformities were not taken into  account  in the DMB formulation,  the
DMB results are of questionable applicability.
       see response to comment 3,  page 4.
p.  32  Possible covariance of impacts from NGS and other coal-fired power
plants makes it difficult to  statistically distinguish the relative effects of NGS
and other plants.
        Trajectory  analyses  using the   wind  fields from the dynamic
meteorological model will allow determination of times that the MPP plume
 and plumes from other sources are jointly present at receptor locations. This
 will facilitate consideration of covariance of impacts from MPP and other

p. 35 No H2O2 measurements were made at or near GCNP during WHITEX.
      Measurements ofH2O2 will be made in the study area, under varying

             Association   International   Specialty   Conference
             'Visibility and Fine  Particles',  October 1989,  Estes
             Park,  CC.
                            Appendix 9

William C. Malm
National Park Service, Air Quality Division
Cooperative Institute for Research in the Atmosphere
Colorado State University
Ft. Collins, CO 80523

Hari K. Iyer
Department of Statistics
Colorado State University
Fort Collins, CO 80523

John Watson
EEEC, Desert Research Institute
Reno, NV 89506

Douglas A. Latimer
Latimer & Associates
P.O. Box 4127
Boulder, CO 80306-4127

     The chemical mass balance (CMB) formalism has been used on a semi-routine
basis to apportion emissions used to mass concentrations at specific receptor sites.
Recently, two other techniques, differential mass balance (DMB) and tracer mass
balance regression (TMBR) have been used to apportion'secondary  aerosols to
sources and source types of a variety of receptor areas. CMB uses known source
and receptor  measured tracer profiles (gradients in tracer concentration at one
point in time) to apportion sources at one point in time. DMB uses gradients in
trace elements across space, while TMBR uses changes in tracers across time to
achieve apportionment of primary as well as secondary .aerosol species.  Assump-
tions and limitations of each approach will be addressed and a unified formalism
building on strengths of all three approaches will be presented.


   Receptor modeling approaches rely on known physical and chemical charac-
teristics of gases and particles at receptors and sources to attribute aerosols to a
source or source type. Historically, the CMS formalism has been used to appor-
tion primary particles. This formalism uses known relationships between emitted
tracers and an assumption that various tracer profiles, stay constant as material
is transported from source to  receptor.  These  tracer profiles are then used to
apportion primary species for each time period that a measurement is made at a
receptor site. Other common types of models include principal component anal-
ysis (PCA) and multiple linear regression "(MLR). Explanations of these models
are given by Watson1'2',3 Chow,4 and Hopke.5 All these models are special cases
of a  General  Mass Balance (GMB) model which is deterministic in nature.  A
regressional model similar to MLR is derivable from the GMB equations and will
be referred to as the TMBR model. The TMBR model incorporates changes in
tracer material over time to apportion both primary and secondary aerosols. Fi-
nally, the  DMB model, a special case of GMB and referred to here as a receptor
oriented model,  is really a hybrid model in that it relies on tracer material to
establish atmospheric dispersion characteristics but deterministically accounts for
deposition and oxidation.  Stevens and Lewis,6 Lewis and Stevens,7  and Dzubay
et al.,8 have used models similar to the TMBR and GMB to create a hybrid model
which they have used for source apportionment.

General  Mass  Balance Equations

   Each special case of GMB has its own set of limiting assumptions and special
requirements for solution. The assumptions that need  to be  satisfied for the
mathematical model to be valid will be apparent during the process  of derivation
of the model equations.  Nevertheless, the assumptions will be explicitly stated
after the derivations of the model equations have been explained.  The statistical
aspects of the estimation of the fractional contribution by a given source and the
calculation of the associated uncertainties will also be presented.

                         Notational Conventions

   The following notation will be used throughout.

               Total number of species under consideration = fn.
               Total number of sources under consideration  = n.
               Total number of sampling periods = s.

The subscript  i will be used for indexing the species, j for sources and  k for
sampling periods.
         =  concentration of aerosol species i at source j corresponding to
             sampling period k.
    C^k  =  concentration of aerosol species i at the receptor, attributable
             to source j corresponding  to sampling period k.
     tjk  =  travel time for the air mass from source j to the  receptor, cor-
             responding to sampling period k.
    rijk  =  a factor that accounts for deposition of aerosol species i from
             source j, for sampling period k.
    r*jk  =  a factor that accounts for the formation of aerosol  species i from
             a parent species i" emitted by source j as well as  its deposition
             during transport for sampling period k.
     djk  =  a factor that accounts for dispersion of aerosol  mixture from
             source j during sampling  period fc, as the mixture travels from
             the source to the receptor.
Whenever  a  subscript i denotes  a secondary aerosol species, then the subscript
i' will denote the corresponding parent aerosol species. For instance, if i denotes
S04 then i"  will stand for
                             Model Equations

    It follows from the definitions that for primary aerosol species
 and fpr secondary aerosol components we have.

                                       k + Ci*jkr~kdjk.                     (2)
 The quantities r,-jt are a function of deposition rates and transport  time while
 r'jlt are functions of deposition and transport times as  well as conversion rates.
                                   9 -3

Simple functional  forms for r^k and r,'^ can be derived if it is assumed that
chemical conversion and deposition are governed by first order mechanisms and
conversion and deposition rates are constant in space over some finite increment
in time.

   Let X(t) denote the mass, at time t  after emission, of a species i in a unit
volume of aerosol mixture. Assume, ignoring dispersion temporarily,

which, when solved, yields

                                             + Kl)t)                     '(4)
where X(0) is the mass at  time 0  in unit volume of aerosol  mixture,  i.e.  the
concentration of the species at the source. The quantities Kc and K\ are  the con-
version and deposition rates, respectively, for the species under consideration. The
conversion and deposition rates have been assumed to remain constant through-
out the transport path  in space and time.  If d(t) denotes  the  dispersion factor
corresponding to t time units after emission of the aerosol mixture, then

                    X(t)x = X(Q}cxp(-(Kc + ffi)0
If now the dispersion factor d(t) is taken into account
From this relation it becomes evident that the factor accounting for the formation
of the secondary  aerosol species from its parent species as well as its deposition
during transport  is of the form

Based on the above arguments, when the conversion and deposition rates of the
various  species remain constant  throughout the duration of transport from 'the
source to the receptor
                  rijk  =  exp(-(Ke(iJ,k) + Kd(iJ,k))tjk)              (11)

                 r.   =             Kc(i'J,k]
                         c(i',j, k) + Kd(i>J, k) - Kd(iJ, k)
          {exp(-Kd(i,j, k)t}k) - exp(-(Kc(i-J,k) + Kd(i',j, k)}tjk}}       (12)

 Kc(i, jt k}  =  conversion rate of species i from source j to its secondary form,
               during sampling period k.
 Kd(i,j, k}  =  deposition rate of species i from source j during sampling period
   Let dk = concentration of aerosol component i at the receptor during sampling
period k. Since the concentration of aerosol component i at the receptor is the sum
of the concentrations attributable to various sources, the mass balance equation
for each sampling period k — 1,2, . . . ,5. From this basic equation various special
cases can be derived.
 CMB Model

    The first special case of the GMB equations to be examined is the Chemical
 Mass Balance formalism.


                             Model Equations

   Suppose our list of aerosol components includes only material that is nonre-
active and maintains relative ratios between various species as material is trans-
ported from source to receptor. In this case Kc(i,j,k] are all zero and Kd(i,j,k]
are the same for all elements i.  Their common value is denoted by Kd(j,k] in-
dicating the nondependence on  i.  This  implies that the quantities Tijk do not
depend on i. Then,
which implies that the signature for source j at the source equals the signature
for source j as perceived at the receptor.

   Let Sjk  = ££:! Cijk-  The quantity Sjk is  the  concentration of the aerosol
mixture at the receptor during sampling period  k that is attributable to source j.
The  fraction a,-jjt defined by

is then the fraction of species i in the aerosol mixture at the receptor attributable to
source j during sampling period fc. Assuming Equation (14) is valid, the numbers
o,-jjt for i  = 1,2, ...,m represent the source signature for source j  for sampling
period k.  From Equations (13) and (15)  it follows that  the set of  Equations in
(16) below also holds.
    If the a,-jfc for all the sources affecting the receptor sites are known, then 16
is a system of linear simultaneous equations in n unknowns 5ut, Su,..., 5n*, for
each of the sampling periods k = 1,2,..., s. These are in fact the chemical mass
balance equations. The rank of the system of equations for each k must be equal
to n in order to uniquely solve these equations. In particular, the numbers of
equations must be greater than or equal to.the number of chemical species (i).

    Solutions to the CMB equations that have been used are: 1) a tracer solution;
2) a linear programming solution-;  3) an ordinary weighted least squares solution
with or without  an intercept; 4) a ridge regression weighted least squares solution
with or without an intercept; and 5) an effective variance least squares solution
with or without  an intercept. An estimate of the uncertainty associated with the
source contributions is an integral part of several of these solution methods.


   Weighted linear least squares solutions are preferable to the tracer and linear
programming solutions because: 1) theoretically they yield the most likely solution
to the CMB equations providing model assumptions are met;  2)  they can make
use of all available chemical measurements, not just the so- called tracer species;
3) they are capable of analytically estimating the uncertainty  of the source con-

   CMB software in current use9 applies the effective variance solution developed
and tested by Watson11 because this solution: 1) provides realistic estimates of the
uncertainties of the source contributions (owing to its incorporation of both source
profile and receptor data uncertainties); and 2) chemical species with higher pre-
cisions in both the source and receptor measurements are given greater influence
than are species with lower precisions. The effective variance solution is derived10
by minimizing the weighted sums of the squares of  the differences between the
measured and calculated values of dk and a,-j. The solution algorithm is an itera-
tive procedure which calculates a new set of Sjk based on the Sjk  estimated from
the previous iteration.

   Watson12 found that  individual sources with similar source profiles would yield
unreliable values if included in the same chemical mass balance.  Henry13 proposed
a quantitative method of identifying this interference between  this similar source
compositions, which  is known as  "collinearity."-  He uses the "singular value de-
composition" define an "estimable space into which resolvable sources should lie."
The-sources which do not fall into this estimable space are collinear, or too similar
to be resolved from the sources which do lie within the estimable space.

   Williamson and Dubose14 claimed that the ridge regression reduces colinear-
ities.  Henry13 tested  the ridge regression solution with respect to the separation
of urban and continental dust and found that the bias resulted in physically un-
realistic negative values  for several of the a,-j. The ridge regression solution .has
not been used in the  CMB  since these  tests  were published.

CMB Model Assumptions

   The CMB model  assumptions are:

    •  Compositions of source emissions are  constant over the period of ambient
      and source sampling.

    •  Chemical species do not react with each other, i.e., they add linearly.
                                   9  -7

   • All sources with a potential for significantly contributing to the receptor
     have been identified and have had their emissions characterized.

   • The sources1 compositions are linearly independent of each other.

   • The number of sources or source  categories is  less  than or equal to the
     number of chemical species.

   • Measurement  uncertainties are random, uncorrelated,  and normally. dis-
         Effects of Deviations from CMB Model Assumptions

   Assumptions 1 through 6 for the CMB model are fairly restrictive and will
never be totally complied within actual practice. Fortunately, the CMB model can
tolerate reasonable  deviations from these  assumptions, though these deviations
increase the stated uncertainties of the source contribution estimates.

   The CMB model has been subjected to a number of tests to determine its abil-
ities to tolerate  deviations from model assumptions.3-:2'13' "•16-17-1Sl 19' 20- 21- 22
These studies  all point to the same basic  conclusions regarding deviations from-
the above-stated assumptions.

   With regard to Assumption 1, source compositions, as seen at the receptor, are
known to vary substantially among sources, and even within a single  source over
an extended period of time.  These variations are both systematic and  random
and are caused by three phenomena:  1) transformation and deposition between
the emission point  and the receptor; 2) differences in fuel type  and operating
processes between similar sources or the same source in time; and 3) uncertain-
ties or differences between the source profile measurement methods.  Evaluation
studies have generally compared CMB results from several tests using randomly
perturbed input data and from substitutions of different source profiles for the
same source type. The general conclusions drawn from these tests are:
   • The error in the estimated source contributions due to biases in all of the
     elements of a source profile is in direct proportion to the magnitude of the

   • For random errors, the magnitude of the source contribution errors decreases
     as the number of components increases.

   The most recent and systematic tests are those of Javitz22 which apply to a
simple four-source urban airshed and a complex ten-source urban airshed. These
tes^s, with 17 commonly measured chemical species, showed that primary mobile,
geological, coal-fired power plant, and vegetative burning source types can be
apportioned with uncertainties of approximately 30% when coefficients of variation
in the source profiles are as high as 50%. This performance was demonstrated even
without the presence of unique "tracer" species such as selenium for coal-fired
power plants or soluble potassium for vegetative burning.  In a complex urban
airshed, which added residual oil combustion, marine aerosol, steel production,
lead smelting, municipal incineration, and a continental background aerosol, it was
found that the geological, coal-fired power plant, and background source profiles
were collinear with the measured species. At coefficients of variation in the source
profiles as low as 25%, average absolute errors were on the order of 60%, 50%, and
130% for  the geological, coal-burning, and background sources, respectively. All
other sources were apportioned with average absolute errors of approximately 30%
even when coefficients of variation in the source profiles reached 50%.  Once again,
these tests were performed with commonly measured chemical species, and results
would improve with a greater number of species which are specifically emitted by
the different source types.

   With regard to the nonlinear summation of species, Assumption 2, no studies
have been performed to evaluate deviations from this assumption.  While these
deviations are generally assumed to  be  small, conversion of gases  to particles
and  reactions between particles are not inherently linear processes. This assump-
tion is especially applicable to the end products of photochemical reactions and
their apportionment to the sources of the precursors. Further model evaluation is
necessary to determine the tolerance of the  CMB model to deviations from this
assumption. The current practice is to apportion the primary material which has
not  changed between source and receptor.  The remaining quantities of reactive
species such as ammonium, nitrate, sulfate, and elemental-carbon are then appor-
tioned to chemical compounds rather than directly to sources.  While this approach
is not as satisfying as a direct apportionment, it at least separates primary from
secondary emitters and the types of  compounds apportioned give some insight
into the chemical pathways which  formed them.  As chemical reaction mecha-
nisms and rates, deposition velocities, atmospheric equilibrium, and methods to
estimate transport and aging  time become better developed, it may be possible
to produce  "fractionated" source profiles which will allow this direct attribution
or reactive species to sources.  Such apportionment will require measurements of
gaseous as well as participate  species  at receptor sites.

    A major challenge to the application of the CMB is the identification of the
primary contributing sources for inclusion in the model, Assumption 3. Watson12


systematically increased the number of sources contributing to his simulated data
from four to eight contributors while solving the CMB equations assuming only
four sources. He also included more sources in the least squares solutions than
those which were actually contributors, with the following results:

    •  Underestimating the number of sources had little effect on the calculated
      source contributions if the prominent species contributed by the missing
      sources were excluded from the solution.

    •  When the number  of sources  was underestimated, and  when  prominent
      species of the omitted sources were included in the calculation of  source
      contributions, the contributions of sources with properties in common with
      the omitted sources  were overestimated.

    •  When source types actually present were excluded from the solution, ratios
      of calculated to measured concentrations were often outside of the 0.5 to 2.0
      range, and the sum of the source contributions was much less than the total
      measured mass. The low calculated/measured ratios indicated which source
      compositions should be included.

    •  When the number of sources was overestimated,  the sources not actually
      present yielded contributions less than their standard errors  if their source
      profiles were significantly distinct from  those of other sources.  The over-
      specification of sources decreased the standard errors of the source  contri-
      bution estimates.

    Recent research suggests that Assumption 3 should be restated to specify that
source contributions above detection limits  should be included in the CMB. At
this time, however, it is not yet possible to  determine the "detection  limit" of a
source contribution at a receptor since this is a complicated and unknown function
of the other source contributions, the source  composition uncertainties and the
uncertainties of the receptor measurements.  Additional model testing is needed
to define this "detection limit."

    The  linear  independence of source compositions required by  Assumption 4
has become a subject of considerable interest since the publication of Henry's13
singular  value decomposition  (SVD) analysis. As previously noted, this analysis
provides quantitative measures of collinearity and the sensitivity of CMB results
to specific receptor concentrations. These measures can be calculated analytically
in each application. Henry13 also proposed an optimal linear combination of source
contributions that have been determined to be collinear.

   Other "regression diagnostics" have been summarized by Belsley23 and have
been applied to the CMB by DeCesar.19- 20  Kim and Henry24 show that  most of
these diagnostics are useless because they are based on the assumption of zero
uncertainty in the source profiles. They demonstrate, through the examination of
randomly perturbed model input data, that the values for these diagnostics vary
substantially with typical random changes in the source profiles.

   Tests performed on simulated data with obviously collinear source composi-
tions typically result in positive and negative values for the collinear source types
as well as large standard errors on the collinear source contribution estimates. Un-
less the source compositions are nearly identical, the sum of these large positive
and negative values very closely approximates the sum of the true contributions.

   With most commonly measured species (e.g., ions, elements and carbon) and
source types (e.g., motor vehicle, geological, residual oil, sea salt, steel production,
wood burning and various industrial processes), from five to seven sources  are
linearly independent of each other in most cases.22

   Gordon15 found instabilities in the ordinary weighted least square solutions to
the CMB equations  when species presumed to be  "unique"  to  a certain source
type were removed from the solution.  Using simulated data with known pertur-
bations ranging from 0 to 20 percent, Watson12 found: "In the presence of likely
uncertainties, sources such as urban dust and continental background dust cannot
be adequately resolved by least  squares fitting,  even though their compositions are
not identical. Several nearly unique ratios must exist for good separation."

   With regard to Assumption 5, the true number of individual sources contribut-
ing to receptor concentrations is generally much larger than the number of species
that can be measured. It is therefore necessary to group sources into source types
of similar compositions so that this assumption is met. For the most commonly
measured species, meeting Assumption 4  practically defines these groupings.

   With respect to Assumption 6 (the randomness, normality, and the uncorre-
lated nature of measurement uncertainties), there  are no  results available from
verification or evaluation studies. Every least  squares solution to the CMB equa-
tions requires this assumption,  as demonstrated by the derivation of Watson.11 In
reality, very little is known about the distribution of errors for the source compo-
sitions and the ambient concentrations. If anything, the distribution probably, fol-
lows a log-normal rather than a normal distribution. Ambient concentrations can
never be negative, and a normal distribution  allows for a substantial proportion
of negative values, while a log-normal  distribution allows no negative values. For
small errors  (e.g., less than 20%), the  actual distribution may not be important,
but for large errors, it probably is important.  A symmetric distribution becomes

                                  9 -11

less probable as the coefficient of variation of the measurement increases.  This
is one of the most important assumptions of the solution method that requires

                     Model Input and Output Data

   The chemical mass balance modeling procedure requires:  1) identification of
the contributing  sources types; 2) selection of chemical species to be included;
3) estimation of the fraction of each of the  chemical species  which is contained
in each source, i.e., the source compositions); 4) estimation  of the uncertainty
in both ambient concentrations and source compositions; and 5) solution of the
chemical mass balance equations, and  6) validation and reconciliation. Each of
these steps requires different types of data.

   Emissions inventories are examined to determine the types of sources which
axe most likely to influence a receptor.  Principal components  analysis applied to
a time series of chemical measurements  is also a useful method of determining the
number and types of sources. After these sources have been identified, profiles
acquired from similar sources25 (identify most of the available  source profiles) are
examined to select the chemical species to be measured. Watson12 demonstrates
that the more species measured, the better the precision of the CMB apportion-

   The ambient concentrations of these species,  C,-, and their fractional amount
in each source-type emission, F,j, are the measured quantities  which serve as
CMB model input data. These values require uncertainty estimates, era and crpij.
which are also input data.  Input  data  uncertainties are used  both to weight the
importance  of input data values in the solution and to calculate the uncertainties
of the  source contributions. The output consists of: 1) the source contribution
estimates (Sj) of each source type; 2) the standard errors, of these source  contri-
bution estimates. 3) the amount contributed by each source type to each chemical

TMBR Model

   The TMBR model is a multiple regression based model which may be used to
apportion an aerosol species of interest  measured at a receptor site to the  various
contributing sources. The actual regression analysis may be performed using the
method of ordinary least squares. However, since the independent  variables in
this  model are ambient concentrations of various aerosol components which are
measured with error, the method of Orthogonal Distance Regression (ODR) is ex-
                                 9  -12

pected to give better estimates of the source contributions.  A detailed theoretical
discussion of the method of ODR may be found in the book by Fuller (1987).26

                             Model  Equations

   In this  section it is shown that, under appropriate assumptions, the general
mass balance model can be reduced -to a simpler linear model. Let aerosol com-
ponent  i =  I be  a secondary aerosol with i* = 2 denoting the corresponding
parent species.  It is  of interest to determine the fractional contribution to the
ambient concentrations of this secondary aerosol component by  a distinguished
source which will  be  denoted by the subscript j — 1. We will also assume that
aerosol  species i\ is a tracer for this distinguished source. Let sources j = 2 thru
j =  ill have an  associated tracer species z'j, sources j — n2 + 1 thru j = n3 have
an associated tracer species i$ etc., and sources _;' =  n/,_i -f 1 thru j = rih have
an associated tracer i/,. Sources j = n^ + 1 thru j =  n may be unknown sources
or may be  known sources with tracers  that are not measured at the receptor. For
the sake of uniformity of notation we .let n! = 1. Thus the n sources have been
partitioned into h + 1  groups, each of the first h groups of sources being associated
with a unique tracer specfes or with a fraction of some reference species that has
been calculated using CMS or some other appropriate mode!.

   In general for 1 < u < h and nu_i + 1 < j < nv we have
                          cljk=        0ukCiujk = piukcittk             (is)
where /3,-u^- is defined as
 For n/j + 1 < j < n let
 The general mass balance equation then reduces to the equation
                               = Att +   jSi.tC.-.fc                       (21)

 for each sampling period k = 1, 2, . . . , s.


   If the quantities /?,-„* are all independent of k for each u, /?,•„* = /#{„, and the
above set of equations reduce to
The quantities C,-ujt are ambient concentrations of the tracer species z'i, 12, ..., ih
and are assumed known.  The quantities C\k are the ambient concentrations of
the aerosol species being apportioned and are also assumed known. We thus have
a set of 3 linear equations in h + 1 unknowns 0o, /?;,, /3,-2, ..., /3,-h. If the system of
equations has rank h + 1,  then these unknown beta coefficients may be obtained
by solving the above system of linear equations.  The apportionment of the species
of interest to the various groups of sources is then carried out by calculating  the
individual terms of the equations above.

    In certain instances it is known that the beta coefficients will differ significantly
from one time period to another.  In such cases it may be possible to determine,
based on physical and chemical reasons, a function of the field measurements,  the
sampling period and  the source, which we denote by jk, such .that it is more
reasonable to assume the quantities /?,-„*/<£,-£  are constant for all sampling periods
rather than the quantities $„*. In  such cases we  define 7;u  =  0iuk/jk .   For.
uniformity of notation we define 70 to be equal to /?o- This results in the system
of linear equations


We may refer to  this  set of equations as the TMBR model.  Again, if this set
of equations has  rank h + 1 then  we may solve for  the gamma coefficients and
consequently calculate the individual terms  of the equations. This will yield the
apportionment we seek. Note that if we take  4>jk = 1 then this set of equations
reduces to the set of equations in (22).

Tracer Mass Balance (TMB) Model

    This is a special case  of the TMBR model.and'is  obtained  by  partitioning
the sources contributing a particular secondary aerosol species, (say species i  = 1"
with associated parent species designated as species i* = 2), into two groups rather
than h + 1 groups. That is, we take  h = 1 in the  TMBR model. The two groups
are: (i) A distinguished source labeled j = 1 with associated tracer species i — i'i,
and (ii) All other sources.  In this case, the TMBR model reduces to

                                 = fok + AjfcC,-,*                         (24)

                                 9  -14

As before , if we assume that the beta coefficients are independent of the sampling
period, then the TMB model equations further reduce to
                                 = fa + faCM                          (25)
If the quantities Cu- and Cilk are known, and if the set of linear equations in (32)
have rank 2 then we can solve for the unknown beta coefficients and consequently
carry out the apportionment of species 1  by  computing the individual terms of
the above equations.

   In certain instances it is known that the beta coefficients will differ significantly
from one time period to another.  In such  cases  it may be  possible to determine,
based on physical and chemical reasons, a function of the field measurements, the
sampling period and  the.  source,  which we denote by ^u-, such that  it is more
reasonable to assume the quantities P^k/Pik are constant for all sampling periods
rather than the quantities /3,-jt.  In such  cases we define 7,^ .= fi^k/fak •  For
uniformity of notation we define 70 to be equal  to fa.  This results in the system
of linear equations
                            Cik - 7o + 7i-jC;,i:<£u                        (26)
We may refer to the above system of equations as the TMB model. Again, if
this set of equations has rank 2, then we may solve for the gamma coefficients and
consequently calculate the individual terms of the equations.  This will yield the •
apportionment we seek.

                              A  Special Case

   The simplest versions  of  the TMBR model  use uk  = 1 for all time periods
and source groups. However, if Ke or K& are dependent on other variables such as
solar radiation,  concentration of key atmospheric chemicals and so forth, it may
be possible to chose a form of ^  that will linearize the TMBR model.

   In apportioning a secondary aerosol, the constant 0iujk  derived from the GMB
model had the form

      T*..  —  	—	'—	'	;	;	 X

                            , k)tik) - exp(-(Kc(i'J, k) + Kd(i-J, k)}tjk)}   (28)


                                    iv,j,k} + Kd(iu,j,k})tjk)             (29)


If the species iu does not convert and its deposition rate is the same as that of
the secondary aerosol species i being apportioned, then

                        I-HJ*  =  exp(-Kd(i.j,k)tjk)                     (30)

so that the ratio r^k/riujk reduces to Kc(i",j, k)tjk after using the approximation

                 exp(x) « 1 + x (when x is sufficiently small).              (31)

The full infinite series expansion for exp(x) is given  by

                                        x2   x3
                       exp(x) = l + x + — + — + --

and a first order approximation has been used in (31). It is possible to use higher
order approximations of exp(x) in these derivations  but this is not pursued  here.

    An an example of the above approximation consider a case where Kc(i',j^ k) is
proportional to RHuk with proportionality constant Bi-j. Then the ratio r"jfc/r,-uj-fc
is equal to Bi-jtjkRHuk which gives

                          AujJt = Bi-jtjkRHuk — — •                       (32)

                             7.\,* = /3i,k/RHuk                           (33)
and assuming that 7,-,,* are constant for all sampling periods rather than the
quantities fruk suggests the use of RHuk as a linear factor in the TMBR model
equation (23).
    The use of RH as a linearization parameter does not necessarily imply that
the RH dependence of Kc is grounded in some basic chemical process. Rather, in
the case of 502  to SO* oxidation,  RH may be thought of as a surrogate variable
depicting the amount of time that SO? spends  in contact with clouds where ox-
idation is accelerated. Therefore, assuming RHuk — RHk, the TMBR model for
the 502 - SO 4 system becomes
     Csouk are
all observed with error. We shall denote the true  values by Cut, C,-0jt  and uk and
the observed values by the quantities Cut,  C,-ujt and 4>uk- We then assume that

                             Cik  =  Cik + f-c

The quantity f.clk is a random error with mean 0 and standard deviation
The quantity €.cl k  is a random error with mean 0 and standard deviation (?civk-
Likewise, the quantity e^ is a random error with mean 0 and standard deviation
From this we obtain the estimated fractional contribution  Fujt of species 1 by
source group u for sampling period k as
The estimated fractional contribution for the entire sampling period, by source
group u, is denoted by Fv and is calculated as
   To calculate the uncertainties to be associated with these estimates we may
use the following procedure. We construct several (say, 100) synthetic data sets
by perturbing the estimates of the true values Cut, C^k and  $uk using gaussian
random deviates with mean zero and standard deviations equal to the respective
measurement uncertainties. Each such synthetic data set is subjected to an ODR
analysis to obtain estimates of contributions  and fractional contributions of the
various source groups to the receptor as explained above. This procedure  results
in a whole collection of estimates '(say, 100) for the various quantities of interest.
The root mean square error is then calculated for each quantity of interest using
the collection of estimates obtained from perturbed synthetic data sets  and using
the" initial estimates obtained from the actual data set as if they were the true
values. This root mean square error associated with a given quantity of interest
is used to quantify the uncertainty associated  with that quantity.  Recall that if 0
represents the true value of a quantity and 0* represents an estimate of 6 obtained
from the qtk  synthetic data  set, then the root  mean square error is calculated by
                 Root Mean Square Error =
Alternatively we may quantify the uncertainty associated with a given estimate
using confidence intervals but we do not discuss that approach here.

    Second Approach.    In this section we discuss an approximate method of
calculating the uncertainties associated with the model outputs. The concentra-
tions (7iujt of species 1 (secondary species  of interest) associated with each trace
element iu for each  time period may be calculated by  multiplying the measured
values of Aivk — Ctakuk for each trace element by the respective estimated regres-
sion coefficients as follows. (•% would just be the estimated intercept representing
the estimated contribution from all sources not explicitly accounted for by any of
the reference species used in the TMBR model.)


                              C'luk = 7i.  x A,uk                          (35)

   The uncertainties for each of these concentrations C~uk may be calculated by:
The quantities aA-uk are the uncertainties in the measured values Aiuk and  is
assumed to be known. In the special case discussed in the previous section, these
uncertainties  are part of the WHITEX data base.  The quantities 7,-u may be
obtained as outputs from the regression packages that are used.  Errors in A,-,,*
and the estimated regression coefficients have been assumed to be independent in
the calculation of Equation (36).

    The total calculated amount of species 1, Cik for each time period is the sum
of the C*ujt's summed over all the reference aerosol species iu and the intercept 70-

                             Cik = 7o + £ C'iuk                         (37)
The uncertainty associated with the total calculated concentration of species 1 for
each time period is:
                                        + £                             (38)
assuming the covariance terms arising in the derivation are negligible.

    The estimated fraction Fuk of species 1 from each source for  any given time
period is equal to the amount of species 1 associated with the trace element divided
by  the total  calculated concentration of species 1:

The uncertainty for each of these fractions is:

                                    £Lt + C&gc»                      (40)
                                    (~*1        f**^

The mean fraction Fu of species 1 attributed to each source type is estimated by
the mean species 1 concentration Cu for that source type divided by the mean
total calculated concentration of species 1, C, as follows:

                                  JU = S±                              (41)



   The uncertainties for Cu and C are calculated by:
                                   1   *i
   The uncertainties associated with the mean fractions Fu are calculated by
   The uncertainty formulas are all derived using propagation of error methods
and  assuming the covariances between various terms occurring in the derivation
are negligible. These assumptions will not be true in practice and so the usefulness
of the above approximations will depend upon how severely the assumptions used
in the above  derivations are violated.

                           Model Assumptions

   The TMBR model assumptions are:

   • The chemical species used as tracers in the model are assumed to be uniquely
     emitted by non-overlapping groups of sources.  In particular none  of the
     species other than the tracer associated  with the source of interest can be
     emitted by another source unless there is an independent method such as
     CMB to partition the ambient species concentrations into components at-
     tributable to the various groups of sources.

   • The composition of source emissions are constant over the period of ambient
                                 9 -20

   • Deposition and conversion are constant from one sampling period to the
     next for each subgroup it.

   • Measurement errors are random, uncorrelated, and normally distributed.

   For the special case where kc was assumed to be proportional to RH the addi-
tional assumptions are:

   • Exponential forms of deposition and conversions can be represented by first
     order approximations.

   • The RH at the receptor site  is indicative of the amount of time that air
     parcels spend  in contact with clouds and therefore can be used as an  indi-
     cator of oxidation rate.
                 Potential Deviations from Assumptions

   It is highly unlikely that deposition and conversion are constant in space and
time and in many cases one can expect source profiles to change over the course
of the study. These assumptions are implicit to the assumption that background
and fractionation coefficients are time independent. Whether of not a lineariza-
tion  scheme is appropriate can be  examined through goodness of fit tests of the
proposed model and possible by direct experimental verification. The uniqueness
of tracer species can be assessed by source testing and by releasing unique tracers
from sources of interest.

   Deviation from any of the assumptions will increase the calculated uncertainty
in the final apportionments.  The extent to which the inflation of uncertainty
occurs will depend on how variable the regression  coefficients are. Research into
the effect of deviation from assumptions on apportionments is needed.

                               Model Inputs

   The model requires the following quantities as inputs:

   •  The ambient concentrations of the aerosol species  being apportioned.

   •  The ambient concentrations of the reference tracer species.

   •  Relative humidity  at the receptor for each of the sampling periods, when
   • The uncertainties in the above quantities, when-ODR is used to estimate
     the 7 coefficients, rather than OLS.

                             Model Outputs

   The model outputs include:

   • Estimates of the actual amount of the contribution and the fractional con-
     tribution of the aerosol species of interest by the source or source type of
     interest to the receptor, along with the associated uncertainty estimates.

   • Estimates of the average amount and the average fractional amount of the
     aerosol species of interest contributed by each source or source type  of in-
     terest along with the associated uncertainty estimates.

Differential Mass Balance  (DMB) Model

   The DMB model is a receptor model combined with elements of a deterministic
model. In this approach dispersion is accounted for by ratioing ambient trace ma- •
terial concentrations attributed to a source by known, trace material release rates
while deposition and conversion are explicitly calculated. The name "Differential
Mass Balance" refers to the use of difference  in trace  material concentration to
account for dispersion.

                             Model Equations

   Suppose a particular source is of interest and we wish to determine  the frac-
tional contribution of some aerosol species  to the receptor by that  source.  We
shall designate the aerosol species of interest by the subscript ii and the source
of interest by j. If species i is a secondary species then the corresponding parent
species will be denoted by the subscript i". For example, if 5C?4 is of interest, then
i stands  for SO '4 and i* stands for SOi. We are then  interested in the  quantity
djk for each of the sampling periods.  We have, from Equation (2) that

                                      jk + Ci'jkr'jkdjk _                   (47)
If i represents a primary species, then r"jk is zero for all  k.  If i represents a
secondary aerosol species that is not emitted  as a primary  aerosol, the quantity
    is zero for all k.  Therefore, the above equation simplifies to

                              Cijk — CijkTijkdjk                           ' ' 8)

                                 9  -22

when i is a primary species and

                                           kdjk                          (49)
when i is a secondary species. A characteristic feature of DMB model applications
is that the dispersion  factor djk is determined based on field measurements.  If a
unique tracer is available for source j then djk may be calculated based on  this
unique tracer.  It can also be calculated based on a reference aerosol species that
may not be a  unique  tracer for source j by first calculating the amount of  this
reference species contributed to the receptor  by the source of interest. Chemical
mass balance model may be applied for this purpose. Other approaches are  also

    The following discussion assumes that a unique tracer is available for source
j of interest. This source will be referred to as St. The  tracer material may be a
naturally emitted primary aerosol species or  may be introduced artificially.  The
aerosol species is denoted by the subscript iQ.  Therefore Equation (48) becomes

                             Cigjk = Ci0jkri0jkdjk.                          (50)

Dividing the quantity djk by the quantity Ci0jk we get,

                               yijk    Cijk Tjjk
when species i is a primary aerosol and

                               &j'fc _ fr-jfc r»jfc                           (52)
 when species i is a secondary aerosol. It follows from this that

                                         -^—Ciajk                         (53)
 for primary aerosols i and

 for secondary aerosols.

    Since aerosol component io is a tracer for source j, the quantity C,-OJ-fc is the
 same as the quantity C;0jt which is the ambient concentration of species i0 at the
 receptor and can be measured. If furthermore the quantities /Q(i, j, /:), Kc(i,j, k)
 are known when species i is primary, or, Kc(i*,j,k),  Kj(i',j,k)  and Kj(i,j, k)


are known when species i is secondary, and if in addition, /C*(*o, J, &), Kc(io,j, k),
tjk as well as the ratio c,-.jfc/c;0jjt are known, then the contribution of the source
of interest to the concentrations of the species of interest at the receptor can, in
principal, be calculated.

   If T represents a unique nonconverting, non depositing tracer for source j ' = 1,
then for a species that is  directly emitted by source j =  1, Equation (53)  for
primary aerosols reduces to

                             Cuk = - rilkCT,k-                        (55)

If the ratio Ciik/CT,i,k is known, the form of rm, is
For a species that is not directly emitted, but is a secondary species  which is
absent at the source, the equation for the DMB reduces to

                             Cuk = — — r'ikCr,k-                         (56)

The ratio Q.u/cr.i,* is assumed known and the form of r*lk in this case  is
                  il*   Ke(i; 1, *) + Kd(i; 1, k) - Kd(i, 1, fc)
          {exp(-Kd(i, 1, k)tlk) - exp(-[Ke(i-, 1, k) + Kd(i', 1

 A%(z, 1, k}  =  conversion rate of species z from source 1 to its secondary form,
                during sampling period k.
 Kd(i, 15 k)  =  deposition rate of species : from source 1 during sampling period
Considering a specific example for SO^ and SO^ Equation (56) becomes
                         r      _    a ,._,     r                       ,. >
                         0504,1,* - — - Tso*,\,k^T,k                     COY)
                                           2, 1, Jb) - Kd(SO<, 1, A:)


    {exp(-Kd(S04, 1, k)tlk) - exp(-[Ke(SOz, 1, fc) + Kd(S02, 1, k)]tlk)}    (59)
              rso,,i,* = ezp(-(tfc(502, 1, fc) + ^(502, 1, fc))iu).          (60)
From now on  we shall use the notation A'c = #c(502, 1, Jt), /^ = Kd(SO2, 1, Jfc)
and /
may be fitted and the adequacy of the fit judged by the resulting R2 value and the
closeness of the beta coefficients to one. Coo .- t refers to the total contribution
of SO* by source group t: to the receptor. If the chosen parameter combination
results in a high R2 value and beta values are not significantly different from one,
then the chosen parameter  values v\,vi,Kc may be judged  as being consistent
with observed data. The best possible value of R2 obtained, by varying the values
of ui,i>2 and A'c over their entire range of values suggested in the  literature, may
be denoted by R\pt. The values v\ =  ui,opt, ^2 = V2,opt,  and Kc  = KCt0pt which
result in the best R2 may be used  to calculate the daily St contributions to SO*
and  502 at the receptor. By calculating the ratio of the total St contribution
over the entire sampling period to total ambient concentrations over the same
period we can calculate the fractional  504 and 50j contributions by  5t during
the experimental period.

                        Uncertainty Calculations

   Uncertainties in the final results are primarily due to three sources.

   • Uncertainties in tjk.

   • Uncertainties in the model parameters such as Kc> K\ and K^.

   -• Uncertainties in the measured values.

   • Uncertainties in the extent to which the model assumptions are violated.

   Uncertainties in  the  Model Parameters.    The model parameters in
question are Ke, K\ and KI which are not  known.  Suppose  a review of the
literature suggests deposition velocities v\ for 502 ranging from  l\ to ui cm/sec
and  U2  for 504 ranging from /2 to u^ cm/sec.  In addition suppose  the  sulfur
dioxide oxidation rates varied from Kc = lc to Kc = uc percent per hour.

   Clearly, not all combinations of values of Ui,t>2 and Kc are physically possible.
To judge which combinations of these parameters are reasonable, the following
procedure may be adopted.  A grid of  values for fi,U2 and Ke may be chosen by
taking all possible combinations of these parameters resulting from

                       i'i = /i  to Ui in increments of S\.

                       t'2 = /2 to u-i in increments of <52.
                       A'c = le to ue in increments of 6e.


To  decide whether a particular combination of values of ul5u2 and Kc are rea-
sonable the regression model suggested by Equation 63 can be exercised and the
adequacy of the fit may be judged by closeness of beta values to one and the re-
sulting R2. The best possible value of R* for /? values close to one over the range
of these parameters is denoted by R2opt.  A value  E% less than R2opt but close to
it is chosen, based on subjective judgement, as a criterion value for judging the
reasonableness of various combinations of the parameter values. Parameter com-
binations resulting in an R* equal to  R% or greater may be considered reasonable.
The set of all  such parameter combinations will be denoted by the symbol  A. St
contributions  can be calculated1 for each of the parameter combinations in the set
A . This will  result in a whole  range of values for the daily St contributions and
the overall average St contributions. The mean and the standard deviation for this
range of values (as .well as the minimum and the maximum values) may be  calcu-
lated to assess the uncertainty in the estimated St contributions due to imprecise
knowledge of  the  model parameters. The measured values of concentrations of
species are assumed to be exact in these calculations.
                           .   ' I
    Uncertainties in the  Measured Values.   To assess the effect of errors in
measurements on the estimated &  contributions to SOi and SOi at the receptor,
the values of Vi,v-t and Kc are fix^ed at their optimum values obtained as explained
in the previous subsection. The measured values used in the calculations are: (1)
The ambient T concentration,  Cx,k,  (2) The ambient  SO^ concentration Cso<,k,
(3)  -The ambient SO? concentration  Cso3,Jt) (4) Relative Humidity RHk  at the
receptor, and  (5) Transport time tst,k for the aerosol mixture from St to arrive at
the receptor.  Suppose each pf  these  measurements have associated with them a
standard deviation characterizing the uncertainty in the respective measurements.
We generate a number of synthetic data sets (one hundred is sufficient for most
purposes) on the computer by perturbing the measured values using random gaus-
sian deviates with zerd means and standard deviations associated with each of the
measured values. For each  synthetic  data set thus generated, the daily St  contri-
bution to 504 and 502 at the  receptor as well ,as the average contributions over
the entire sampling period are  'calculated. The range of values thus obtained for
each of these quantities gives an indication of the uncertainty that would  be due
to imprecise measurements alone.  The results are reported in the form of  means
and standard  deviations of each of the quantities of interest calculated from the
synthetic data sets.  Throughout- this exercise, the model parameters, viz, -the
conversion and deposition parameters, are to be kept  constant at  their optimum

    Uncertainties in the  Extent  to which the Model Assumptions are
Violated.  Assessment of the uncertainties in reported results arising from model
                                  9  -27

violations can be evaluated by conducting extensive sensitivity studies involving
various perturbations in the model assumptions themselves.

   Overall Uncertainties.    Since the first two categories of uncertainties are
expected to be "independent", the total uncertainty  due to these two sources may
be characterized by the effective total standard deviation

                            Total = (*\ +


   A set of deterministic general mass balance (GMB) equations describing how
primary and secondary aerosols and gases  are transported and  transformed as
they pass through the atmosphere were formulated. From the GMB equations it
is possible, with a variety of limiting assumptions, to derive the chemical mass
balance, the differential mass balance equations, and the tracer mass balance  re-
gression  model.  Derivation  of these receptor modeling approaches from  a first
principle model allows  for an examination of model assumptions, and deviations
from assumptions. With assumptions identified it possible to make a better de-
termination of how to incorporate measurement uncertainty and how to estimate
model uncertainty associated with an imperfect knowledge of model parameters.

   1. J.G. Watson, Overview of receptor model principles. JAPCA, 34, 620, 1984.

   2. J.G. Watson, J.C. Chow, D.L. Freeman, R.T. Egami, P.  Roberts  and R.
     Countess, Model and Data Base Description for California's Level I PM10
     Assessment Package. DRI Document 8066-002.1D1, Draft Report, Prepared.
     for the California Air Resources Board, Sacramento, CA, 1987.

   3. J.G. Watson, J.G.,  J.C. Chow and N.F.  Robinson,  Western States  Acid
     Deposition Project Phase I:  Volume 4~An  Evaluation of Ambient Aerosol
     Chemistry in the Western United States. Prepared for the Western States
     Acid Deposition Project by  Systems Applications, Inc., San  Rafael, CA,
     SYSAPP-87/064, 1987,

   4. J.C. Chow, Development of a Composite Modeling Approach to Assess Air
     Pollution Source/Receptor Relationships.  Doctor t>f Science Dissertation,
     Harvard University,  Cambridge, MA, 1985.

   5. P.K. Hopke, Receptor modeling in environmental chemistry. Chemical Anal-
     ysis, 76, John Wiley & Sons, New York, NY, 1985.

   6. R.K. Stevens, .C.W. Lewis,  Hybrid receptor modeling.  In: Extended Ab-
     stracts for the Fifth  Joint Conference on Applications of Air Pollution Me-
     teorology with APCA, November  18-21,  1986, Chapel Hill,. N.C. Published
     by the American Meteorological Society, Boston, Massachusetts, 1987.

   7. C.W. Lewis, R.K. Stevens, Hybrid receptor model for secondary sulfate from
     an SOi point source. Atmos. Environ. 19,6:917-924,  1985.

                                 9 -29

 8.  T.  Dzubay, R.K.  Stevens, G.E. Gordon,  I. Olmez,  A.E. Sheffield, W.J.
    Courtney,  A composite receptor method applied to  Philadelphia aerosol.
    Environ. Sci.  & Technol., 22, 1, 1988.

 9.  J.G. Watson, Transactions, Receptor Models in Air Resources Management,
    Air and Waste Management Assoc., Editor, Pittsburgh, PA, 1989.

10.  H.I. Britt,  and R.H. Luecke, 1973: The estimation of parameters in nonlin-
    ear, implicit models. Technometrics, 15, 233, 1973.

11. i J.G. Watson, J.A. Cooper and J.J. Huntzicker, The effective variance weight-
    ing for least squares calculations applied to the mass balance receptor model,
    Atmos. Environ., 18, 1347, 1984.

12.  J.G. Watson, Chemical Element Balance Receptor Model Methodology for
    Assessing  the Sources of Fine and  Total Particulate Matter. Ph.D. Disser-
    tation,  University  Microfilms International, Ann Arbor, MI, 1979.

13.  R|C. Henry, Stability analysis  of receptor models that use least squares
    fitting.  Receptor Models Applied to Contemporary Air Pollution Problems,
    Air Pollution Control Association, Pittsburgh, PA, 1982.

14.  H.J. Williamson, and D'.A. DuBose, 1983:  Receptor model technical series,
    volume III: User's  manual for chemical mass balance model. EPA-450/4-83-
    014, U.S. Environmental Protection Agency, Research Triangle  Park, NC,

15.  G.E. Gordon, W.H. Zoller, G.S. Kowalczyk and S.H. Rheingrover, Composi-
    tion of source components needed for aerosol receptor models. Atmospheric
    Aerosol: Source/Air Quality Relationships. Edited by  E.S. Macias and P.K.
    Hopke, American  Chemical Society Symposium Series #167, Washington,
    B.C., 1981.

16.  L.A. Currie, R.W. Gerlach, C.W. Lewis, W.D. Balfour, J.A. Cooper,- S.L.
    Dattner, R.T. DeCesar, G.E. Gordon, S.L. Heisler, P.K. Hopke, J.J.  Shah,
    G.D. Thurston and H.J. Williamson, Interlaboratory  comparison of source
    apportionment procedures: results for simulated data sets. Atmos.  Envi-
    ron., 18, 1517, 1984.

17.  T.G. Dzubay, R.K. Stevens, W.D.  Balfour, H.J.  Williamson, J.A. Cooper,
    J.E. Core, R.T. DeCesar, E.R. Crutcher, S.L.  Dattner, B.L. Davis, S.L.
    Heisler, J.J. Shah, P.K. Hopke and D.L. Johnson, Interlaboratory compar-
    ison of  receptor model results for Houston aerosol.  Atmos.  Environ., 18,
    1555, 1984.   •

18.  J.G.  Watson, and  N.F. Robinson, A method to determine accuracy and
    precision required of receptor model measurements. Quality Assurance in
    Air Pollution Measurements, Air Pollution Control Association, Pittsburg,
    PA, 1984.                                               '

19.  R.T. DeCesar, S.A. Edgerton, M.A.K. Kahlil and R.A. Rasmussen, Sensi-
    tivity analysis of mass balance receptor modeling: methyl  chloride as  an
    indicator of wood smoke. Chemosphere, 14, 1495, 1985.

20.  R.T. DeCesar, S.A. Edgerton, M.A. Khalil and R.A. Rasmussen, A tool for
    designing receptor model studies to apportion source impacts with specified
    precisions.  Receptor Methods for Source Apportionment: Real World Issues
    and Applications, Air Pollution Control Association, Pittsburgh, PA, 1986.

21.  H.S.  Javitz, J.G. Watson,  J.P. Guertin and P.K. Mueller, Results of a re^
    ceptor modeling feasibility study, JAPCA, 38, 661, 198S.

22.  H.S.  Javitz, J.G. Watson, and N. Robinson, Performance of the chemical
    mass balance model with simulated  local-scale aerosols, Atmos. Environ.,
    22, 2309, 19S8.

23.  D.A. Belsley, E.D.  Kuh and R.E. Welsch, Regression Diagnostics: Identify-
    ing Influential Data and Sources of Collinearity. John Wiley and Sons, New
    York, NY, 1980.

24.  B. Kim,  and R.C.  Henry, Analysis  of multicollinearity indicators and in-
    fluential  species for chemical mass balance receptor model,  Transactions,
    Receptor Models in Air Resources Management,  J.G. Watson, ed., Air and
    Waste Management Assoc., Pittsburgh, PA, 1989.

25.  J.C. Chow, and J.G. Watson, Summary of particulate data bases for recep-
    tor modeling in the United  States,  Transactions, Receptor Models in  Air
    Resources Management, J.G. Watson, ed., Air and Waste Management  As-
    soc.,  Pittsburgh, PA, 1989.

26.  W.A. Fuller, Measurement Error Models, John Wiley and Sons, New York,
    New York, 1987.

                         Project MOHAVE  Summary
Study Component
Description of Study Component
Purpose and
Purpose: Respond to Congressional mandate for "Mohave Power Plant tracer study."
Study Objectives: Estimate frequency and magnitude of any perceptible impact of
Mohave Power Plant to visibility at Class I areas; Estimate impacts of other sources
upon visibility in the southwest; Develop and evaluate tools for subsequent regional haze
Detailed intensive study periods nested within year-long study period. Results and
conclusions to be based upon evaluation and reconciliation of multiple analysis
Field Study: September 1991   November 1992
Winter Intensive: January 1992  (30 days)
Summer Intensive: July   August 1992 (50 days)
Draft Report: July 1993  Final Report: December 1993
Continuous stack release of perfluorocarbon tracer during intensives.  Monitoring with
35 samplers at 31 sites.  Release different tracers from southern California (Los Angeles
Basin and San Joaquin Valley) during summer intensive.
Continuous SOX and NOX stack monitoring during intensives.
Detailed source profiling using daily samples during intensives.
Air Quality
Full IMPROVE samplers at 10 sites, IMPROVE channel A + SO2 at 21 sites during
intensives (12 & 24 hour sampling). Sampling two days per week at  10 sites with
IMPROVE samplers during non-intensives.  DRUM sampling (8 size  ranges, 6 hour
resolution) at six sites during intensives.  Sampling with medium volume particle
samplers at three sites during intensives.  Hydrogen peroxide sampling for a portion of
summer intensive.
 Optical Monitoring
 Continuous monitoring entire study period.  Nephelometers at all receptor sites, a
 transmissometer at Meadview,  in addition to ones already at IMPROVE sites.  Time-
 lapse photography at several sites.
 Continuous vertical wind profiling for 12 months at two sites using radar wind profilers.
 Two additional profilers during intensives.  Surface meteorology at all wind profiler sites
 and receptor sites.  Doppler sodar at two sites.  RASS temperature profiling at two sites.
 Deterministic meteorological modeling (wind, turbulence, moisture, etc.) every day for
 12 month period.  Calculation of influence functions.  Detailed chemistry modeling
 (RADM, RPM) for selected cases. Monte Carlo transport modeling with linear
 chemistry every day for 12 month period.
 Data Interpretation
 Statistical study of historical sulfur concentrations and plant output.  Spatial pattern
 (eigenvector) analysis. Hybrid receptor modeling utilizing artificial and endemic tracer
 data.  Calculation of extinction budget.  Reconciliation of modeling results. Source
 Quality Assurance
 QA audit by independent reviewer covering all portions of the study.
 Potential SCE
 Upper air monitoring, particle monitoring (endemic tracers), chemical modeling, tracer
 release, data analysis, aircraft measurements, stack sampling, and data base

             O       Points of Reference
             O       Coal Fired Power Plants

             •       Receptor Sites
             •       Other Class I sites
                       Background Sites
             *       LA Basin Pass Sites
                     Low Elevation Transport
                     High Elevation Transport
                                                                                   ndian Gardens
                                                                            Hopi Point
                                                               w Truxton
                                                     Dolan Sprinfls
 * Tehachapi Summit
                          T Baker               \CTMohave  A Hulapai Mt. Park  T Seligman

                                                       » Yucca               Sycamore Canyon
                                             Needled                      A
                                                   \                      Cnr
                                                                                      Petrified Forest
            Cajon Summit
                                              Camp Wood
                                                  Prescott (airport)
  \               • San GorgonJo
        Angeles              • Joshua Tree
                                                                                       • Tonto
                                                                                 O Phoenix
                                    Monitoring Locations
Grid 1
         Number of Points   Spacing
  Grid    x     y     z    (kml
   1     100   60    44
   2     104   72    44
   3     144   144    44
   4     80    80    44
   5     80    80    44
                                  Meteorological Modeling Grids
           For more information,  contact Mark Green at (702) 798-2182.

                                          Begin returning
                                          tracer samplers
                           Finish returning
                           tracer samplers
                                                                       Begin analysis
                                                                          of tracer
                                          Complete tracer
                                           tracer data
                                           Ship tracer
                                           NEW YEAR'S
                                                          Ship tracer
               tracer and


Tuesday     Wednesday     Thursday
                               W»*nAVWp*'''"'.K:T ^yc"*;
                                                                                       . vv^fc •

                 Shaded days represent intensive sampling days,  (prepared \d\

Project MOHAVE Winter Intensive
September -
November -
11/20- 11/26
12/4 - 12/9 or
12/20, 12/23
1/6 - 1/11
1/11 7am MST
1/14 7am MST
1/14 7am MST -
2/13 7am MST
2/16 7am MST
2/20 7am MST
Begin year-round paniculate monitoring at
receptor and other Class I area sites;
Begin year-round optical monitoring
Install radar wind profilers/RASS
Deploy tracer samplers for background test
Pickup tracer samplers, return to Brookhaven
Analyze tracer samples
Distribute background tracer data
Assess readiness for field program; if OK, then
the following schedule will hold. If major
problems exist, re-evaluate study.
Ship tracer material to NOAA-Idaho Falls
Ship tracer samplers to Lake Mead for winter
Deploy tracer and paniculate samplers
Start tracer sampling
Start paniculate sampling
Release tracer
Stop paniculate sampling
Stop tracer sampling, except at Meadview and
Hopi Point
Air Resource
Idaho Falls,
          (schedule as of 12/17/91)