United States
Environmental Protection
Agency
Office of Air Quality
Planning and Standards
Research Triangle Park, NC 27711
EPA-454/R-95-006
December 1995
AIR
 & EPA
          INTERAGENCY WORKGROUP
     ON AIR QUALITY MODELING (IWAQM):
            ASSESSMENT OF PHASE 1
     RECOMMENDATIONS REGARDING THE
               USE OF MESOPUFF II

-------
INTERAGENCY WORKGROUP ON AIR QUALITY MODELING (IWAQM):
          ASSESSMENT OF PHASE 1 RECOMMENDATIONS
               REGARDING THE USE OF MESOPUFF H
               Office of Air Quality Planning and Standards
                  Technical Support Division (MD-14)
              Research Triangle Park, North Carolina 27711

                        National Park Service
                         Air Quality Division
                       Denver, Colorado 80225

                        USDA Forest Service
                         Office of Air Quality
                     Fort Collins, Colorado 80526

                     U.S. Fish and Wildlife Service
                          Air Quality Branch
                       Denver, Colorado 80225

                             April  1995

-------
                                 DISCLAIMER

The information in this document has been reviewed in its entirety by the U.S. Environmental
 Protection Agency (EPA), and approved for publication as an EPA document.  Mention of
    trade names, products, or services does not convey, and should not be interpreted as
           conveying official EPA approval, endorsement, or recommendation.

-------
                                     PREFACE

A memorandum of understanding (MOU) was developed by interested Federal agencies, viz.
the Environmental Protection Agency, the U.S. Forest Service, the National Park Service, and
 the U.S. Fish and Wildlife Service, to provide a focus for development of technically sound,
 regional air quality models for regulatory assessments of pollutant source impacts on Federal
   Class I areas. Based on this MOU the Interagency Workgroup on Air Quality Modeling
  (IWAQM) was formed.  Although no States are signatories, their participation  in IWAQM
                        functions is explicitly noted in the MOU.

 The Guideline on Air Quality Models (Code of Federal Regulations, Appendix W to Part 51)
 suggests MESOPUFF n may be considered on a case-by-case basis for use for assessing long
 range transport impacts.  In the interim recommendations (EPA, 1993), IWAQM outlined a
 manner in which MESOPUFF II might be applied in such instances. This report documents
 results from a case study to apply the MESOPUFF n air quality modeling system following
                        the IWAQM  interim recommendations.

     As will be seen in the discussion, limitations in resources necessitated only partial
 implementation of the interim recommendations. This was deemed acceptable because the
  purpose of the exercise was not to develop a meaningful assessment of actual air pollution
  impacts, but was to identify and summarize the decisions made; record and summarize the
  resolution process for these decision;  and provide a written record of the resources used to
  complete  the effort.  The contractor was given a free hand in suggesting and implementing
 strategies to automate processes and to accelerate the computations.  Acceptance of these or
   similar strategies in an actual assessment can only be addressed on a case-by-case basis
        involving the relevant review authorities in the context of an actual situation.

    IWAQM concludes that the findings confirm the need for active interaction between
applicant and all the reviewing authorities (EPA, State, Federal Land Managers), and that this
  interaction should occur as soon as is feasible.  Such interaction is needed to confirm the
  assessment endpoints of interest from the Federal Land Managers. Assessments involving
  Air Quality Related Values are evolving as experience is gained. It would be erroneous to
  believe that the previous report, EPA (1993), and this report provides all the information
   needed to define the modeling requirements. For instance, when IWAQM first issued its
  suggestions regarding regional haze assessment, (EPA, 1993),  one hour visibility impacts
 were considered appropriate by the Federal Land Managers participating in IWAQM. When
   IWAQM asked the contractor to perform the analyses summarized in this report, 3-hour
 visibility impacts were deemed appropriate. As of the release of this report, consideration is
 being given to 24-hour impacts as providing a better representation of regional haze impacts
                                than 3-hour impacts.

 The IWAQM members anticipate issuing additional publications related to progress toward
 meeting the IWAQM goals and objectives, the results of model evaluation studies, proposed
   and recommendations on modeling systems for regulatory applications, and other topics
                        related to specific objectives in the MOU.
                                         IV

-------
                            ACKNOWLEDGEMENTS

   The members of IWAQM acknowledge the special efforts of H. Andrew Gray, Mary P.
Ligocki and Christopher A. Emery of Systems Applications International in conducting the
analyses and summarizing the results presented in this report under EPA Contract No. 68-D-
30019, Work Assignment 2-94 with John S. Irwin as the Work Assignment Manager.

   The members of IWAQM acknowledge the peer review comments provided by several
separate groups. We found the comments to be insightful and helpful towards providing a
complete and understandable description of work. Recognizing the constraints of resources
and time, every effort was  made to address the comments received. Despite everyone's
assistance, however, some errors and inadequacies no doubt exist, which of course are
IWAQM's sole responsibility. Review comments were received from:

Utility Air Regulatory Group (UARG) comments received from Robert J. Paine and David
W. Heinold of ENSR Consulting & Engineering, commissioned by Lucinda Minton
Langworthy of Hunton & Williams.

Stephen F. Mueller of the Tennessee Valley Authority.

Northeastern States for Coordinated Air Use Management (NESCAUM) comments
coordinated by Paul Wishinski  of the Vermont Air Pollution  Control Division.

Southern Appalachian Mountains Initiative (SAMI) comments received from Kenneth L.
McBee of the Virginia Department of Environmental Quality.

Western States Air Resources (WESTAR) comments from Clint Bowman of Washington
Department of Ecology, Patrick Hanrahan of the Oregon Department of Environmental
Quality, and  Steven F. Weber of the North Dakota Division of Environmental Engineering.
                                        v

-------
                                 Contents


Acronyms and Abbreviations	      vi

1  INTRODUCTION	     1-1

      Study Objective 	     1-2
      Study Approach 	     1-3

2  REVIEW OF THE SCRAM BBS FILES AND EXAMPLE PROBLEM	     2-1

      Evaluation of BBS Files	     2-1
      Implementation of Improvements  	     2-4

3  DEMONSTRATION OF THE PHASE 1 MODELING SYSTEM	     3-1

      Modeling Domain Definition	     3-2
      Sources to Be Modeled	    3-10
      Modeling Episode Definition	    3-18
      Development of Meteorological Inputs  	    3-19
      Development of Background Air Quality Inputs	    3-33
      MESOPUFF II Application 	    3-35
      Data Analysis	    3-36
      Results  	    3-39

4  SOURCE IMPACTS AS A FUNCTION OF DISTANCE	     4-1

      Input Development	     4-1
      Application of MESOPUFF H for Ring Sources	     4-3
      Application of MESOFILE for Ring Sources  	     4-5
      Results from Ring Source Simulations  	     4-6
      Problems Encountered	    4-33
      Conclusions 	    4-33

5  SUMMARY	     5-1

      Status of the MESOPUFF H Modeling System	     5-1
      Demonstration Application  	     5-2
      Impacts as a Function of Distance  	     5-7

References 	    R-l

Appendix A: INTERAGENCY WORKGROUP FOR AIR QUALITY MODELING
Appendix B: BASELINE DETERMINATION
Appendix C: DESCRIPTION AND USERS GUIDE FOR NEW
          MESOPUFF IIPRE- AND POSTPROCESSORS
Appendix D: ISCST2 APPLICATION AND INTEGRATION

                                    vi

-------
                           Acronyms and Abbreviations
AIRS
AQRV
BBS
CD 144
DRI
EPA
FLM
IMPROVE
IWAQM
JRFW
km
MB
mb
MESOFILE
MESOPAC

MESOPUFF II

MHz
MM4
NCDC
NFS
NPS
OAQPS
PC
PM10

QA
SAI
SCRAM
SNP
TD-3240
TD-6200
UTM
VDEQ
WBAN
WRCC
Aerometric Information Retrieval System
Air Quality Related Value
Bulletin Board Service
standard 144 column format for surface weather observation data
Desert Research Institute (University of Nevada, Reno)
Environmental Protection Agency (U.S.)
Federal Land Manager
Interagency Monitoring of Protected Visual Environments
Interagency Workgroup on Air Quality Modeling
James River Face Wilderness
kilometer
mega-byte (computer disk storage)
millibar (atmospheric pressure)
postprocessing program for MESOPUFF II
meteorological preprocessing program for MESOPUFF U (the version
used is MESOPAC H)
mesoscale puff model; version II (all other program names in this
report are boldface)
megahertz
mesoscale meteorological model (four-dimensional data assimilation)
National Climatic Data Center
National Forest Service (U.S.)
National Park Service (U.S.)
Office of Air Quality Planning and Standards (U.S. EPA)
personal computer
particulate matter with aerodynamic diameters less than or equal to 10
microns
quality assurance
Systems Applications International (San Rafael, California)
Support Center for Regulatory Air Models
Shenandoah National Park
standard format for precipitation measurement data
standard format for upper air sounding data
Universal Transverse Mercator (coordinate projection)
Virginia Department of Environmental Quality
National Weather Service station identifier
Western Regional Climate Center (@DRI)
                                        vn

-------
List of Illustrations
Figure
3-1
3-2(a)
3-2(b)
3-2(c)
o o
J-J
3-4
4-1
4-2
4-3
4-4
4-5
4-6
4-7
4-8

The MESOPAC II meteorological modeling domain.
The MESOPUFF n computational domain.
Non-gridded receptors located within Shenandoah NP.
Non-gridded receptors located within James River Face W.
Display of the MESOPUFF U computational domain,
showing positions of all PSD sources compiled for the
MESOPUFF Phase 1 demonstration.
Location of Wampler-Longacre, source modeled with
ISCST2.
MESOPUFF computational domain for ring source analyses,
showing the location of the ring sources.
Highest simulated 3 -hour average SO2 (|ig/m3)
concentrations from ring sources for January, April, July,
and October, 1988.
Highest simulated 24-hour average SO2 (|ig/m3)
concentrations from ring sources for January, April, July,
and October, 1988.
Highest simulated monthly average SO2 (|ig/m3)
concentrations from ring sources for January, April, July,
and October, 1988.
Highest simulated 4-month average SO2 (|ig/m3)
concentrations from ring sources for 1988.
Highest simulated per-source 4-month average SO2 (|ig/m3)
concentrations from ring sources for 1988.
Highest simulated monthly average NOX (|ig/m3)
concentrations from ring sources for January, April, July,
and October, 1988.
Highest simulated 24-hour average PM10 (|ig/m3)
concentrations from ring sources for January, April, July,
and October, 1988.
Page
3-5
3-7
3-8
3-9
3-15
3-43
4-2
4-8
4-9
4-11
4-12
4-13
4-15
4-16
        Vlll

-------
List of Illustrations (Continued)
Figure
4-9
4-10
4-11
4-12
4-13
4-14
4-15
4-16
4-17
4-18
4-19
4-20
4-21
4-22

Highest simulated monthly average PM10 (|ig/m3)
concentrations from ring sources for January, April, July,
and October, 1988.
Seasonal effects for sulfate, nitrate and total PM10 monthly
average concentrations (|ig/m3) for the 125 km ring for 1988.
Highest simulated 4-month average PM10 concentrations
(|ig/m3) from ring sources for 1988.
Highest simulated per-source 4-month average PM10
concentrations (|ig/m3) from ring sources for 1988.
Highest simulated monthly cumulative total S deposition
(kg/hectare) from ring sources for January, April, July, and
October, 1988.
Components of simulated January 1988 total S deposition
from ring sources.
Components of simulated April 1988 total S deposition from
ring sources.
Components of simulated July 1988 total S deposition from
ring sources.
Components of simulated October 1988 total S deposition
from ring sources.
Highest simulated monthly cumulative total N deposition
(kg/hectare) from ring sources for January, April, July, and
October, 1988.
Components of simulated January 1988 total N deposition
from ring sources.
Components of simulated April 1988 total N deposition
from ring sources.
Components of simulated July 1988 total N deposition from
ring sources.
Components of simulated October 1988 total N deposition
from ring sources.
Page
4-17
4-19
4-20
4-21
4-23
4-24
4-25
4-26
4-27
4-28
4-29
4-30
4-31
4-32
              IX

-------
List of Tables
Table
3-1
3-2
o o
J-J
3-4
3-5
4-1
4-2
C-l

Original set of sources to be modeled in the MESOPUFF n
demonstration application for Shenandoah NP and James
River Face W.
Final condensed set of sources modeled with MESOPUFF U
for the SNP demonstration application.
Assumed correspondence between GIS land use categories
and MESOPUFF U land use categories.
PSD and AQRV parameters to be calculated from
MESOPUFF H output.
PSD and AQRV parameters calculated from MESOPUFF II
output.
Emission rates and stack parameters for idealized sources.
Source ring characteristics.
Comparison of array dimension parameters between
"original" and "demonstration" versions of the MESOPUFF
II modeling system.
Page
3-13
3-17
3-32
3-37
3-40
4-4
4-4
C-l

-------
                                  1 INTRODUCTION
   Recognizing the immediate need within the permitting community for evaluating the
   impacts of sources of air pollution located more than 50 kilometers from Class I
   wilderness areas and national parks, the Interagency Workgroup on Air Quality Modeling
   (IWAQM) drafted an interim Phase 1 recommendation from existing "off-the-shelf-
   techniques"  (EPA, 1993).  The interim approach recommended the use of the Lagrangian
   puff model,  MESOPUFF-II (Scire et al., 1984), to evaluate the impacts of pollutants from
   sources located more than 50 kilometers from Class I areas, and up to several hundred
   kilometers from Class I areas.  The impacts of concern are the allowable Prevention of
   Significant Deterioration (PSD) Class I increases in pollutants (increments), the National
   Ambient Air Quality Standards (NAAQS), and  the Air Quality Related Values (AQRVs).
   AQRV impacts include such effects as visibility degradation and acidic deposition.  The
   interim recommendation was envisioned as suitable for single source impact analyses, as
   well as for multiple source (cumulative) impact analyses.

   It is important to note that by restricting the modeling techniques to "off-the-shelf,"
   certain limitations were incurred. These include limits in considering the effects of
   terrain on long range transport  and dispersion, an underestimation of the conversion of
   sulfur dioxide to sulfate when polluted air interacts with clouds, and an overestimation of
   particulate nitrate when a limited number of sources are considered.  Furthermore, the
   estimations  of the impacts of sources on regional visibility are simplistic and do not
   account for  all of the processes important to regional visibility. Nonetheless, the
   IWAQM considers the techniques to provide a useful assessment of air quality impacts in
   Class I areas.

   In the course of developing the Phase 1 interim  recommendations, it was recognized that
   even if the modeling techniques could be agreed upon between the various federal
   agencies, there would still remain both technical and policy issues in implementing an
   assessment of pollutant impacts on  a Class I area. In general, the fact that Class I area
   analyses focus on a fixed piece of real estate sets them apart from Class II analyses.  In
   Class n analyses, the area  of concern is a circular area (typically of radius 50 kilometers
   or less) centered on the source  in question; whereas in Class I analyses the source and the
   Class I area  are usually separated by a significant distance.  This occasions a unique set of
   issues which need consideration from and cooperation among a variety of organizations,
   given the need to assess the incremental increase in concentration values, the need to
   evaluate AQRVs, and the added role of the Federal Land Mangers (FLMs).
94030rl.02
                                            1-1

-------
   STUDY OBJECTIVES

   In an attempt to address some of the concerns unique to Class I area analyses, it was decided
   that a case study would be conducted to apply the MESOPUFF n air quality modeling
   system following the IWAQM interim recommendations (EPA, 1993). This study would
   identify and summarize the decisions made, would record and summarize the resolution
   process for these decisions, and would provide a written record of the  resources used to
   complete the effort. The objective was to learn by experience where the difficulties are in
   the process of conducting such an analysis, and when possible, to provide a means for
   resolving these difficulties. It was not an objective to provide a meaningful assessment of
   PSD, NAAQS or AQRV impacts for the Class I areas considered in the study. As will be
   seen in the following discussion, on several occasions significant departures were made in
   conducting this study from that which would be expected if a realistic  assessment were to be
   developed. For instance, the  source inventory considered only some of the states
   surrounding the Shenandoah National Park and the James River Face Wilderness, and thus is
   incomplete.  And in order to conserve resources, the sources were consolidated into ten
   surrogate sources for the purposes of this study.  These departures allowed the emphasis of
   the project to be focused on a critique of the process and resource needs, which were the
   primary study objectives.

   A realistic assessment, following the interim IWAQM recommendations, would require that
   all important sources be modeled (without consolidation).  If the modeling objective is to
   determine PSD impacts, then all relevant sources that consume PSD increment must be
   considered.  If one desires to  determine the impact of a single new (or modified) source, then
   the PSD increment from the new source must be added to all pre-existing PSD sources.  It
   would be possible to model the impacts from a single source and then add those  impacts to
   prior MESOPUFF II results, assuming the prior results were available. If not, it  would be
   necessary to model all relevant PSD sources to assess the total PSD increment consumed.

   The demonstration modeling assessment described in this report did not  completely follow
   IWAQM's interim recommendations. One of the goals of the study was to perform the
   demonstration assessment using the MESOPUFF U modeling system as developed for a
   personal computer (PC) system. This presented some significant limitations regarding the
   ability to follow the interim IWAQM recommendations. As an example, for a realistic
   assessment of multiple sources, the interim IWAQM recommendations require that two five-
   year MESOPUFF II modeling exercises be conducted; one run using all relevant sources to
   determine impacts to secondary NAAQS pollutants (secondary particulate matter)  and
   AQRVs (visibility and deposition), and a second model run using only sources beyond 50
   km from a receptor for SO2 and NOX (and primary PM10). The MESOPUFF II results for
   SO2 and NOX are then to be added to results from a Gaussian model (such as ISCST2) for
   sources within 50 km of the receptor. In the MESOPUFF U modeling demonstration
   described in this report, only  one run was performed (for three years) using sources beyond
   50 km of Shenandoah National Park. The MESOPUFF U results (for one month) were
   added to ISCST2 results to demonstrate the integration process.
94030rl.02
                                           1-2

-------
   STUDY APPROACH

   In order to assess the impact of implementing the Phase 1 recommendations, and ultimately to
   improve the Phase 1 modeling system, the following tasks were carried out:

             The MESOPUFF II model and associated processors were tested using the
             example problem intended for Support Center for Regulatory Air Models bulletin
             board (SCRAM BBS) distribution.

             The SCRAM BBS example problem computer files were evaluated and some
             suggested improvements were implemented.

         •   A five-year meteorological data set suitable for input to the MESOPUFF n model
             was developed for a multi-state area surrounding Shenandoah National Park.

             A modeling protocol was developed that outlined the approach and procedures to
             be followed in the MESOPUFF n demonstration modeling assessments.

         •   Demonstration model simulations were performed for the assessment of visibility,
             acidic deposition, and PSD increments for a set of real sources in the states
             surrounding Shenandoah National Park.

         •   Model simulations were performed to test the sensitivity of concentrations to the
             distance between sources and receptors using a set of "pseudo" sources placed in
             successive rings around Shenandoah National Park.

         •   An assessment report was developed (this report), documenting the modeling
             process, and the results obtained.

   Throughout the project, there has been coordination with IWAQM members and coworkers
   regarding implementation of the Phase 1 recommendations for assessing visibility and acidic
   deposition impacts in Federal Class I land associated with the Shenandoah National Park.
   IWAQM members representing FLMs (National Park Service, NFS; and National Forest
   Service, NFS), have provided assistance in modeling protocol related to the determination of
   visibility and acid deposition in Class I areas. The Virginia Department of Environmental
   Quality (VDEQ) and the EPA Region HI office provided assistance and guidance regarding
   receptor locations, emission source data collection, and PSD source selection.  The Office of
   Air Quality Planning and Standards (OAQPS) provided direct oversight of the project. The
   IWAQM members who participated in this study are listed in Appendix A.

   This report documents the procedures followed to demonstrate the implementation of the
   Phase 1 recommendations for assessing the PSD and AQRV impacts at Class I area land
   associated with Shenandoah National Park. Section 2 describes the testing of the SCRAM
   BBS example problem and the improvements made to those files. Section 3 describes the
   development and application of the MESOPUFF II modeling system for assessing the impacts
94030rl.02
                                           1-3

-------
   of regional point sources on Shenandoah National Park.  Section 4 describes the analysis
   performed to assess the relationship between distance of a source and impact at a receptor in
   the park. The study results are summarized in Section 5, including a discussion of what was
   learned during the modeling process.

   The technical approach for this study is described briefly below:
   Testing of the Example Problem

   The MESOPUFF II software and documentation were retrieved from the SCRAM BBS,
   installed on a PC, and run for the example problem. The model output was compared with the
   sample output provided to ensure that the model was running properly on our system. Errors
   and/or omissions in the model documentation were identified.

   A set of recommended improvements for the MESOPUFF n system and example problem
   were developed. The recommended improvements considered input files, documentation and
   user instructions, and modifications to the MESOPUFF n code. The recommended
   improvements were implemented and a revised version of the MESOPUFF II PC files were
   placed on the SCRAM BBS.
  Demonstration Application of MESOPUFF II to Shenandoah National Park

  MESOPUFF II was applied to the estimation of PSD and AQRV impacts at Shenandoah
  National Park as a demonstration of implementing the IWAQM interim (Phase 1)
  recommendations. The necessary data were collected and processed, including
  meteorological,  emissions, receptor, and land use data.  Data processing procedures were
  developed to prepare model inputs and for postprocessing of model output. The process
  involved in implementing the IWAQM Phase 1 recommendations was documented, including
  the identification and resolution of key modeling issues and the resources used to complete the
  application. The products of this application were: (1) an assessment of the experience in
  performing the modeling exercise, including suggestions for improving the Phase 1
  recommendations, (2) a demonstration assessment of the impacts at Shenandoah National
  Park for a set of PSD sources surrounding the park, and (3) a five-year meteorological data
  set, to be made available for distribution.
  Application of MESOPUFF II for Distance versus Impact Analysis

  In addition to the application intended to demonstrate the use of MESOPUFF II to assess the
  impacts of real PSD sources on Shenandoah National Park, a second analysis was performed
  to provide some indication of the relationship between distance of a source and impact at a
  receptor in the park. MESOPUFF n was applied to assess the impacts of sources placed on
  rings at specific distances from Shenandoah National Park.
94030rl.02
                                          1-4

-------
         2  REVIEW OF THE SCRAM BBS FILES AND EXAMPLE PROBLEM
  The Support Center for Regulatory Air Models bulletin board (SCRAM BBS) maintains a
  number of air quality models for distribution, including the MESOPUFF n modeling system.
  The SCRAM BBS MESOPUFF II modeling package consists of the model and processors,
  associated documentation, and an example problem.  The initial task in evaluating the
  IWAQM Phase 1 recommendations regarding MESOPUFF II was to review these BBS
  distribution files and example problem, and suggest improvements to the package.
  EVALUATION OF SCRAM BBS FILES

  We were successful in locating and retrieving the MESOPUFF II modeling package and
  associated documentation from the SCRAM bulletin board. Downloading the four
  MESOPUFF II "zipped" files required approximately three hours (at 2400 baud rate).  No
  problems were encountered in expanding the files and executing the test case. The test case
  output obtained on our system (PC/486) was identical to that provided with the model, and no
  problems were encountered in the actual mechanics of running the model. In addition,
  WordPerfect files for MESOPUFF n documentation were downloaded from SCRAM and
  printed. No problems were encountered with that operation.

  It should be noted that no test case was provided for the first MESOPUFF II meteorological
  pre-processor, called READ62. Thus, this processor was not tested. READ62 reads upper-
  air meteorological data from tape. Between the execution of READ62 and the execution of
  MESOPAC, the user must fill  in missing data by hand.  This process is likely to cause more
  difficulties than the actual execution of the model.  As will be described later in this report, a
  significant investment of manpower is necessary to create the input files for MESOPAC.

  Some omissions were identified in the instructions provided for executing the test case that
  might cause confusion for new users. In addition, suggested improvements were made to the
  test case post-processing files.  Each of these is described in detail below. During the testing
  process, one error in an input file supplied for the test case and one error in the model code
  itself were identified.  These are also described below.
  Instructions for Executing the Sample Problem (README.TXT)

  The text file README.TXT was included in one of the MESOPUFF H "zipped" files. It
  begins with a discussion of how to "unzip" the files, which appears unnecessary since users
  who do not know how to unzip files will not be able to access README.TXT.
94030rl.02
                                          2-1

-------
   The next paragraph discusses disk space and memory requirements. We recommended that a
   statement be added that an 8087 math coprocessor or equivalent capability is required to run
   the model. Also, the total disk space required (12.65 MB) should be included. The disk space
   required for each of the unzipped files is listed, but the total disk space needed is larger, since
   the zipped and unzipped files must co-exist for at least a short time. It should be stated that an
   additional 2.57 MB are required to execute the test case, bringing the total disk space
   requirement to 15.22MB.

   The next section discusses documentation for the model and its application.  The 1984 version
   of the User's Guide is cited, rather than the revised 1993 User's Guide. Since several changes
   have been made in the model and associated processors since 1984, we recommended that the
   1993 Revised User's Guide be cited, and perhaps even included in the documentation
   available from SCRAM. (Key sections of the 1993 Revised User's Guide will  be made
   available by EPA in the near future.)

   The following sections describe how to execute the test case.  A typographical error was
   found in the description of step 4, where it states that the user should type "MESOPAC" to run
   the model, rather than "MESOPUFF". Also, in that section it states that three output files are
   created, when actually four output files are created. The missing file is PUFF.LST. It might
   also be helpful to indicate that all output files with the .DAT extension are binary files,
   whereas those with .LST extensions are ASCII text files.

   In step 6, the instructions for executing the MESOFILE post-processor are not provided. A
   line should be added stating that the user should type "FILE" to run the post-processor.  If the
   user types "MESOFILE" rather than "FILE", an error message will  result.
   Error in Input File PUFF.INP

   This file contains all non-meteorological parameters needed to run MESOPUFF n, including
   modeling domain definition, number of pollutants, and source and receptor information. Line
   14 of this file contains chemistry parameters, including the background ozone concentration
   (CO3B) and the background ammonia concentration (CTNH3). For the test case, the value of
   CO3B is set to 80 ppb and the value of CTNH3 is set to 10 ppb. These are the default values
   specified for these parameters (Scire et al., 1984). However, these values were read by
   MESOPUFF n as 8.0 ppb and 1.0 ppb, respectively, because the decimal point was omitted in
   the input file. The values read by the model are echoed to the output file, PUFF.LST, which
   provided confirmation of the error.

   The background O3 value is not critical in the test case because hourly ozone values are
   provided. However, the background value is used to fill in for missing hourly data.  The error
   in CO3B should be corrected because users are likely to copy the test case input file for use in
   other applications and may believe that they are using a value of 80 ppb rather than 8 ppb.
94030rl.02
                                           2-2

-------
   The total ammonia concentration, on the other hand, is critical to the test case results.
   CTNH3 is used to calculate the distribution of total nitrate between nitric acid (HNO3) and
   aerosol nitrate (NO3"). The NO3" concentration is extremely sensitive to the specification of
   CTNH3.  Changing CTNH3 from 1 ppb to 10 ppb results in a tenfold increase in NO3" in the
   test case.

   Correction to the input file can be accomplished simply by moving the CO3B and CTNH3
   values one place to the left,  and adding decimal points after each value.

   We recommended that further thought be given to the default values of CO3B and CTNH3.
   The value of 80 ppb for ozone on a regional scale is higher than the values assumed for other
   regional modeling applications, particularly in winter. A default value of 40-60 ppb may be
   more appropriate for CO3B. The background ozone level  is used in the model to calculate the
   conversion rates of SO2 to sulfate and NOX to nitrate.  The value of 10 ppb for CTNH3 also
   appears quite high for a regional background level. In fact, CTNH3 should not represent the
   total background ammonia since the model contains no background sulfate or nitrate for the
   sources not being simulated. It represents what could be termed  "excess ammonia", the
   amount remaining after background sulfate and nitrate have been fully neutralized. Given
   this, 10 ppb is almost certainly an upper bound; hence, aerosol nitrate concentrations predicted
   by MESOPUFF II may represent upper bounds. A more representative value for the excess
   ammonia concentration in the eastern U.S. is between 0.5 and 3 ppb, depending on the season.
   For the test case, however, we recommended that the values of 80 ppb (for ozone) and 10 ppb
   (for ammonia) be retained, since they are the default values given in the existing
   documentation.
   Error in MESOPUFF II Subroutine CHEMTF

   The model code was not scrutinized in detail for this task. However, one error was found in
   the chemistry subroutine CHEMTF.  In calculating the concentration of ammonia that is
   available to combine with nitric acid to form aerosol nitrate, the amount of ammonia taken up
   by sulfate is subtracted from the total ammonia:

         CANH3 = CTNH3 - PPB(2)

   The equation should read:

         CANH3 = CTNH3 - 2.*PPB(2)

   because ammonium sulfate contains two ammonium ions per sulfate ion.


   Test Case Post-Processing Files

   The test case input files for the MESOFILE postprocessor only calculate 24-hour average
   concentrations and deposition rates for a single species, SO2.  If the model is to be used for
94030rl.02
                                           2-3

-------
  visibility and acid deposition assessments, much more postprocessing will be required.
  MESOFILE is configured such that only one pollutant can be processed at a time. Thus, the
  user must either run a large number of small post-processing jobs for each MESOPUFF II
  simulation, or set up a file that will execute all desired MESOFILE runs in the appropriate
  order.  A test case example would be helpful in demonstrating how such a file could be set up,
  and illustrating  some of the additional features of MESOFILE.

  We recommended that the test case postprocessing files be expanded to provide, at a
  minimum, 24-hour average concentrations and deposition rates for all five species. It might
  be desirable to calculate cumulative total (wet plus dry) deposition for total sulfur and total
  nitrogen.  These calculations can be accomplished with the existing MESOFILE system. In
  order to calculate visibility parameters such as extinction, additional software development
  would be needed. Therefore it does not appear feasible at this time to expand the post-
  processing test case to include extinction calculations.
   IMPLEMENTATION OF IMPROVEMENTS

   Three recommended tasks were identified:

         (1) Correct and expand the instructions for running the test case in README.TXT.

         (2) Expand the test case post-processing files.

         (3) Modify the PUFF.INP file to correct the format errors in CO3B and CTNH3, and
             modify the CHEMTF code to correct the CANH3 calculation.
   The first task was implemented to correct and expand the instructions for running the test
   case.  The README.TXT was updated, including editing the few typographical errors, and
   adding more detailed disk space information.

   The second task, development of new post-processing software, was not included in the
   revised SCRAM BBS example files because sufficient documentation on the postprocessing
   program (MESOFILE), contained in the 1993 MESOPUFF II User's Guide, was not yet
   available on the SCRAM. However, for the distance versus impact analysis (see section 4 of
   this report), a post-processing procedure using MESOFILE was developed and applied.

   The third task was approved and implemented.  The SCRAM BBS files were revised to reflect
   the changes to PUFF.INP and the CHEMTF subroutine.
94030rl.02
                                           2-4

-------
             3 DEMONSTRATION OF THE PHASE 1 MODELING SYSTEM
   The primary task of this research effort was to conduct a demonstration application of the
   MESOPUFF II modeling system, following the IWAQM Phase 1 recommendations (EPA,
   1993). At the outset, the specific air quality setting that would dictate the use of the
   MESOPUFF II system was undefined.  The focus of the project was to go through the process
   of gathering and processing data, exercising the models, and analyzing results, in order to
   identify issues and problems that users of the MESOPUFF n modeling system would
   inevitably confront. In addition, because the pre-processing of meteorological data is the most
   time-consuming step encountered during MESOPUFF n application, an important objective
   was to prepare a five-year set of MESOPUFF II meteorological inputs for a multi-state region
   surrounding Shenandoah National Park that would be suitable for future applications.

   The first step in the demonstration application was the development of a modeling protocol.
   The EPA protocol for MESOPUFF II (EPA 1992a) and the interim IWAQM
   recommendations were consulted in development of the protocol.  The first draft of this
   document was prepared in the form of an outline, and was used as a focus for discussion in
   initial meetings with the IWAQM. After several key issues were resolved, a more detailed
   modeling protocol was prepared. This document identified a proposed modeling domain and
   proposed approaches for development of inputs and processing of outputs.  The issues
   presented in the modeling protocols, and the decisions that were ultimately made regarding
   modeling protocol, are presented in the remainder of this section.

   In order for the results of the demonstration application to have some value, beyond simply
   demonstrating that the Phase 1 modeling system could be exercised, the IWAQM decided that
   the sources to be modeled should consist of existing sources subject to PSD requirements and
   located within 200 km of Shenandoah National Park.

   The original modeling plan, as described in the protocol, was to:

       • Prepare 5 years (1988-92) of meteorological data using MESOPAC, suitable for future
        applications in the region.

       • Simulate 5 years (1988-92) with MESOPUFF H using a single set of PSD  sources. The
        results would indicate the cumulative PSD increment and AQRV impacts due to all the
        modeled sources.

   As the source data were being gathered, it became apparent that identification of "PSD
   sources" is not straightforward, that there are multiple triggering dates for PSD provisions for
94030rl.30
                                           3-1

-------
   major and minor sources and for differing locations, and that it might not be possible to
   identify and model all PSD sources.  Some of these issues are described in Appendix B.
   These uncertainties diminished the possible relevance of the MESOPUFF n results but do not
   impair the objectives of the demonstration application.

   Further discussions with IWAQM members led to the idea of reducing the scope of the PSD
   source application of MESOPUFF II and adding an additional  analysis using hypothetical
   sources. A proposal was developed, whereby the original  scope would be changed to:

      • Prepare 5 years (1988-92) of meteorological data using MESOPAC, suitable for future
        applications in the region. This task would remain the same.

      • Simulate 3 years (36 contiguous months selected from 5 year period 1988-92) using
        MESOPUFF n with selected PSD sources, using actual  source data collected from EPA
        Region HI states.

      • Perform 24 additional MESOPUFF II simulations using hypothetical sources at varying
        distances from  Shenandoah NP to provide some insight into the relationship between
        distance from Shenandoah NP  and potential PSD and AQRV impacts.

   This modified modeling approach still accomplished the original objectives of preparing 5
   years of meteorological data for future applications, going through and documenting the
   modeling process for actual PSD sources, and using the modeling results to evaluate the long-
   term PSD and AQRV impacts (over a 3-year period, instead of 5 years).  By modifying the
   scope in this way, the  total number of MESOPUFF II simulations remained constant, and the
   additional effort to prepare the source ring data was not significant.  Hence, we were able to
   obtain two useful modeling results instead of just one, and did not relinquish the original
   objectives of the demonstration modeling exercise.

   The remainder of this  section describes model input preparation and application for the PSD
   source analysis.  This  section is structured to provide discussions of issues, proposed approach
   (protocol), problems, and resolutions  for each step of the modeling process.  The impact
   versus distance analysis is described in Section 4.
   MODELING DOMAIN DEFINITION

   MESOPUFF n is suitable for simulating the mesoscale transport and dispersion of air
   pollutants from a source or group of sources, and estimating their impacts on remote receptors
   at distances often to hundreds of km downwind (Scire et al., 1984). The MESOPUFF II
   modeling system utilizes three modeling grids. The meteorological grid is the largest; the
   computational grid may be smaller than the meteorological grid or the same size, but must be
   of the same resolution. The sampling grid may be smaller than the computational grid and
   may be of higher resolution.  The errors resulting from the projection of a curved surface onto
   a
94030rl.30
                                           3-2

-------
   rectangular grid become more pronounced for larger dimensions. Maximum recommended
   dimensions for the modeling domain of 1000 km in the east-west direction and 600 km in the
   north-south direction are suggested. Grid resolution is recommended to range from 10 to 50
   km, based on the overall size of the domain and limitations in data storage.
   Meteorological Domain

   This domain is defined through inputs to the MESOPAC n meteorological preprocessor
   (hereafter referred to as MESOPAC), and is the basic reference frame for all spatially varying
   input data to MESOPUFF H. In the Phase 1 demonstration implementation of MESOPUFF II,
   PSD and AQRV impacts were estimated for receptors located in and around Shenandoah
   National Park (SNP) and James River Face Wilderness (JRFW) in Virginia. These Class I
   areas are 250-300 km west of the Atlantic Coast. Prevailing winds tend to be from the
   southwest but range from south through northwest. Considering the regional meteorology
   along with known PSD source emission distributions in the eastern U.S., we suggested in the
   protocol that the meteorological domain extend further to the west from SNP than it does to
   the east. Consequently, we developed a preliminary domain to cover a full 1000 by 700 km
   region, extending about 300 km to the north, east and south of SNP, and 500+ km to the west.

   We then proposed two optional extensions to this domain. In the first  option, the southern
   boundary of the preliminary domain could be extended southward by 100 km to encompass a
   larger potential source region, which could strongly impact the Shenandoah area under
   prevailing southwest flow conditions. In the second option, the northern boundary could
   similarly be extended 100 km northward, which would encompass the  important source
   regions of northern Ohio as well as include all of EPA Region in.

   Although both extensions would safely contain (within an ample number of buffer cells) all
   potential source areas to be modeled by future MESOPUFF U applications, we proposed
   implementing the preliminary 1000 by 700 km domain (without extensions) for the Phase 1
   MESOPUFF II demonstration.  This decision was primarily based on projected data and
   storage requirements and the fact that the longitudinal and meridional extents of the
   preliminary domain were at recommended limits. Beyond these issues, however, we felt that
   this grid represents the best balance between adequate geographical coverage and the
   performance limits of the MESOPUFF II model; i.e., sources located in regions beyond the
   boundaries of this grid would not be well modeled by MESOPUFF U as the model was not
   designed for travel distances beyond 500 km.  Further, although application of MESOPUFF U
   will not necessarily utilize the full meteorological domain in this or future studies, it was
   desirable to generate relatively well resolved meteorological fields on a large regional domain
   so that we could establish a flexible meteorological database for future MESOPUFF II
   applications.

   Recommended grid spacing for MESOPUFF II is 10-50 km. The IWAQM selected a grid
   spacing of 20 km for the demonstration study. To minimize storage and CPU requirements,
   grid spacing of 40 or 50 km could have been used, but may have degraded model estimates.
94030rl.30
                                           3-3

-------
   We had also recognized that future coordination with MM4 (to be used in Phase 2) modeling
   grid definition may have been desirable, although differences in mapping projections between
   the two models would have to be addressed.

   Figure 3-1 displays the final meteorological domain selected for the Phase  1 demonstration.
   By specifying a grid size of 20 km, an array of 35 by 50 grid points covers the recommended
   1000 km by 700 km domain within the 100 by 100 grid limit currently set in MESOPAC and
   MESOPUFF II.  We had estimated in the protocol that the resulting five years of hourly
   meteorological data generated on this grid would require about 3.3 gigabytes of storage.
   Computational Domain

   All sources and receptors to be modeled must be contained within the computational domain.
   Puffs are not tracked after they leave this grid.  Overall size of this domain can be set equal to
   the full dimensions of the meteorological domain, or specified to occupy a subset of the grid.
   All sources and receptors should be at least 20-50 km from the boundaries of this grid to avoid
   underestimating concentrations by immediately losing puffs from near-boundary sources
   and/or missing short-term recirculation events at near-boundary receptors.  The size of the
   computational domain does not affect the volume of output produced, but does affect the CPU
   requirements of the modeling.  Grid spacing must be the same as the meteorological domain.

   For the demonstration study, sources were to be located as far as 200 km from SNP.
   Provision for a 4 grid (80 km) buffer zone on all sides of this source region yielded a
   computational domain of 30 by 30 grid points (Figure 3-1; dotted inset). We proposed using
   this smaller computational grid, rather than the entire meteorological domain, to minimize
   computing requirements while still giving an adequate representation of the sources to be
   studied. Future studies that include different source regions can use the same meteorological
   fields, but specify a computational domain appropriate to the application.
   Sampling Domain

   MESOPUFF II allows output concentrations and fluxes to be reported for both gridded and
   non-gridded receptors.  Gridded receptors consist of all grid points within a user specified
   sampling domain.  Non-gridded receptors can be located anywhere within the computational
   domain. Gridded receptor concentrations and fluxes can be used to produce spatial isopleth
   maps; as such, specification of a sampling grid is useful for general characterizations of an
   area.

   The sampling domain may be specified as a subset of the computational domain, with a
   maximum of 40 by 40 grid points. Grid spacing may be smaller than on the meteorological
   and computational grids. This is accomplished by specifying an integral number of divisions
   of the computational grid spacing (usually 2). Note that an increase
94030rl.30
                                           3-4

-------

JH '
HO
Ğ
ğ
	 •- 	 i 	 : 	 \
                                                        *r^&
Q

"5
,S
'w

"5

9



i
                                                                                           u
                                                                                           •<
                                                                                           a,
                                                                                           o
                                                                                           Ğ
                                                                                           w
                                                                                           x
                                                                                                  e
                                                                                                 Ğn
94030rl.30
                                                           3-5

-------
   in the number of divisions reduces the overall coverage of the sampling domain when the
   maximum number of grid points are specified. Since the size/resolution of the sampling
   domain controls the amount of MESOPUFF n output produced, it should not be larger than
   absolutely necessary.

   In preparing the Phase 1 demonstration protocol, we recognized that the use of gridded
   receptors greatly increases the amount of output produced by MESOPUFF II. In an effort to
   control MESOPUFF II computation time and the size of sampling output we did not anticipate
   using the sampling grid in the demonstration application of MESOPUFF n. Model
   concentrations and fluxes were calculated at non-gridded receptor locations only.
   Non-Gridded Receptor Locations

   As discussed in the protocol, receptor sites were to be located within SNP and JRFW.
   Receptor sites within SNP were to be distributed in a manner that represented the entire park,
   but not so densely that nearly identical results would be obtained at neighboring receptors.
   For a grid spacing of 20 km and sources located 50 to 200 km from the park, a spacing
   between receptors of 5-10 km was expected to be adequate.  Additionally, we had planned to
   place several receptors  outside SNP and JRFW, between the sources and the Class I area
   receptors. These were to be selected after the locations of sources to be modeled had been
   provided. Decisions on receptor selection for future MESOPUFF II applications should be
   made on a case-by-case basis.
   Selection of Non-Gridded Receptors

   We received UTM coordinates for 200 receptor locations in SNP and 67 receptor locations in
   JRFW from VDEQ (Browder, 1993).  The specification of 267 non-gridded receptors within
   MESOPUFF II would have resulted in larger amounts of output than would have been optimal
   for post-processing. Furthermore, many of the receptors were located very close together, and
   the model was not expected to be able to resolve the differences between them.  Therefore, we
   selected a subset of the receptors to be used in the MESOPUFF II demonstration modeling.

   A simple approach was developed to estimate the minimum spacing for which concentration
   differences would be modeled.  Since the sources were to be at least 50 km from the receptors,
   we examined downwind plume widths at 50 km using the Pasquill-Gifford-Turner dispersion
   curves. The plume width at 50 km for D stability is approximately 8 to 9 km. Half the
   expected plume width, or roughly 4 to 5 km, was used as a guideline in selecting non-gridded
   receptors for MESOPUFF H.

   Figure 3-2(a) displays the entire MESOPUFF U computational domain, where the insets
   denote the placement and extent of the SNP and JRFW receptor areas shown in Figures 3-2(b)
   and 3-2(c), respectively. Figure 3-2(b) presents the array of receptors for SNP, and identifies
   those selected as MESOPUFF U receptors (denoted as circles). Receptors
94030rl.30
                                           3-6

-------
        420
                                                                                   ğ

                                  II                       SI




                                       11
               by             Mtf (c).

94030rl.30
                                                  3-7

-------

                    680
                                    o
                                          v*Ti.  f"t #r%,

                                 r
                               0
                             0


                                                                                   4280
                                                                                  •4S40
                                                                                  •4220
                                              OT*
                1-1(1)),                                            NP.
                                fcf iğ IffiMXnifF I            tie              and
                       aU 200                           The 29 IB i;
-------
       83-Q
835

X!M *
X
x ^
e x x
M
X X * g
k!
!*r i

x,0
<
X
X
	 ..s 	 ""W-Jf1 	 ğ•—•——
X
                                 X     X !

                                    M  *|
                                  Qtyxlj

                                                                           •4J6O
                                                                            4155
                                                                           '4150


            3-2CC),                                             W.
                            far it           I           (be              ml
                   ;ğy ir                               iin 20 km            grtt
           ox
94030rl.30
                                              3-9

-------
   were selected to cover all boundaries of the park. In total, 55 of the 200 receptors were
   selected for modeling.

   Figure 3-2(c) shows the array of receptors for JRFW, and identifies those selected as
   MESOPUFF n receptors (again denoted as circles).  This is a much smaller region than SNP,
   extending less than 10 km in each direction.  Nine of the 67 receptors were selected for use in
   MESOPUFF H.

   Once receptors for the Class I areas were selected, however, no additional outside receptors
   were specified. Reduction in the number of PSD sources to be modeled, and consequent
   movement of resulting aggregated sources to average locations (to be discussed below),
   clouded the importance of monitoring concentrations between sources and receptors. When
   we considered the uncertain benefits from such an analysis, we found it difficult to justify the
   additional output and complications to postprocessing.
   SOURCES TO BE MODELED

   Selection of Sources: Issues and Protocol

   MESOPUFF II can be used to simulate impacts from existing sources, proposed sources, or
   hypothetical sources. For the demonstration study, the IWAQM decided that existing sources
   that began operations since the beginning of the PSD program would be modeled. The
   IWAQM felt that this would provide a good foundation for future modeling in the area. In
   addition, since the PSD increments are expressed as the cumulative impact of all changes in
   emissions, it was of interest to assess how much of the available Class I PSD increments have
   already been "consumed" to date.

   The IWAQM agreed to limit the sources to be modeled in the present study to those within
   EPA Region HI, comprising the states of Delaware, Maryland, Pennsylvania, Virginia and
   West Virginia, and the District of Columbia.  This decision was based upon practical
   considerations such as ease of communications between affected states. The IWAQM
   recognized from the outset that sources outside EPA Region in (located at distances further
   than 200 km from SNP, such as those in Ohio) also have the potential to impact air quality
   and deposition in SNP, and that therefore this demonstration application would not provide a
   complete answer of the impact of existing sources.  For that reason, the meteorological
   modeling domain was specified to be large enough to encompass these sources and transport
   corridors. Sources located within EPA Region HI and at a distance of between 50 km and 200
   km from SNP became the  focus of this study.

   It was proposed that all PSD sources or a subset of sources would be modeled, depending
   upon the number of such sources and data availability.  The identification of existing PSD
   sources includes the determination of a baseline date, i.e., the date after which any new
   construction or modification of existing sources will be considered as part of the accumulative
   PSD impact. A discussion of the issues involved in determining the various baseline dates
   (for
94030rl.30
                                           3-10

-------
   major and minor sources) to be applied for PSD impact analysis, determination of AQRV
   impacts, and the associated inventories needed for each analysis is presented in Appendix B
   (Cimorelli  1993).

   Therefore,  the actual number of sources to be modeled in the demonstration study was not
   specified in the modeling protocol.  This caveat was also based upon recognition that the CPU
   time needed to run the model increases with the number of sources, and can become an
   important factor in the decision of number of sources to model.  As originally configured,
   MESOPUFF II could treat up to 1000 point sources. Although MESOPUFF II can also be
   used for area sources, no area sources were  to be modeled in this study.  It is important to note
   that when multiple sources are included, only the cumulative effect is modeled — individual
   source attribution is not possible.

   For each source, the model requires location, stack parameters (height, diameter, temperature,
   flow rate),  and emission rates of SO2, sulfate, and NOX. If primary non-sulfate aerosol is to be
   modeled, two MESOPUFF II simulations are needed, with the primary aerosol labeled
   "sulfate" in the second simulation. Primary non-sulfate aerosol was not to be modeled in the
   present study,  as available resources did not allow for two sets of simulations, and since it was
   expected that the majority of visibility impacts due to PSD sources was to be captured by
   examining only sulfate and nitrate effects.

   The Phase  1 recommendations state that, for sources within 50 km, MESOPUFF II should not
   be used to estimate PSD impacts but can be used to estimate visibility and deposition impacts.
   This means that in a situation where sources both closer and further than 50 km are to be
   modeled, two MESOPUFF II simulations would be needed, one with  and one without sources
   within 50 km.  In order to avoid the need for two model runs,  it was decided that sources
   within 50 km of the park would not be modeled with MESOPUFF II in the demonstration
   study, but one or two would be modeled with ISCST2, in order to demonstrate the integration
   of MESOPUFF n and ISCST2 results. We planned to follow the Phase 1 recommendations
   concerning application of ISCST2.

   Subsequent to the development of the modeling protocol, the  decision was made to conduct a
   parallel set of MESOPUFF n analyses using uniform idealized sources placed in rings at
   various distances from  SNP.  The ring source analysis is described in  Section 4. The
   remainder of this section is devoted to the PSD source analyses.
   Processing of PSD Source Data

   For the demonstration application, source data were received from the states of Pennsylvania,
   Maryland, and Virginia. The data provided by Virginia also included some sources that are in
   the states of Maryland and West Virginia, but are traditionally included in Virginia PSD
   analyses. Therefore, some duplicate information was obtained for the state of Maryland. In
   processing these data for input to MESOPUFF II, the following operations were carried out:

      1.  Sources within 50 km  of any SNP receptor were identified and removed.  There were
          seven such sources, all in the state of Virginia.
94030rl.30
                                           3-11

-------
      2.  Sources where two or more stacks are present at the same location were combined into
          a single source with emissions equal to the sum of the individual emissions, exit
          velocity equal to the average of the individual exit velocities, and stack diameters
          adjusted to give the appropriate total combined mass fluxes.

      3.  All information provided in English units were converted to metric units.  Also, UTM
          coordinates provided in zone 18 were  assigned zone 17 equivalents by first converting
          them to latitude-longitude and then converting back to UTM.

      4.  Small sources, with emissions of both SO2 and NOX less than 5 g/s, were excluded.

   The resulting set of 27 sources is shown in Table 3-1; their locations are displayed in Figure
   3-3 (indicated by crosses). Several issues arose during the processing of these sources. These
   issues, and our approach in resolving them, are described below:

      1.  Some emission totals were provided as both short-term and annual rates. In these
          cases, the short-term rates were used, since the model was to be used primarily to
          determine the highest short-term impacts. The resulting annual average impacts will
          overestimate the true annual average impacts.

      2.  There were some differences between the emission totals reported by Maryland and
          Virginia for several Maryland sources. In case of differences, the higher total was
          used.

      3.  In two cases, emission offsets were reported as negative emissions.  Negative
          emissions cannot be input into MESOPUFF n. In one case, where the offset occurred
          at the same location as actual emissions, the offset was subtracted from the total
          emissions.  In the other case, the offset was at a different location. This offset cannot
          be modeled in a single model run.

      4.  MESOPUFF II allows emissions of primary sulfate. The sulfur emissions data
          provided were in terms of SO2 and did not specify emission levels of primary sulfate.
          For most fuel burning sources, primary sulfate emissions are roughly equal to 3
          percent of SO2 emissions (Cass, 1980).  Therefore, for purposes of this demonstration,
          primary sulfate emissions of 3 percent (as sulfur) of the SO2 emissions were included
          in the MESOPUFF n input file.  Accounting for the difference in molecular weight
          between sulfate (SO4=) and SO2, the SO2 emission rates were multiplied by 0.045 to
          obtain the sulfate emission rates.

   This set of 27 sources was presented to IWAQM as the set of sources to be modeled in the
   demonstration application. However, attempts to model these 27 sources with MESOPUFF II
   for the first month, January 1988, were unsuccessful. On an 386/25 MHz PC, 16 hours were
   required to process the first 76 hours of the simulation.  Hour 77 consumed 38 CPU minutes
   alone. Although these CPU times would be reduced somewhat on the 486/50 MHz PC that
   was planned for use in the demonstration application, it would clearly be impossible to
   complete  a full 60  months of MESOPUFF
94030rl.30
                                           3-12

-------
   TABLE 3-1. Original set of sources to be modeled in the MESOPUFF II demonstration application for Shenandoah NP and James River Face
   W.
Name
50 - 100 km from SNP
PEPCO Hb
Warrier Run
Ogden-Martin
Patowmack
SEI Birch
N Branch
100 - 150 km from SNP
LG&E Altavista
Multitrade
Old Dominion
Doswell
Westvaco
Cogentrix, Richmond
Brandon
Chalk Pt.
Mettiki Coal0
P.H. Glatfelterd
UTM-E
(km)
806.1
693.6
826.9
800.3
822.5
643.5
653.2
653.3
704.6
813.2
588.6
815.2
885.2
875.6
638.0
853.5
UTM-N
(km)
4345.7
4385.0
4289.4
4330.9
4241.6
4346.9
4109.3
4107.8
4082.6
4191.4
4183.7
4151.0
4346.3
4275.4
4353.0
4421.4
xa
36.3
30.7
37.3
36.0
37.1
28.2
28.7
28.7
31.2
36.7
25.4
36.8
40.3
39.8
27.9
38.7

23.7
25.3
20.5
22.5
18.1
23.3
11.5
11.4
10.1
15.6
15.2
13.6
23.3
19.8
23.7
27.1
Dist
from
SNP
(km)
64
83
77
54
90
95
115
117
135
107
108
134
140
127
103
147
StkHt
(m)
65
82
88
30
123
109
67
53
134
58
147
76
187
65
43
69
Diam
(m)
6.87
3.75
2.26
5.49
4.72
3.96
2.44
3.84
6.77
5.03
3.44
2.64
6.71
11.25
2.90
4.10
ExVel
(m/s)
35.4
23.6
19.2
37.2
0.3
15.1
23.6
19.1
14.4
11.2
24.3
16.8
27.3
33.7
27.4
13.7
Temp
(K)
260
398
405
854
339
444
341
472
323
389
444
339
413
789
333
430
SO2
(g/s)
317.2
54.8
20.3
31.0
27.7
102.2
17.9
6.8
145.2
74.7
0.0
49.2
1893.6
136.4
9.9
0
NOX
(g/s)
179.3
26.0
82.5
51.1
69.4
60.3
28.7
13.9
309.1
68.7
23.0
113.5
630.8
177.0
23.3
61.3
94030rl.30
                                                             3-13

-------
   TABLE  3-1.    Concluded.
Name
150 -200 km from SNP
Cogentrix, Dinwiddie
LG&E Hopewell
Mecklenburg Cogen
Cambria Cogen
Coal Dynamics
ColverPP
Ebensburg
Harrisburg
Lancaster
Solar Turbine
York Co.
UTM-E
(km)
815.2
829.5
720.8
695.1
597.0
686.4
690.8
850.6
871.5
869.0
865.6
UTM-N
(km)
4121.2
4134.1
4053.3
4482.2
4413.5
4490.9
4480.4
4465.0
4444.6
4435.0
4436.6
xa
36.8
37.5
32.0
30.8
25.9
30.3
30.5
38.5
39.6
39.5
39.3
ya
12.1
12.7
8.7
30.1
26.7
30.5
30.0
29.3
28.2
27.8
27.8
Dist
from
SNP
(km)
152
156
166
170
171
181
169
178
176
168
167
StkHt
(m)
76
67
84
76
31
107
76
59
93
20
95
Diam
(m)
2.64
2.44
3.51
2.74
2.74
3.48
2.59
3.66
2.82
2.00
2.42
ExVel
(m/s)
16.8
23.6
22.8
27.7
25.6
18.0
18.3
8.1
19.4
15.7
20.0
Temp
(K)
339
341
339
422
478
411
422
477
405
488
389
SO2
(g/s)
49.2
14.9
49.4
133.3
28.5
144.7
9.8
2.8
4.1
0
7.2
NOX
(g/s)
94.6
27.5
116.0
77.5
14.0
72.4
5.9
46.7
40.1
48.9
15.3
   a Meteorological grid cell coordinates.
   b Combined values for three units, and including an emission offset for NOX
   0 Emission data from Maryland. Virginia lists emissions from this facility of 2.4 g/s SO2 and 3.8 g/s NOX.
   d Combined values for two units.table 3-1 (cont)
94030rl.30
                                                                3-14

-------
410
                                 sa§
        21!-
                                 11                      21

                                      II Computational

               I~S,        of fie           1            doiMta.               of ill
                           for the                I demoBSBatloo.  Crosses        jJi 23
                                           lie 10     anpefsied sĞuseĞ med for (he
94030rl.30
                                               3-15

-------
   II simulations within the time frame of the project.  A series of 24-hour test simulations was
   performed to determine which input parameters had the largest effect on CPU time. These
   simulations, described below, clearly demonstrated that the number of puffs was the key
   variable.  They also suggested that the number of puffs must be reduced by a factor of 10 to
   achieve the needed reduction in CPU time.

   The number of puffs is governed by the puff release rate and the number of sources. The puff
   release rate had been set at the default value of 4 puffs per hour.  Although MESOPUFF II, as
   originally configured, allows up to 1000 point sources, it only allows for a total of 10,000
   puffs. With 1000 sources and a puff release rate of 4 per hour, the limit on the number of
   puffs would be reached after 2.5 hours. With 27 sources and a puff release rate of 4 per hour,
   the original configuration for the demonstration application, the puff limit would be reached
   after 92 hours.  In order to run MESOPUFF II in this configuration for one full month,
   therefore, it would be necessary to increase the maximum number of puffs substantially. This
   would require re-compiling MESOPUFF n and would increase the size of the executable
   code.  Since the MESOPUFF n executable, at 3.3 MB, is near the limit for use on a PC with 4
   MB memory, the maximum number of puffs might be a limiting factor even if CPU time
   considerations were not.

   In order to reduce the number of puffs by a factor often and still preserve at least a portion of
   the original intent of the demonstration application, the puff release rate was reduced from 4
   puffs per hour to 1 puff per hour. This deviates from the Phase 1 recommendations; however,
   because the output averaging interval used for the demonstration application was 3 hours
   rather than 1 hour, this  change was considered acceptable.  The number of sources was
   reduced from 27 to 18.  Under these conditions, 25.8 CPU hours were required for the January
   1988 simulation.  Although this was a major improvement over the 27 source case, it was still
   too slow to allow completion of the project on schedule. Further reductions in the  puff release
   rate were judged to be infeasible. Therefore, the number of sources was reduced further from
   18 to 10.  This reduced the CPU time to less than 12 hours for January 1988. This set of 10
   sources was used for the demonstration application of MESOPUFF U.

   In reducing the number of sources to be modeled, the goal was to retain as much as possible
   of the total emissions mass, and reduce the number of sources by consolidation of nearby
   sources rather than elimination of sources.  In the condensation of the original 27 sources to
   18 sources, sources within one grid cell of each other were consolidated. No sources were
   eliminated. In order to reduce the number of sources to 10, however, it was necessary to
   eliminate  three small sources that were not located near other sources, and sources within 2-3
   grid cells  were  consolidated. The consolidated sources were represented as point sources,
   with coordinates and stack parameters obtained as an arithmetic mean of the individual
   sources. The three sources  eliminated, Westvaco, Chalk Pt, and Coal Dynamics, have total
   emissions of 165 g/s SO2 and 214 g/s NOX.  This amounts to 5 percent of the total SO2
   emissions of the original 27 sources, and  8 percent of the total NOX emissions. Table 3-2
   shows the final set of 10 consolidated sources, the facilities that are included in each, and the
   location and emission parameters
94030rl.30
                                           3-16

-------
   TABLE 3-2.  Final condensed set of sources modeled with MESOPUFF II for the SNP demonstration application.

1
2
3
4
5
6
7


8
9

10


Sources Included
PEPCO H, Patowmack
Warrier Run
Ogden-Martin, SEI Birch
N Branch, Mettiki
LG&E Altavista, Multitrade
Mecklenburg, Old Dominion
Doswell, Cogentrix-Richmond,
Cogentrix-Dinwiddie,
LG&E Hopewell
Brandon
Cambria Cogen, Colver PP,
Ebensburg
P.H. Glatfelter, Harrisburg,
Lancaster, Solar Turbine, York
Co.
xa
36.2
30.7
37.2
28.0
28.7
31.6
37.0


40.3
30.5

39.1


ya
23.1
25.3
19.3
23.5
11.5
9.4
13.5


23.3
30.2

28.0


StkHt
(m)
48
82
106
76
60
109
70


187
86

67


Diam
(m)
8.77
3.75
3.24
4.69
4.44
7.11
5.35


6.71
5.08

6.60


ExVel
(m/s)
36.3
23.6
10.0
21.3
21.4
18.6
14.0


27.3
21.3

15.4


Temp
(K)
557
398
372
389
407
331
352


413
418

438


SO2
(g/s)
348.2
54.8
48.0
112.1
24.7
194.6
252.1


1893.6
287.8

14.1


so4=
(g/s)
15.7
2.5
2.2
5.0
1.1
8.8
11.3


85.2
13.0

0.6


NOX
(g/s)
230.4
26.0
151.9
83.6
42.6
425.1
426.4


630.8
155.8

212.3


   ' Meteorological grid cell coordinates.
94030rl.30
                                                           3-17

-------
   for each. Figure 3-3 also displays the locations of these sources in relationship to all 27
   original PSD sources (indicated by circles).
   Run Time as a Function of Number of Puffs

   A series of 24-hour test simulations was performed to ascertain the dependence of
   MESOPUFF II execution time on input parameters including puff release rate, number of
   sources, number of receptors,  ozone data specification and chemistry. The results of these
   simulations showed that decreasing the number of puffs, whether by decreasing the puff
   release rate or the number of sources, had a major effect on execution time. Changes in the
   other parameters had much smaller effects on CPU.  Specifically, a 50 percent reduction in the
   number of puffs reduced CPU by 56 percent.  A 50 percent reduction in the number of
   receptors reduced CPU by 6 percent. The use of a single default ozone value rather than
   hourly values at monitoring stations had virtually no effect on CPU.  Even running the model
   in an inert mode (i.e. with all chemistry turned off) resulted in only an 8 percent reduction in
   CPU time.

   These results show that there is a greater than linear dependence of CPU time on number of
   puffs, even within the first 24  hours of a simulation.  The departure from linearity increases
   for longer simulations.  Based on the CPU requirements for the month-long simulations, a
   simulation with 17 percent of the original number of puffs required only 6 percent of the CPU
   time.
   MODELING EPISODE DEFINITION

   Following the Phase 1 recommendations, a five-year period was to be modeled. In order to
   supply the MESOPUFF II modeling system with the most recent environmental input data, the
   IWAQM agreed to the five year period from 1988 through 1992. As planned, the MESOPAC
   meteorological preprocessor was run to generate MESOPUFF II input fields for this entire
   period.  As discussed above, however, a decision was made to reduce the Phase 1 PSD
   demonstration to three years (1988-1990), with the remaining 24 months of MESOPUFF n
   integrations reserved for the ring source analyses.

   The default value and Phase 1 recommended value for output averaging interval is one hour;
   however, substantial reductions in disk storage requirements can be achieved by increasing
   this interval.  Based on the observation that the PSD increments and AQRV parameters are all
   based at a minimum on 3-hour averages, we judged that an averaging interval of 3 hours
   would be sufficient to produce each desired output.

   It should be noted that this deviates somewhat from the Phase 1 recommendations, which state
   that visibility effects should be calculated using  hourly concentrations and hourly relative
   humidities.  The intent of this recommendation was to capture the diurnal effects and avoid
   calculating visibility on the basis of 24-hour averages. The use of 3-hour averages still allows
   for the determination of diurnal trends, and would seem  to preserve the intent of the
   recommendation.
94030rl.30
                                           3-18

-------
   MESOPUFF II cannot simulate seasonal or annual time scales in a single model run,
   especially on a PC. One problem we had identified during review of the SCRAM BBS
   example problem at the start of the project was the demand put upon PC disk storage by
   MESOPAC and MESOPUFF II input/output (I/O).  We had anticipated executing
   MESOPAC and MESOPUFF II in one-month intervals, using the MESOPUFF H restart
   option for all but the first month of each year.  One month of MESOPAC output for the 50 by
   35 grid point domain produced 50-55 megabytes of output. This is a manageable amount for
   the 200 MB PC hard drive systems we used; other potential users of the meteorological data
   are likely to have similar systems. Monthly MESOPUFF II output was significantly less
   (about 1  MB).

   During the demonstration applications, we found that a single MESOPUFF n integration over
   several months would have also stressed MESOPUFF n memory requirements. This is
   related to simplistic puff accounting within MESOPUFF n, which continuously adds new
   puffs as they are emitted, yet does not purge old puffs that exit the computational grid.
   Particularly during stagnant conditions, the number of puffs quickly grows to the maximum
   dimension  of the puff array, which slows MESOPUFF II considerably.  This problem is
   discussed further in following sections.
   DEVELOPMENT OF METEOROLOGICAL INPUTS

   MESOPAC requires the following input meteorological data: (1) twice daily upper air
   temperature and wind soundings in NCDC TD-6200 format at up to 20 radiosonde stations;
   (2) scheduled airways surface observations in NCDC CD144 format, which include hourly
   winds, temperature, relative humidity, cloud cover and ceiling height at up to 100 stations;
   and (3) hourly precipitation data in NCDC TD-3240 format.  The CD144 data sets are input
   directly into MESOPAC, whereas the upper air and precipitation data must be scanned for
   missing data and reformatted using several MESOPAC preprocessors. MESOPAC and its
   preprocessors expect the data in these specific NCDC formats.

   Prior to the start of this project, SAI already possessed in-house upper  air and hourly surface
   data for 1988 covering the entire U.S. However, we had to procure precipitation data for
   1988-1992, and upper air and surface data for 1989-1992. At the time, we anticipated that
   some of the data sets we were to acquire would not necessarily fit the MESOPAC formats,
   and either the data would have to be reformatted for the processors, or the processors would
   have to be revised. Furthermore, the protocol called for the development of a procedure to fill
   in missing data in the upper-air soundings. We recognized that for a 5-year database, it would
   not be feasible to fill  in missing data by hand.  Consequently, we needed to develop a
   processor that uses a  specified set of rules to automatically fill in missing data. Based upon
   the size of the proposed meteorological domain (see Figure 3-1), it was also expected that the
   MESOPAC limits on the number of surface and upper air sites would probably need to be
   increased.
94030rl.30
                                          3-19

-------
   Raw Data Processing for MESOPAC and Related Problems

   Data Procurement

   SAI managed to procure the remainder of the necessary raw meteorological data sets through
   the Western Regional Climate Center, Desert Research Institute (WRCC/DRI) of the
   University of Nevada, Reno. These nationwide data sets include TD-6201  upper air
   soundings and TD-3280 hourly surface observations for the years 1989-1992, and TD-3240
   precipitation data for the years 1988-1992.

   Two options were available for procuring meteorological data: to order through WRCC/DRI,
   or through the National Climatic Data Center (NCDC) in North Carolina. The least expensive
   format through DRI was to receive surface, upper air, and precipitation data sets in terms of
   annual files covering the entire U.S.  Special processing to extract monthly data for the
   MESOPUFF II domain would have required extra charges for running their processors, and
   would have doubled or tripled processing time. DRI did not insist on prepayment, and was
   able to start delivering data within two weeks of the order.  Data transfer was simplified by the
   fact that DRI processed all data on a  SUN/Unix system and wrote to Exabyte 8 mm tape
   cartridges. This was a significant advantage since we expected to handle the large raw data
   sets on our Trace/Unix mainframe systems, which possess Exabyte tape drives. Finally, DRI
   subscribes to the Internet nationwide communications network, which enables data transfer
   from DRI to purchasers on Internet directly, removing mail time.

   Alternatively, NCDC charges are based on the amount of data, not the amount of processing.
   To process the same surface, upper air, and precipitation data, but only within states covering
   the MESOPUFF II modeling domain, NCDC quoted SAI about a 25% higher cost than for the
   nationwide data sets from DRI.  Hence, full nationwide data sets through NCDC would have
   been more than twice the DRI price.  Further, NCDC requires payment for  the total amount
   before they begin processing data,  after which transfer of data tends to take about a month.
   Combining correspondence time between NCDC and SAI to request an estimate, time
   required to arrange for and transmit payment, and processing time at NCDC, we had estimated
   a wait of six or more weeks. Time constraints, along with the slightly higher cost, rendered
   this option infeasible. Realizing that breaking up nationwide data sets would demand more
   in-house labor and computer time than working with smaller regional data  sets, we
   nevertheless decided to order through DRI.

   Future users of the MESOPUFF II system may prefer to order raw meteorological data via
   NCDC if time constraints are not so crucial. MESOPAC front-end processing work may also
   be reduced substantially if data are requested for stations within a specific area for a particular
   time window. However, if a specific region is requested, NCDC requests a list of station
   identification (WBAN) numbers or a list  of states to extract.  Obviously, increased processing
   also escalates the time NCDC requires to process and deliver the data. NCDC writes all files
   to 9-track tape, and the purchaser must specify the exact tape format (variable vs. constant
   record length, byte density, blocking factors, etc.). In the future, it may also be possible to
   obtain the raw data on CD-ROM format.
94030rl.30
                                           3-20

-------
   Protocol for Raw Data Processing

   Aside from the sheer volume of data to be processed, development of several new
   MESOPAC preprocessors was necessary for the following reasons. First, all data files were
   delivered in slightly modified formats from standard NCDC formats, such that long single
   station records were simply broken up to maintain 80-character width files to ease file transfer
   and quality assurance procedures. Further, each of the three data sets were delivered in
   several subfiles per year to facilitate transfer and handling.  New processors were needed to
   read the altered formats, concatenate all subfiles for each year, and extract data for the time
   (monthly) and space windows of interest. Second, while all precipitation and upper air data
   we had on order and/or in-house were in the proper format for the MESOPAC preprocessors,
   it was necessary to convert all TD-3280 surface data to CD144 format before MESOPAC
   could be run.  Finally, an automated system was required to scan and fill extracted surface and
   upper air data sets for missing values,  and conduct simple quality assurance checks.

   The first preprocessor that we had planned to develop was to reformat the TD-3280 surface
   data to CD144 format.  It was expected that this would be a rather straightforward task, and
   should not have demanded much time. We planned to add provisions in the program to scan
   the data set for missing values and take appropriate steps to fill these gaps.  At the time, we
   expected to simply linearly interpolate over time for periods of missing values of less than 6-
   12 hours. Linear interpolation over time is best for short periods (a few hours) but is
   increasingly inappropriate for longer periods since diurnal patterns, such as diurnal
   temperature waves, are eliminated. Therefore, a contiguous period of 12  hours was
   considered the maximum length of time considered for linear time interpolation.  For sites
   with data missing for the bulk of an entire day or more, but with ample data coverage for the
   month, we proposed to  spatially interpolate data from nearby sites to the station location using
   a distance-weighted average. For sites missing significant portions of data for a particular
   month (e.g., two or more weeks), the station was to be disregarded for that month.

   We also planned to develop a processor for the upper air data that replaced manual editing
   between the READ62 and MESOPAC programs. READ62 only flags missing data within a
   sounding, or writes a warning message if an expected sounding is completely missing.  The
   new processor was to read the READ62 output; for missing data in the vertical, it would
   linearly interpolate between pressure levels.  We realized that vertical interpolation is suitable
   only for relatively shallow portions of a sounding, since atmospheric conditions between two
   vertically distant levels are often decoupled. Therefore, if a large portion of a sounding was
   missing, or if a sounding was missing  altogether, the program would spatially interpolate data
   from nearby sites to the station location using a distance-weighted average.

   The revised MESOPUFF n user's guide (EPA, 1994) suggested simply replacing missing
   soundings with the nearest representative soundings. However, we felt that a better approach
   was to use spatial averaging, as upper  air stations are typically separated by at least a few
   hundred kilometers, and no single sounding can properly represent a missing profile
   (particularly near the surface). Aloft, the scales of horizontal atmospheric motions are on the
94030rl.30
                                            3-21

-------
   order of 200-1000 km or larger (a distance that is well resolved by sounding data), so
   horizontal interpolation is acceptable above about 1000 m.  Near the surface, this approach
   loses validity since local effects (1 km or less) dominate the vertical profile. However, we
   believe that averaging from several sites is far better than simply replacing a missing sounding
   with another.

   Throughout the process of generating meteorological data input files for MESOPAC, we
   encountered a multitude of problems related to missing or substandard raw data.  In
   developing a system of data extract!on/filling/processing algorithms to work in tandem with
   existing MESOPAC preprocessors, we attempted to account for as many contingencies as
   possible. However, numerous unexpected patterns of missing data were flushed out during
   actual processing that required  significant amounts of time in which to identify the problems
   and develop alternative programming approaches. As discussed below, many preprocessors
   had to be altered and rerun several times, which proved to be a very labor intensive process.
   The development and subsequent dissemination of a five year MESOPUFF n meteorological
   database was accomplished in part to alleviate this demanding step for future MESOPUFF II
   users.  However, for those users who find it necessary to develop a new database, we present
   below a description of all meteorological preprocessing steps, along with a review of all
   significant problems we have encountered with the data and existing preprocessors.
   Surface Data Processing

   A preprocessor was developed that reformats the TD-3280 surface data to CD144 format. We
   added provisions in the program to scan the data for missing values and take appropriate steps
   to fill these gaps. A flow diagram showing the necessary steps to provide CD144 files to
   MESOPAC is shown below:

                              Concatenated Raw TD-3280 File
                                            !
                        GETSTN/RESFC Tape Extraction  Programs
                                            !
                                 Intermediate Monthly File
                                            !
                        TOCD144 Data Reformatting/Filling Program
                                            !
                                   Monthly CD 144 File
                                            !
                              PARSE Station Splitting Program
                                            !
                               Monthly CD144 Station Files
                                            !
                                       MESOPAC
94030rl.30
                                           3-22

-------
   The first two programs, GETSTN and RESFC, are used together to extract hourly surface
   data for a given time and space window from a single raw TD-3280 file. Because of file size
   and time required to operate the tape extraction programs, GETSTN and RESFC were run on
   our Trace Multiflow 14/300 Unix mainframe system.

   As delivered to SAI, the TD-3280 data were broken down into many subfiles per year,
   arranged by station WBAN number.  Since WBAN numbers are not necessarily ordered
   geographically, it was expected that all subfiles would have to be run through RESFC to
   ensure that all available surface data were extracted for the modeling domain. Hence, our
   original plan was to run GETSTN/RESFC programs on each raw annual subfile, and operate
   a rather complex concatenating procedure on the resulting intermediate monthly subfiles to
   obtain a single monthly intermediate file. For the 1988 surface data set this approach was
   satisfactory since only four subfiles existed for the year. Data sets for 1989-1992 were instead
   split into 13 subfiles per year.

   Realizing that our original approach would be quite inefficient for the last four years, we
   decided instead to identify only those subfiles containing station data within our domain, and
   concatenate those raw TD-3280 subfiles together first and run GETSTN/RESFC for each
   single concatenated data set.  In this way, we were able to disregard 5 of the 13 subfiles, and
   concatenated the remaining eight.  The resulting raw files required about 250-300 MB of disk
   space per year. In terms of labor, about one-half hour was spent identifying which subfiles to
   concatenate, after which setting up and running the concatenating procedure required about an
   hour per year on our Unix mainframe system (yet very little CPU time).

   GETSTN and RESFC already existed for previous in-house processing of meteorological
   data; however, CD 144 format required slight modifications to RESFC to output different
   types of meteorological variables.  The product of these two programs is an intermediate
   monthly surface data file for  stations located within the meteorological modeling domain.
   These programs were run for all 60 months  of 1988-1992, extracting data for a spatial window
   within the latitude/longitude  ranges 34-42°N and 71-86°W, which resulted in hourly surface
   observations for up to 48 stations.  Although RESFC required substantially more mainframe
   CPU time to process the larger annual concatenated files, significant labor was saved by
   removing RESFC setup time for all subfiles and removing the post-RESFC concatenating
   procedures. On the Unix system, processing of each month took between 20-30 minutes of
   mainframe CPU time (6 hours per year); about an hour of labor per year was spent setting up
   and executing RESFC, and inspecting output diagnostics and data files.

   In order to reduce processing time further, we developed a process that screened stations
   within the intermediate surface file for missing data before running the TOCD144 program.
   If a station was missing more than 50% of the data for a particular month, we deleted it from
   the file before running TOCD144. The process utilized rapid awk/sed editing (a Unix-based
   protocol), which requires very little time. For 1988, two surface stations had to be repeatedly
   deleted from all twelve months of data.
94030rl.30
                                          3-23

-------
   TOCD144 is a new PC preprocessor that reads the monthly intermediate files, fills missing
   values, and reformats the data into monthly files in CD 144 format. The program was written
   to fill  1-6 hour intervals of missing data using linear interpolation over time. For longer
   periods of missing values, data from surrounding surface stations are used to spatially
   interpolate information to the station using an inverse-distance-squared weighting technique.
   A minimum  of 2 stations and a maximum of 4 stations are used within 200 km of a station
   needing spatial interpolation. If all data for a particular station is missing for more than one
   week, the station is deleted from the database for that month.  TOCD144 was compiled using
   Microsoft (MS) FORTRAN 5.0; processing of each month took about 5  minutes of PC CPU
   time using an 486/66 MHz processor (this PC was used throughout all raw meteorological
   data preprocessing). Total labor and computer time required to run CD144 averaged about 1.5
   hours  per year.

   We first encountered a problem with the 1988 TD-3280 surface data in that it did not contain
   a "present weather" field, which is needed in CD 144 files to assign precipitation data as either
   liquid or frozen (for use in wet deposition calculations in MESOPUFF II).  When the TD-
   3280 data was originally ordered,  a decision was made to exclude present weather  in the data
   set to reduce file size.  Peculiarities of the present weather field format, along with
   complexities involved to map TD-3280 present weather descriptors to CD144 descriptors,
   would have necessitated the addition of highly complex logic structures to the RESFC
   program.  After consulting one of the authors of MESOPUFF II,  it was determined that
   MESOPAC uses input temperature to determine liquid/frozen precipitation states  if present
   weather fields are missing from the CD144 data (Scire, 1993). Therefore, all present weather
   fields  were set to missing values in the processed CD144 files.

   It was eventually determined far into processing of surface data with TOCD144 that pressure
   values were incorrect and appeared to be in wrong units.  Upon further investigation, we
   realized that  pressures read from the raw TD-3280 data set were  station pressures in inches of
   mercury, rather than sea level pressure in  millibars. Instead of restarting surface data
   processing from the raw TD-3280 level to extract the proper pressure field, we decided to
   convert station pressure to millibars and calculate a sea level pressure using the hypsometric
   equation (via station temperature and elevation) within TOCD144.

   At the same time, we made a decision not to interpolate for missing cloud ceiling height
   fields, either in time or space, due to the often large temporal  and spatial variations observed
   in cloud heights, and the potentially large and uncertain impacts on MESOPUFF II
   applications. MESOPAC requires cloud heights and cloud cover to adjust a daytime clear-
   sky solar insolation index, which is in turn used along with wind speed to estimate Pasquill-
   Gifford-Turner stability class. When cloud cover is less than 50 percent, MESOPAC ignores
   cloud  height effects on solar insolation index. At night, only cloud cover and wind speed are
   used to determine stability class. For the  most part, missing cloud heights occurred at night,
   as these observations are typically estimated visually; they only periodically occurred during
   daylight hours.  In TOCD144, sky cover percentage was  used to set cloud height: if sky cover
   was less than 50 percent, cloud height was set to an "unlimited" ceiling;  if cloud cover was
   greater than
94030rl.30
                                           3-24

-------
   50 percent, cloud height was set to 5000 ft, a height which altered the classification of solar
   insolation within MESOPUFF H to "low cloud" values (refer to page 5-13 of the MESOPUFF
   II User's Guide; EPA, 1994). Our binary "all or nothing" methodology drastically simplified
   the procedure and only occasionally influenced the solar insolation index under daytime
   broken/overcast conditions.

   A final simple processing program called PARSE was then written to split monthly CD 144
   files into numerous monthly station files, which are ready to be used directly by MESOPAC.
   Processing of single monthly files and splitting into separate station files just before running
   MESOPAC for each month simplified intermediate file management and provided faster
   meteorological data processing.  Typically, data for 45-48 surface stations needed to be split
   into separate station files each month, taking just about 1 minute of PC CPU using MS
   FORTRAN 5.0,  and virtually zero labor.
   Upper Air Data Processing

   Although TD-6201 upper air data were generally in the correct format for READ62 (some
   minor modifications were necessary), the READ62 program only flags missing data within a
   sounding, or writes a warning message if an expected sounding is completely missing.  We
   therefore had to develop a processor that replaces manual editing between the READ62 and
   MESOPAC programs with an automated procedure. A flow diagram showing the necessary
   steps to provide upper air data files to MESOPAC is shown below:

                                    Raw TD-6201 File
                                            !
                             REUPR2 Tape Extraction Program
                                            !
                                  Monthly TD-6200 File
                                            !
                             FILLUPR Data Filling Program
                                            !
                               Monthly Filled TD-6200 File
                                            !
                                   READ62 Program
                                            !
                                 Monthly MESOPAC File
                                            !
                             PARSE Station Splitting Program
                                            !
                             Monthly MESOPAC Station Files
                                            !
                                       MESOPAC

   The first program (REUPR2) is a modified version of a program existing in-house for
   previous processing of upper air data.  Its purpose is to extract 12-hourly upper air data for a
   given time
94030rl.30
                                          3-25

-------
   and space window from a single raw nationwide/annual TD-6201 file. It was modified to
   output TD-6200 format readable by READ62, as the previous version output a format useful
   for another meteorological model.  Because of file size and time required to operate the tape
   extraction program, REUPR2 was also run on our Unix mainframe system. Also, in order to
   reduce file sizes and expedite filling of missing data, REUPR2 writes out sounding data
   between the surface and about 7000 m, and deletes non-mandatory levels if they have missing
   pressure. The resulting TD-6200 file contains intermediate monthly upper air data for stations
   located within the meteorological modeling domain. The program was run for all 60 months
   of 1988-1992, extracting data for a spatial window within the latitude/longitude ranges 34-
   42 °N and 71-86°W, which resulted in upper air data from as many as 8 rawinsonde stations.
   REUPR2 required about 15-20 minutes of mainframe CPU per month, yet very little labor
   was involved.

   A new processor (FILLUPR) reads the intermediate TD-6200 file and scans the sounding
   data for missing values. The program linearly interpolates between pressure levels for data
   gaps less than 200 mb deep. If a significant portion of the sounding  is missing (i.e., > 200
   mb), or if the sounding is missing altogether, the program spatially interpolates data from
   nearby sites to the station location using a distance weighted average (at mandatory levels
   only). The value of 200 mb was a rather arbitrary choice; in  the lower troposphere such a
   pressure depth is equivalent to about 2000  m. Vertical variations in hydrodynamic variables
   across this depth result from larger-scale atmospheric dynamics, and are thus better estimated
   via horizontal interpolation from other soundings.  Vertical interpolation across large depths
   in a single sounding may overly simplify a sounding aloft.

   We made the decision to use the program FILLUPR on the output of REUPR2 instead of
   running the data through READ62 first, which would flag the missing values or soundings.
   The main reason for this is that READ62 eliminates non-mandatory sounding levels that may
   be important for performing a vertical interpolation on missing data.  Since pressure is known
   for all levels up to 7000 m, only missing height, temperature, and winds are filled (in that
   order).  FILLUPR follows the processing  scheme outlined below:

   (1)  In a first pass through the data, the program finds the first level where there is a missing
       value (either height, temperature, or wind).  The program proceeds up the sounding until a
       non-missing value is found. If this gap is greater than 200 mb deep, the mandatory levels
       within the missing block are flagged for spatial interpolation of this particular variable.  If
       the gap is less than 200 mb  deep, linear interpolation (using height, or log-pressure if
       height is not available) is performed for the variable at all levels (mandatory or otherwise)
       within the gap.  For winds, interpolation is done for vector components.

   (2)  The program then proceeds farther up the sounding and repeats the procedure for data
       gaps aloft. If no valid data is found up  to the top of the 7000 m sounding, and the data gap
       is greater than 200 mb, all mandatory levels above the level of good data are flagged for
       spatial interpolation. If the data gap to the top of the sounding is less than 200 mb, the
       data are extrapolated using the highest two levels with valid data (this saves on spatial
       interpolation time).

   (3)  Data for all soundings needing spatial interpolation are written to a temporary direct
       access file to ease memory requirements and quicken run time. In the second pass
94030rl.30
                                           3-26

-------
      through the data, a matrix containing the top 5 closest stations to each of the stations in
      the file is computed.  Then the flags for missing data at mandatory levels are checked for
      spatial interpolation.

   (4) When a mandatory level requiring spatial interpolation for a particular variable is
      identified, data from the closest 2 to 4 upper air stations within a radius of 500 km are
      used in an inverse-distance weighted average.  Stations containing spatially interpolated
      values at the same mandatory level are not used in this calculation.

   (5) After a temperature sounding has been completely filled via spatial interpolation, the
      resulting temperature gradients are checked against a slightly super-adiabatic lapse rate to
      insure thermodynamically realistic values. The limiting lapse rate specified was -0.15
      K/m; this limit was violated only infrequently.  When the violation did occur, the
      interpolated temperatures were adjusted to adhere to an adiabatic lapse rate (-0.01 K/m)

   FILLUPR was compiled using MS FORTRAN 5.0, and took about 5 minutes PC CPU time
   per month. Labor and CPU time together required about two hours per year to process upper
   air data through FILLUPR.

   In processing 1988 upper air data, many soundings were found at hours 6 and 18
   supplementing or replacing the standard hour 00 and 12 soundings.  Since MESOPAC
   expects only hour 00 and 12  soundings, these extra  sounding data were deleted during the
   FILLUPR step. Upper air data for one particular station during August 1988 contained so
   many missing records that the resulting direct access file ran up against PC disk storage limits.
   Since so much data were missing, we decided to delete that station from the intermediate file
   and rerun FILLUPR for August 1988.

   Several array overflows occurred during test extraction of 1989 upper air data. These were
   tracked to inadequate  array dimensioning; array dimensions for the number of input sounding
   levels in REUPR2 were increased from 100 to 150, and 1989 data were re-extracted.
   FILLUPR and READ62 worked well for all of 1989 except for November.  One of the
   stations was missing data for the entire last week of the month, causing FILLUPR to fail.
   Rather than alter FILLUPR to handle this situation, we simply deleted the station from the
   November 1989 intermediate data file and FILLUPR and READ62 were rerun. We then
   discovered that extracted data for 1990 was unusable, due to an unexpected limitation in
   REUPR2 that did not allow processing of as much  data as it was supplied. This was
   apparently not a problem for previous years. REUPR2 was revised and 1990 data were re-
   extracted.

   When processing upper air data for 1991  and  1992, we found numerous stations commonly
   reporting at hours 11 and 23. Since MESOPAC cannot accept this, we elected to simply
94030rl.30
                                           3-27

-------
   relabel data at hour 11 to hour 12, and data at hour 23 to hour 00 on the following day.  This
   presented a serious logistical problem for soundings at the end of each monthly file since they
   needed to be transferred to the beginning of the files for the following months. It was
   determined that such a task would be far too labor intensive; instead, the REUPR2 extraction
   program was revised to find these incidences and properly relabel soundings before writing
   the monthly files to disk. Raw 1990-92 upper air data were then extracted once again using
   the modified version of REUPR2

   Data from only four upper air stations were available for August 1991. When attempting to
   process the data through FILLUPR, the program was unable to spatially interpolate data to
   the other four locations because of significant data gaps in the existing data. Additionally,
   upper  air data for July 1992 contained only 2 stations. This leaves serious questions
   concerning the feasibility of using upper air data from August 1991 or July 1992 for future
   MESOPAC/MESOPUFF H applications.  Nevertheless, all months of 1991 and 1992 were
   processed following the procedures outlined above.

   The resulting filled monthly TD-6200 files were then processed through READ62 to produce
   monthly upper air data files in MESOPAC input format. Although only about a minute of PC
   CPU time was required to run READ62 per  month, about one hour of labor was spent
   performing quality assurance checks on a year's worth of READ62 output.

   Finally, the PARSE station splitting program was run to provide monthly station upper air
   files for MESOPAC. Typically,  eight  separate upper air station files were generated by
   PARSE, taking only a few seconds  of PC CPU time.
   Precipitation Data Processing

   The TD-3240 precipitation data was basically in the correct format for the PXTRACT
   preprocessor. No extra processing (e.g. data filling) was necessary for these data sets; it was
   therefore anticipated that processing precipitation information would be the most
   uncomplicated of our meteorological processing tasks. A flow diagram showing the necessary
   steps to provide precipitation data files to MESOPAC is shown below:

                                    Raw TD-3240 File
                                            !
                                   PXTRACT Program
                                            !
                               Monthly TD-3240 Station Files
                                            !
                                   PMERGE Program
                                            !
                                 Monthly MESOPAC File
                                            !
                                       MESOPAC

   The PXTRACT preprocessor reads in an annual/national raw TD-3240 precipitation file and
   extracts a monthly TD-3240 file for each station within the specific states that cover the
   meteorological modeling domain.  These monthly files are then read into PMERGE, which

94030rl.30                                    3-28

-------
   processes the data into hourly precipitation rates, taking into account several data
   quality/accumulation flags, and merges all station data into a single file ready for use by
   MESOPAC

   Initial runs of PXTRACT for all 12 months of 1988 led to unreadable monthly output files.
   Although the raw TD-3240 precipitation data contained all information required by
   PXTRACT as outlined in the MESOPUFF II User's Guide, we found that these files were not
   in the exact format that PXTRACT was expecting.   The raw data was reformatted using a
   Unix awk process (taking about 1.5 hours per year of combined computer and labor time),
   and PXTRACT was rerun for 1988. PXTRACT required about 10 minutes of PC CPU time
   per month; about 1 hour of labor per year was required to run PXTRACT and examine
   program diagnostics and output.

   During processing of 1988 precipitation data, PXTRACT produced about 550 separate station
   files per month for the states covering the meteorological modeling domain. Although
   PMERGE is written to process any number of station files through repeated rerunning and
   merging, the sheer number of stations was rather unexpected. FORTRAN programs compiled
   with MS FORTRAN and operating in DOS environments only allow for 15-20 I/O files to be
   opened at one time, which would necessitate that PMERGE to be run 35-40 times per month.
   PMERGE was instead recompiled with Lahey 5.2 FORTRAN, which, if the PC is configured
   correctly, allows for up to about 250 I/O files to be opened at any one time.

   A large percentage of the precipitation stations were located outside the meteorological
   domain. An intermediate program (run between PXTRACT and PMERGE) was written to
   flag those stations outside the domain for deletion from the precipitation  database. The
   remaining stations within the modeling domain were further reduced in number by arbitrarily
   deleting every other precipitation station from the database (which still left a sufficient
   number of stations to adequately represent the spatial distribution of precipitation).  The total
   number of precipitation stations was reduced in this way to just below 200. One PMERGE
   run per month processed each set of 200 files in just a few minutes; processing a full year took
   about half an hour, requiring about 2 hours per year of labor (mostly in QA).

   We initially had many problems running PMERGE; each time we corrected for one type of
   data problem, PMERGE would stop on another. An extensive review of the code revealed
   that PMERGE did not allow for several possible combinations of data quality flags. We
   determined that it was easier to alter the PMERGE program to allow for all contingencies we
   had found within the precipitation data, rather than hand-edit or write yet another processor to
   handle the hundreds of monthly files.

   The program PMERGE required modification in order to give it the capacity to handle
   effectively certain types of precipitation records it might encounter while processing TD3240
   data.  The data records in a TD3240 file may be flagged as 'M' (missing), 'A
   (accumulation period), T (incomplete), or 'D' (deleted).  All of these files are expected to
   come in pairs. In other words, the first record flagged as 'A signals the start of an
   accumulation period, and the next record with that flag signifies the end of that same type of
   period.  When the PMERGE program was first run on the TD3240 data  for the domain, it
   crashed because it was unable to process records flagged as T. A few lines of code were
   added to enable the  program to deal with records containing this flag.  The  program required
94030rl.30
                                           3-29

-------
   further modification to enable it to handle situations in which the beginning and end of
   accumulation periods (records flagged with 'A') were separated with missing, incomplete, or
   deleted periods of records.  It was a common occurrence in the TD3240 data to have one or
   several pairs of deleted, missing, or incomplete data records sandwiched between valid
   accumulation period records, but the PMERGE program did not have the ability to deal with
   this structure. The same modification was required to enable the handling of missing record
   pairs being separated by incomplete or deleted periods of records. It appears that all
   contingencies are now handled by the revised PMERGE program.

   We found during initial precipitation processing that the raw TD-3240 format did not contain
   any information about each station's physical location (i.e., latitude/longitude or UTM
   coordinates). TD-3240 format only contains station ID number, which includes a code
   denoting the state in which a particular station resides.  The revised MESOPUFF II User's
   Guide (EPA, 1994) does not mention anything about where to obtain such information.
   Fortunately, we were able to download from WRCC/DRI an NCDC precipitation "Station
   History File" via Internet linkup. This file contains all station history and location  information
   for the entire U.S. A program was written (GRIDDIT) to cross reference station
   identification numbers within each month's MESOPAC-ready precipitation data file with
   station numbers in the history file, and write out all station grid coordinates in the proper
   MESOPAC input format.  This same program was used to calculate station grid coordinates
   for all surface and upper air stations from latitude/longitude information within respective
   intermediate files.
   Development of Gridded Land Use for MESOPAC

   Both MESOPAC and MESOPUFF II require a "typical" land use type to be specified for each
   grid point in order to determine surface characteristics (i.e. water vs. land) and to set default
   canopy resistances and surface roughness lengths.  A single representative land use type for
   each grid point is designated within the MESOPAC user input file; from there, the user may
   wish to specify roughness lengths for each land use type, or have MESOPAC use default
   values. For the current study, default surface roughness values and canopy resistances were
   selected.

   SAI has possessed a nationwide land use file in-house for many years; it was originally
   procured from a Geographical Information System (GIS) database maintained by the U.S.
   Geological Survey (USGS). This file contains the distribution of 11 land use categories over
   the entire U.S. at 1/4 degree longitude by  1/6 degree latitude resolution. The data were first
   mapped to the meteorological grid in terms of the percentage of each of the 11 categories in
   each cell, using the UAM preprocessing program PRELND.  These land use categories,
94030rl.30
                                           3-30

-------
   however, do not agree with the categories required for MESOPUFF n. A new program was
   written to identify the dominant GIS land use category in each cell and map it into an
   appropriate MESOPUFF n category. The assumed mapping arrangement of GIS-
   MESOPUFF II land use categories is displayed in Table 3-3.  All steps in developing a
   MESOPUFF II land use field were executed on our Unix mainframe, where the land use
   database and PRELND preprocessor reside.

   For those future MESOPUFF II/MESOPAC users finding it necessary to develop gridded
   land use for some other modeling domain, there are several land use databases available.  SAI
   has supplied the National Park Services (NFS) with the same land use file used in this analysis
   as part of the delivery of the NFS Air Quality Modeling System (Morris and Chang, 1992).
   Other common databases are available through the USGS, including the GIS database
   described above, and newer, higher resolution (200m) GIS databases covering most of the
   U.S.
   MESOPAC Applications

   Meteorological fields for all 60 months of the years 1988-1992 were generated for
   MESOPUFF n using the MESOPAC meteorological preprocessor.  The operation demanded
   a high degree of file handling/transfer because of the large number of MESOPAC input files
   (as many as 58 total files), and the huge output files generated by MESOPAC. The following
   steps outline our approach to setting up MESOPAC for each month:

       1. The appropriate MESOPAC input data files for the month were assembled (surface,
         upper air, and precipitation) into a common run directory; to maximize PC disk
         utilization, all MESOPAC input files were compressed together after preprocessing
         using the PKZIP utility.  Assembling the appropriate files simply meant "unzipping"
         them from the compressed files.

       2. The PARSE splitting program was run on the monthly surface and upper air files to
         split them into separate monthly station files, named appropriately for MESOPAC
         (CD1.DAT,  CD2.DAT, etc., for surface data; UP1.DAT, UP2.DAT, etc., for upper air
         data).

       3. The GRIDDIT  station location processor was run to obtain gridded station locations
         for input surface, upper and precipitation data.  It was necessary to run the program
         each month since the collection of station data changed month to month. The output
         from this program contains station locations, in terms of the meteorological grid
         location, in the proper format for the MESOPAC user input file (PAC.INP).

       4. The GRIDDIT  output file was concatenated onto a sample header of the PAC.INP
         file, and the  resulting complete PAC.INP file was edited for the particular month to be
         run.
94030rl.30
                                          3-31

-------
  TABLE 3-3. Assumed correspondence between GIS land use categories and MESOPUFF
  land use categories.
GIS MESOPUFF
Category GIS Land use H Category MESOPUFF H Land use Type
Type
1
2
3
4
5
6
7
8
9
10
11
Urban
Agriculture
Rangeland
Deciduous Forest
Coniferous Forest
Mixed Forest
Water
Barren Land
Non-forested
Wetland
Mixed Ag-
rangeland
Rocky Open Areas
11
1
6
5
5
5
12
6
10
2
8
Metropolitan City
Cropland and Pasture
Subhumid Grassland and Semi-arid
Grazing Land
Ungrazed Forest and Woodland
Ungrazed Forest and Woodland
Ungrazed Forest and Woodland
Lake or Ocean
Subhumid Grassland and Semi -arid
Grazing Land
Marshland
Cropland, Woodland, and Grazing Land
Desert Shrubland
94030rl.30
                                         3-32

-------
      5. MESOPAC was run for the month with all appropriate input files.

   During initial trial runs of MESOPAC, two small inconsistencies were found between the
   model and the input data. First, the new upper air preprocessors wrote the sounding height
   field in six characters, rather than the original five. The appropriate READ statement within
   MESOPAC was altered to allow for this. Second, unlimited ceiling heights are recorded with
   a series of X's in standard CD 144 format; in MESOPAC, a quality assurance routine was
   checking for dash characters (-) rather than X characters. MESOPAC was altered to check
   for the appropriate character.

   MESOPAC contains two date/time routines specific to the Lahey FORTRAN compiler.
   MESOPAC was therefore recompiled using Lahey FORTRAN 5.2; processing a single month
   of meteorological fields required 60-70 minutes of CPU time on a 486DX50 (50 MHz) PC for
   the 50x35 grid meteorological domain.

   The 486/50 MHz PC used for MESOPAC contained a 340 MB hard drive, of which about
   220 MB were available for data processing. Since MESOPAC produced 50-55 MB of binary
   output per month for this particular meteorological domain, and required about 5 MB per
   month of input data, only four months could be processed by MESOPAC at a time.
   Obviously, mass off-line storage became a key importance. As MESOPAC was run for each
   year, meteorological data files were generated on the PC four at a time, and transferred via
   FTP (a network file transfer protocol) to one of the Unix mainframe disks (with a volume of
   nearly 1 gigabyte). After an entire  year was completed and transferred (600 MB total), all data
   were backed up to magnetic 8 mm  tape cartridge via the mainframe Exabyte drive.
   DEVELOPMENT OF BACKGROUND AIR QUALITY INPUTS

   Protocol for Developing Air Quality Inputs

   MESOPUFF II requires input values for background ozone and ammonia. Initial or boundary
   concentration data are not supplied.  Background ozone can be specified as a single time-
   invariant region-wide default value of 80 ppb, or the default value may be overridden by
   designating some other value within the MESOPUFF n input file. MESOPUFF n also
   optionally accepts hourly ozone observations for a large number of monitoring sites via the
   OZONE.DAT input file. MESOPUFF II uses the ozone values to calculate chemical
   conversion rates for NOX and SO2.

   Following the Phase 1 recommendations, we planned to develop hourly ozone input fields.
   This option was chosen for the current study in order to provide both spatially and temporally
   improved estimates of sulfate and nitrate production over the single background level
   approach. Complete ozone data for the years 1988 through 1991 from all  stations reported to
   the Aerometric Information Retrieval System (AIRS) database were available in-house at the
   start of the project. We planned to obtain 1992 ozone data from the AIRS network. We
   realized that it was necessary to pre-process the data to obtain the format needed for input into
   MESOPUFF H.
94030rl.30
                                          3-33

-------
   MESOPUFF II can accept data from up to 50 ozone sites as originally configured.  Most
   AIRS monitors are clustered in urban areas.  We anticipated that in some cases there may be
   multiple sites within one 20-km grid cell. In such cases, we considered averaging the hourly
   values for all monitors within a grid cell. Since MESOPUFF II uses the nearest monitor to
   each puff, we optionally considered the addition of "pseudostations" with background ozone
   values in some rural locations to prevent urban ozone values from being used in those areas.
   Fortunately, several AIRS monitors are located either within,  or very near, SNP.  These were
   expected to allow a good estimate of rural ozone levels in the area of interest.

   MESOPUFF II's default background ammonia concentration is 10 ppb.  At the start of the
   project, we recognized that this value was likely to be too high for the eastern United States,
   but would provide an upper bound for aerosol nitrate. This would also result in a conservative
   (high) estimate of PM10 and visibility impacts. However, we noted that such an approach
   would not provide a conservative estimate of deposition impacts, since overestimating aerosol
   nitrate means that nitric acid would be underestimated, and nitric acid deposits more rapidly
   than aerosol nitrate. Nevertheless, we planned to maintain background ammonia levels at the
   default 10 ppb level.
   Development of Ozone Input Files

   During the course of the Phase 1 Demonstration, SAI procured (independently from this
   project) hourly ozone data for the entire U.S. for the years 1970-1992. The data were
   extracted from EPA's Aerometric Information Retrieval System (AIRS) in AMP-350 data
   work file format. Existing programs were used to withdraw hourly ozone data from 1988-
   1992 annual files into 60 monthly files for a spatial window covering the entire MESOPAC
   meteorological domain.  A new processor was  developed (AIR2MESO) that reformats the
   hourly data into OZONE.DAT format, and produces a separate file containing grid
   coordinates for each AIRS monitor location. The coordinates file was then easily inserted into
   the MESOPUFF n input file as necessary when supplying an OZONE.DAT file to the model.
   In order to simplify processing, we did not average multiple stations within the  same grid cell.
   A casual inspection of the data revealed a fairly high degree of uniformity in ozone levels for
   much of the year. Further, we did not add pseudo stations with background ozone. The total
   number of ozone monitoring sites within the meteorological domain depended highly on
   season, ranging from about 50 or 60 in the winter, to around 150 in the summer. Hence, the
   maximum  allowable number of ozone stations  in MESOPUFF II was increased from 50 to
   200.

   MESOPUFF H identifies missing data within the OZONE.DAT file and fills it with either the
   default (80 ppb) or user-specified background ozone concentration.  Since 80 ppb is a rather
   high value to represent daytime-average background concentrations for the entire year, and
   since such  a high value would likely produce relatively large "spikes" in observed ozone time-
   series (particularly in the wintertime), a background ozone level was supplied for each month.
   The AIR2MESO program was written to calculate daytime/domain average ozone
   concentrations for each month; these values, rounded to the nearest 5 ppb, were supplied to
   each monthly MESOPUFF II input file to specify background ozone levels.
94030rl.30
                                           3-34

-------
   MESOPUFF II APPLICATION

   MESOPUFF II was integrated over all 36 months of the years 1988-1990 on the same 486/50
   MHz PC as used in MESOPAC processing. Each monthly MESOPUFF II run required disk
   space for a 50-55 MB input meteorological data file, the 3.3 MB MESOPUFF H executable
   file itself, and  1.5-1.8 MB of output - a total of about 60 MB. Again, due to the 200 MB disk
   storage limit, MESOPUFF n was run for only 3 or 4 months before another set of
   meteorological input files were transferred to the PC from off-line storage.

   After MESOPUFF II was revised to correct a chemistry routine (during the review of the BBS
   example problem) and to increase the maximum number of ozone sites, it was recompiled
   using Lahey FORTRAN 5.2.  This version of the Lahey compiler combines extended and
   virtual (disk swapping) memory managers into the program executable file. This alleviated
   the necessity of running the 3.3 MB MESOPUFF II executable in a Windows environment,
   and in fact required execution from standard DOS.  We found that some applications loaded
   into memory via the AUTOEXEC.BAT led to runtime problems for MESOPUFF H.  In
   particular, the PC had to be rebooted without SMARTDRIVE (a hard disk drive utility),
   PREDIR, and SHARE (network application sharing software).  It is suspected that
   MESOPUFF II overloaded SMARTDRIVE by filling I/O buffers too quickly for
   SMARTDRIVE to handle

   MESOPUFF II has a restart option that allows puffs from the end of a previous simulation to
   carry over to the beginning of a new simulation. To limit the number of puffs carried over
   into each monthly integration, two days were assumed adequate to supply a reasonable
   amount of puff mass to the system from a zero background state.  A monthly MESOPUFF n
   application therefore involves two steps: (1) "cold starting" MESOPUFF II from a zero
   background state (no restart) two days before the end of the previous month, and running the
   model for a two-day "spin-up" period; and (2) restarting MESOPUFF n at the beginning of
   the current month and integrating for the entire month. MESOPUFF n was "cold started",
   however, from a zero background state for each January. Depending on environmental
   conditions, particularly in regards to the frequency and duration of stagnation events, CPU
   requirements for monthly MESOPUFF U integrations varied substantially.  In general, months
   in the winter and spring ran fastest, while the summer and fall months were markedly slower.
   Monthly runtimes ranged from about 45 to 120 minutes on the 486/50 MHz PC.

   All input data and parameters specified within the MESOPUFF II user input file (PUFF.IMP),
   except for background ozone concentration and averaging interval, were set at default or
   recommended  values, as outlined in Appendix A of the Phase 1 recommendations. The
   concentration/deposition averaging time, as stated above, was set to three hours.
94030rl.30
                                          3-35

-------
   DATA ANALYSIS

   PSD and AQRV Issues and Protocol

   The MESOPUFF n model produces concentrations, wet fluxes, and dry fluxes for each
   sampling grid point and for each non-gridded receptor for each of the following species: SO2,
   sulfate, NOX, nitric acid, and aerosol nitrate. As stated in the working protocol, since PSD
   increments represent incremental concentration totals for specified averaging periods, these
   values could be calculated with the existing MESOPUFF n post-processor MESOFILE.
   Averaging times for each species should correspond to PSD requirements and associated
   standards. The pollutants covered by the PSD regulations, the averaging time for each
   pollutant, calculation method and allowable PSD increment are summarized in Table 3-4.

   For this study, PM10 increments were to be calculated as described in the Phase 1
   recommendations. Modeled concentrations of sulfate and aerosol nitrate were to be converted
   to ammonium sulfate and ammonium nitrate, and summed to estimate PM10 increments.  The
   PSD increments were to be calculated for each non-gridded receptor, and the highest values
   within SNP and JRFW were to be identified.

   Since there are no standards or mandated increments for AQRVs, it was necessary to define
   AQRV criteria before postprocessing could begin. The measures for visibility and deposition
   agreed to by the IWAQM are summarized in Table 3-4.  For visibility, incremental extinction
   coefficients were to be computed from 3-hour average modeled PM10 concentration
   increments using relative humidity data and the equation provided in Appendix B of the Phase
   1 recommendations. The equation to estimate extinction was to be used for both sulfate and
   aerosol nitrate, and no other contributors to visibility were to be considered. This calculation
   required the development of a MESOPUFF II visibility post-processor that reads the 3-hour
   average concentrations and relative humidity, and calculates the incremental extinction.

   The maximum incremental 3-hour extinction over all receptors was to be compared to
   measured total extinction from the  SNP IMPROVE monitoring site.  The number (and
   percentage) of 3-hour periods for which the maximum incremental extinction represents 10
   percent or more of the measured extinction for the 90th percentile cleanest day was to be
   reported.

   Deposition impacts were to be calculated for each non-gridded receptor for total sulfur (SO2
   plus sulfate, expressed  as S) and total nitrogen (NOX plus HNO3 plus nitrate, expressed as N).
   Deposition impacts were to be expressed as the cumulative annual sum of wet and dry
   deposition, in units of kg/hectare.  At the time the protocol was written, a level of concern
   (analogous to the 10 percent increase in the 90th percentile cleanest day for visibility) had not
   been identified for deposition.
94030rl.30
                                           3-36

-------
   TABLE 3-4. PSD and AQRV parameters to be calculated from MESOPUFF II output.
Parameter
SO2
NOX
PM10

Extinction

Averaging
Period
Annual
24-hour
3 -hour
Annual
Annual
24-hour

3 -hour

Calculation Method
PSD Increments
Highest annual average for each year
Highest 2nd-high 24-hour average for each year"
Highest 2nd-high 3 -hour average for each year*3
Highest annual average over each year, expressed as NO2
Highest annual average; sum of ammonium sulfate and nitrate
Highest 2nd-high 24-hour average; sum of sulfate and nitrate
Visibility
Maximum 3 -hour extinction (sulfate plus nitrate), added to 90th
percentile cleanest day extinction. Calculate number (%) of 3-hour
periods for which extinction is increased by more than 10%
Deposition
Allowable Class I
Increment
(Hg/m3)
2
5
25
2.5
4
8



    Total S


    Total N
Annual
Annual
Highest cumulative annual deposition, sum of wet and dry, SO2 plus
SO4=, expressed as S

Highest cumulative annual deposition, sum of wet and dry, NOX plus
HNO3 plus NO3, expressed as N	
      a Defined as follows: calculate 2nd-highest value for each receptor for each year, then find the highest value among all
      receptors.
      b 3-hour averages are fixed interval averages, not running averages.
94030rl.30
                                                           3-37

-------
   Processing MESOPUFF II Output Into PSD and AQRV Measures

   MESOFILE was developed to allow for a great deal of flexibility; as such, it requires a user
   to construct complex command streams in order to identify the desired pollutant, perform the
   desired function, and output the desired quantity. Unfortunately, as the requested output
   quantity becomes more complex  (e.g., total sulfur deposition) it is necessary to expand the
   MESOFILE command streams into highly complicated structures.

   We had become familiarized with the MESOFILE postprocessor during review of the
   SCRAM BBS sample application and associated dissemination files.  To become adept at
   utilizing MESOFILE, as well as to further test the postprocessor, we decided to improve the
   sample MESOFILE runstream to calculate, from the same sample MESOPUFF II output, 24-
   hour average/sums of concentrations/deposition for all species, rather than just for SO2.

   We found the task to develop expanded MESOFILE output for the sample application to be
   overly tedious and complicated, and resulting file manipulation (of both MESOPUFF II output
   and MESOFILE intermediate files) to be burdensome. Clearly, developing a similar
   MESOFILE process to generate averages and sums over an entire year, as required by PSD
   and AQRV measures, would be difficult and inefficient. Furthermore, MESOFILE does not
   produce second highs needed for PSD analysis. Since development of a new program was
   necessary to calculate visibility parameters from MESOPUFF II output anyway, we decided to
   include all PSD and AQRV calculations into a single new postprocessor.

   The single postprocessor approach, designated PSDPOST, allowed for quick and efficient
   calculations of the specific PSD and AQRV impacts as described in Table 3-4, without
   intermediate computations and file management. AQRV computations include the calculation
   of aerosol extinction, which in turn is dependent on relative humidity.  Relative humidity data
   is contained within the large 50 MB MESOPAC output files, only a few of which can fit onto
   the PC hard drive. The objective of PSDPOST, however, was to operate on all 12 months of
   each year continuously. Therefore, it was necessary to write another processor (RHPOST) to
   extract hourly relative humidity data at each surface station from MESOPAC files, and output
   3-hour average relative humidity at each receptor location to a much smaller file for
   PSDPOST

   PSDPOST requires 12 months of binary MESOPUFF II output files and 12 monthly
   RHPOST output files for each year. The  program calculates all concentration averages and
   deposition sums outlined in Table 3-4. As discussed in the protocol, PSDPOST then
   determines humidity-dependent extinction coefficients resulting solely from ammonium
   sulfate and ammonium nitrate. This calculation specifically follows the procedure outlined in
   Appendix B of the IWAQM Phase 1 recommendations. Visibility impacts are then reported
   as the amount of time (both in terms of the absolute number of 3-hour periods and percent of
   each year) the calculated maximum extinction over all receptors is more than 10 percent
   higher than the cleanest observed extinction.  The cleanest observed extinction was taken to
   be the
94030rl.30
                                          3-38

-------
   90th percentile extinction over all years in which statistical analyses of IMPROVE monitoring
   data are available at SNP (1987 through 1991). From Sisler et al. (1993), the 90th percentile
   particulate extinction for this period at the Shenandoah monitor was 0.005 km"1. This was
   combined with a Rayleigh (pure air) scattering extinction of 0.010 km"1 to obtain a total
   observed cleanest extinction of 0.015 km"1.

   Since both MESOPUFF II and MESOPAC were compiled using Lahey FORTRAN 5.2, both
   PSDPOST and RHPOST had to be compiled with the same in order to  read the binary model
   output files. RHPOST required just a few minutes of CPU time per month (on the 486/50
   MHz PC) to extract relative humidity data from MESOPAC output files.  Again, humidity
   files could only be extracted from 3 or 4 MESOPAC output  files before more MESOPAC
   files could be transferred from off-line storage. PSDPOST required about 10 minutes of PC
   CPU per month.
   RESULTS

   Note:  The modeling results presented below are for a partial list of relevant PSD sources.

   Table 3-5 displays a summary of PSDPOST calculations from three years of MESOPUFF II
   output (1988-1990).  Results are given for receptor groups in both SNP and JRFW. All PSD
   and AQRV measures for JRFW are markedly lower than for SNP; this may be caused by the
   relative isolation of JRFW from the bulk of modeled PSD sources, in combination with the
   fact that the small number of receptors within JRFW are spatially compact. As for
   concentrations of criteria pollutants, annual averages of SO2, NOX, and modeled PM10 in both
   Class I areas were predicted to be small fractions of the total allowable Class I increments for
   all years modeled.  Short-term concentration increments were predicted to approach the
   allowable Class I increments, particularly at SNP.  The 2nd highest 3-hour and 24-hour
   average SO2 concentrations approach or exceed the allowable Class I increments, while the
   2nd highest 24-hour average modeled PM10 is between an eighth and one-half the allowable
   limit.  Also note that whereas summer 1988 was characterized by widespread  high pollution
   levels throughout the eastern U.S. (particularly for ozone), all concentration, extinction, and
   deposition predictions were distinctly higher during 1989 than during the other two modeled
   years at both Class I areas.

   The extinction measure reported in Table 3-5 corresponds to the percent of each year in which
   calculated incremental modeled PM10 (sulfate plus nitrate) concentrations lead to incremental
   extinction levels more than 10 percent above clean background levels for SNP. According to
   these calculations, it is predicted that secondary particulate matter from the PSD sources
   modeled in this analysis lead to such conditions between 20 and 28 percent of the year for
   SNP, and about 7 to 8 percent of the year at JRFW. At SNP, maximum 3-hour extinction for
   the years 1988, 1989, and 1990, were 0.0507, 0.3348, and 0.1345 km"1, respectively. Using a
   simple standard estimation procedure for clear-sky visual range (3.0 divided by extinction in
   km"1), the extinction increments at SNP
94030rl.30
                                          3-39

-------
   TABLE 3-5. PSD and AQRV parameters calculated from MESOPUFF II output.
Shenandoah NP
Parameter
SO2
NOX
PM10

Extinction

Total S
Total N
Averaging Period
Annual (|ig/m3)
24-hour (|ig/m3)
3 -hour (|ig/m3)
Annual (|ig/m3)
Annual (|ig/m3)
24-hour (|ig/m3)

3 -hour (% of year)3

Annual (kg/Ha)
Annual (kg/Ha)
1988
0.23
3.92
16.56
0.12
0.08
1.24

19.9

0.38
0.09
1989 1990
PSD Increments
0.32 0.21
5.03 3.15
19.62 9.55
0.15 0.13
0.12 0.07
3.32 1.31
Visibility
27.9 22.4
Deposition
0.36 0.28
0.12 0.07
James River Face W
1988
0.08
1.39
5.38
0.05
0.05
0.85

7.3

0.11
0.04
1989
0.10
2.93
7.35
0.07
0.07
1.25

8.2

0.15
0.16
1990
0.06
1.61
3.65
0.04
0.04
1.31

6.8

0.08
0.03
Allowable Class I
Increment
2
5
25
2.5
4
8





   a Percent of 3-hour periods for which incremental extinction is greater than 10 percent of clear-day extinction levels.
94030rl.30
                                                           3-40

-------
   translate to minimum visual ranges of 59, 9, and 22 km, for each year modeled. The
   maximum extinction increments are quite high when one considers that these estimates were
   calculated solely as a result of the secondary particulate matter generated by these emission
   sources alone, and that other SO2 and NOX sources, and natural/anthropogenic sources of
   organics and dust, were not taken into account.

   The extinction coefficient, as defined in these analyses, results from the product of 3-hour
   average modeled PM10 concentrations and a logarithmic humidity-dependent extinction
   efficiency (ranging from 3.0 m2/g at less than 30 percent relative humidity to 48 m2/g at 98
   percent relative humidity).  High levels of extinction (visibility degradation) are expected to
   occur when a highly concentrated plume of secondary particulate matter is advected into a
   Class I area, coinciding with maximum relative humidity conditions.  The logarithmic
   humidity-dependent approach for calculating extinction, however, can yield large extinction
   coefficients for even modest particulate levels if relative humidity is maximized.  For
   purposes of this study, we attempted to reduce the uncertain influence of particulate extinction
   under the highest humidities by limiting relative humidity in PSDPOST to 95 percent
   (corresponding to an extinction efficiency of 33.8 m2/g).

   On the other hand, the potential certainly exists for very high incremental modeled PM10
   concentrations to be supplied to PSDPOST.  Maximum 3-hour average modeled PM10
   concentrations (sulfate plus nitrate) determined by PSDPOST (but not required for PSD
   analyses) for these years were 6.5, 17.6, and 6.9 |ig/m3, respectively. At these levels, high
   humidity is not necessary to generate large extinction numbers.  During screening of the
   SCRAM BBS example MESOPUFF II problem, we identified that the default value of
   background ammonia (10 ppb) supplied to MESOPUFF n via the user input file was  much
   higher than typically observed.  In view of the fact that there is no background SO2 or sulfate
   specified (from non-modeled sources), the background ammonia concentration is probably far
   too high for annual particulate nitrate calculations. In the eastern U.S., particulate mass is
   generally dominated by sulfate and organic species, with much smaller contributions  from
   nitrate. We anticipated that 10 ppb of ammonia would consistently produce the maximum
   potential (i.e. conservative) ammonium nitrate concentrations. Although separate 3-hour
   average sulfate and nitrate levels were not analyzed by PSDPOST, it is likely that maximum
   3-hour average modeled PM10 concentrations would be lower if the input ammonia level is
   reduced to more realistic levels.

   Predicted maximum accumulated sulfur and nitrogen deposition for both sets of receptors is a
   small percentage of commonly observed deposition loadings for the area.  As reported by
   Olsen (1988), annual sulfur deposition in the area around Virginia and West Virginia in 1986
   ranged from 25 to 30 kg/ha, while nitrogen deposition ranged between 15 and 20 kg/ha.
   Predicted annual sulfur and nitrogen deposition within SNP from the PSD sources modeled in
   this study is typically about 10 percent and 5 percent of these  observations, respectively.
   Predicted annual deposition loadings for JRFW are about 1/3  lower than the SNP numbers.
94030rl.30
                                           3-41

-------
   ISCST2 Application and Integration

   The IWAQM Phase 1 recommendations (EPA, 1993) require the use of a steady-state
   Gaussian model for evaluating the incremental impacts of PSD sources within 50 km of a
   receptor. These impacts should be added to the MESOPUFF n results for sources farther than
   50 km from the receptors. The ISCST2 Gaussian model was applied in order to demonstrate
   the reconciliation of the ISCST2 and MESOPUFF H results.

   The ISCST2 model is described in detail elsewhere, including instructions for its application
   (EPA, 1992b), so no description of the model will be presented here. The ISCST2 model was
   used to simulate the 3-hour average concentrations of SO2 and NOX at each of the
   MESOPUFF II receptors corresponding to one of the months modeled with MESOPUFF n.
   Seven of the sources considered for MESOPUFF II application were within 50 km of a SNP
   receptor. Of the  seven, the largest NOX and second largest SO2 emission source, Wampler-
   Longacre, is situated 13 km from the western tip of SNP (see Figure 3-4). This source was
   modeled with ISCST2 for the month of July 1990.

   The objective of the ISCST2 application was to generate output files that provided the same
   PSD measures that were created by the MESOPUFF II demonstration application. The
   ISCST2 results and MESOPUFF U results would then be combined. A postprocessing
   program, ISCMPF, was developed by VDEQ that we would use for combining the model
   results (Browder, 1994). This program was designed to read from a single binary hourly
   MESOPUFF II output file (for up to five species) and four separate binary ISCST2 output
   files for each species (SO2 and NOX) modeled; hourly concentrations, three-hour averages, 24-
   hour averages, and annual averages.  Since each MESOPUFF II output file generated during
   the demonstration exercise was for a single monthly period and the concentrations were
   output as three-hour averages, ISCMPF could not be exercised for one-hour or for annual
   average data. Therefore, ISCST2 was used to produce three-hour and 24-hour average output
   only.

   Surface and upper air meteorological data for Richmond, Virginia were obtained from the
   SCRAM BBS for use in ISCST2 modeling.  ISCST2 input files were constructed with the
   following data (separate runs are required for SO2 and NOX):

      Source
      Wampl er-Longacre

      Location
      UTM coordinates: 692.8 E, 4278.0 N, zone 17

      Emission rates
      SO2:30.21bs/hr(3.81g/s)
      NOX: 98.4 tpy (2.8 g/s)

      Stack Parameters
      base elevation: 1020 ft (310.9 m)
      stack height: 45 ft (13.7 m)
94030rl.30
                                          3-42

-------


                                 =v  >.        v   :

                                  ,„
                          ->'  i  ^5-"



                         tay'  P
                             ,%i;
                            %

                       "e~5r"'"


                                                                          42BO


                                         HP
               1-4.                                      wth
94030rl.30
                                             3-43

-------
      temperature: 400 F (477 K)
      velocity: 3720 fpm (18.9 m/s)
      diameter: 2 ft (0.61 m)

      Receptors
      64 discrete receptors, located as shown in Figure 3-2(b,c)

   The ISCST2 model runs also produced ASCII output files.  Two tables from these output
   files, showing the highest second high three-hour average SO2 and NOX concentrations, are
   reproduced in Appendix D.  As expected, the peak impacts from Wampler-Longacre were
   predicted to occur at the closest SNP receptor (located at UTM 703.58 E, 4270.24 N).
   Examination of the ISCST2 results show that the highest second high three-hour average
   concentrations (during July 1990) were 8.16  |ig/m3 for SO2, and 6.00 |ig/m3 for NOX. The
   modeled peak impacts from this one nearby source was comparable in magnitude to the
   cumulative impacts from the set of PSD sources that were modeled with MESOPUFF II,
   although the impacts were at different times and receptor locations.  For example, the
   maximum three-hour average SO2 concentration during July 1990 for MESOPUFF II sources
   was about 10.5 |ig/m3 (in the northeast corner of SNP), whereas the maximum three-hour
   average SO2 concentration from ISCST2 for just Wampler-Longacre was 11.4 |ig/m3 (at the
   receptor closest to the source).

   Following execution of ISCST2, the results from ISCST2 and MESOPUFF H were merged
   using ISCMPF.  (Note: In order to have compatible binary file formats, ISCMPF must be
   compiled with the same compiler, e.g., Lahey 5.0, that was used to compile the MESOPUFF
   II and ISCST2 programs.)  The ISCMPF program adds the SO2 and NOX results from
   ISCST2 and MESOPUFF H, but summarizes the results for all five MESOPUFF species.
   Only MESOPUFF II is capable of modeling secondary species and providing results for
   species other than SO2 and NOX.  For the secondary species, the statistical summary from
   ISCMPF only includes the PSD increment measures (see Table 3-4), whereas the results from
   PSDPOST include the estimation of the AQRV parameters for extinction and deposition. For
   this reason, the PSDPOST output is preferred over the ISCMPF output for these species. In
   addition, PSDPOST was designed to read 12 monthly input files per year in order to compute
   annual PSD and AQRV statistics. Since the MESOPUFF II output files were monthly,
   ISCMPF could not be used for annual statistics.  To provide annual statistics, it would be
   necessary to either (1) edit the ISCMPF program to accept  12 monthly files instead of one
   annual file, (2) combine all the binary MESOPUFF II monthly output files  into one binary
   annual file for input to ISCMPF, or (3) run MESOPUFF II for an entire year.

   The results of the model output integration indicate that the highest second high three-hour
   average SO2 concentration (during July 1990) for all sources was 10.44 |ig/m3. Since this
   occurred during the same hour as the highest second high from ISCST2, the contribution
   during this hour from MESOPUFF II sources (sources other than Wampler-Longacre) can be
   computed as 2.28 |ig/m3 (10.44 - 8.16). Similarly, the highest second high three-hour average
   NOX concentration for all sources was 6.24 |ig/m3. This also occurred during the same hour as
   the highest second high from ISCST2, so the contribution during this hour from  only
   MESOPUFF II sources can be computed as 0.24 |ig/m3 (6.24 - 6.00).
94030rl.30
                                          3-44

-------
   The results of this exercise demonstrate that it was possible to combine the SO2 and NOX
   results from MESOPUFF II and ISCST2 using the ISCMPF program.  Some concerns were
   raised regarding the ISCMPF program, however. The program could be more general.
   Currently, the program is hardwired into accepting only annual average MESOPUFF n output,
   and the user must provide ISCST2 output files for four averaging times (1-hour, 3-hour, 24-
   hour, and annual).  If, for example, only  1-hour ISCST2 output were available, then the
   program should be flexible enough to compute the data for the other averaging periods (as it
   does for the MESOPUFF II data). In this case, since we only had 3-hour output from
   MESOPUFF II, it was necessary to alter the ISCMPF program to accept and process 3-hour
   data. Additionally, the program has the ability to produce tables of maximum concentrations
   and second high concentrations for each receptor, and a table of N-maximum concentrations
   (these three ISCMPF output tables corresponding to 3-hour average SO2 concentrations are
   shown in Appendix D). However, the PSD increment parameters required for SO2 are the
   highest second high (3-hour and 24-hour averages) concentrations (see  Table 3-4). The N-
   maximum table does not provide this result, so one is forced to find the highest value on the
   table of second high concentrations in order to ascertain this required PSD increment
   parameter.  It is recommended that the ISCMPF program be revised in order to be more
   flexible regarding input files, and that it be able to readily display the necessary PSD
   increment parameters.

   An additional minor difficulty in using the ISCMPF results is that the coordinate locations for
   the discrete receptors are input from the MESOPUFF n output file in grid cell units.  These
   grid cell coordinates are later scaled into meters, however, the appropriate UTM coordinate  of
   the modeling grid origin needs to be added to convert the locations into proper UTM
   coordinates.  The receptor coordinates shown in the output file (see Appendix D) are almost
   meaningless. Since the receptor numbers (indices) were also provided, it was necessary to
   refer to the discrete receptor list from ISCST2 in order to determine the actual locations.

   For future users of MESOPUFF II and ISCST2, one option for integration of results would  be
   to obtain the ISCMPF program (it may be made available on the SCRAM BBS), and edit the
   program to match the MESOPUFF n output structure (i.e., monthly files, 3-hour data). A
   second  (probably simpler) option would  be to  revise the PSDPOST program to read an entire
   year of ISCST2 3-hour averages and add them to the MESOPUFF II (SO2 and NOX) 3-hour
   average concentration results before processing as before.
94030rl.30
                                          3-45

-------
                4  SOURCE IMPACTS AS A FUNCTION OF DISTANCE
   A set of 24 MESOPUFF II simulations were performed with sources at varying distances from
   Shenandoah National Park (SNP) to provide some insight into the relationship between
   distance from SNP and potential PSD and AQRV impacts.  Four months of one year
   (representing each season of one of the years simulated with the PSD sources) were modeled
   under six different source scenarios.  Each source scenario included a number of identical
   hypothetical point sources placed on a ring at a constant distance from the park. The number
   of sources was proportional to the ring circumference; in other words, the linear density of
   sources was held as constant as possible.

   This section describes the preparation of input data for this "ring source" analysis, the
   application of MESOPUFF II, and analysis of modeling results.  This exercise provided an
   opportunity to utilize the MESOPUFF II post-processor, MESOFILE, which was not used in
   the PSD source analysis described in the previous section. The results are presented as
   maximum impacts as a function of distance, and as maximum impacts normalized to the
   emission strength and number of sources.
   INPUT DEVELOPMENT

   For the ring source analysis, the modeling domain, meteorological input files, land use, and air
   quality input files were identical to those described in Section 3 for the PSD source analysis.
   The modeling episode for the ring source analysis included the months of January, April, July,
   and October of 1988.  The months were selected to include one representative month from
   each season.
   Development of Ring Source Data

   Six non-circular rings of hypothetical point sources were developed to encompass SNP at a
   range of distances from the SNP "spine". The park's spine was defined as a line segment
   connecting the most northeastern and most southwestern modeling receptors. The source
   rings were established at 50, 100, 125, 150, 175, and 200 km from the spine. Each ring
   consisted of two semicircles connected by two line segments, the segments being identical in
   length and parallel to the SNP spine (Figure 4-1).

   The distribution of hypothetical point sources around each ring was determined in the
   following manner. First, SO2  and NOX emission rates for all available PSD point sources
   between 50 and 200 km of Shenandoah NP (see Table 3-2) were summed to obtain total
94030rl.40
                                           4-1

-------
                               620
ISO
                                          1-	-I ; j ;  ;  I  I   '  I., ,,J. . I  I


                                11                      tl




                                    1! Computational

                4-1, MCSCHPUiFF                   far
94030rl.40
                                                 4-2

-------
   SO2 and NOX emission rates of 3330 g/s and 2480 g/s, respectively. Based on these total
   emissions and the total area covered by the PSD sources in Table 3-2, an average emission
   density (emission rate per unit area) was determined. For each ring, this emission density was
   multiplied by the area of an annulus extending ħ25 km from the ring to determine the total
   emission rates for the ring.

   The next step was to determine how many individual sources to place on each ring. In order
   for the results to be somewhat representative of typical PSD sources, average SO2 and NOX
   emission rates were calculated for the PSD sources in Table 3-2. The average emission rates
   were 123 g/s for SO2 and 92 g/s for NOX, although the emission rates varied over two orders
   of magnitude.  In order to obtain a 1-source increase per 25  km increase in ring distance from
   SNP, slightly higher emission rates of 181 g/s for SO2 and 135 g/s for NOX were selected for
   the hypothetical sources.

   The emission rates and stack parameters used for the hypothetical sources are summarized in
   Table 4-1.  The stack parameters are average values from the PSD source data in Table 3-2.
   Emission rates of primary sulfate were specified at 3 percent of SO2 rates, with an additional
   1.5 factor to account for the larger sulfate molecular weight. Source characteristics of each
   ring are summarized in Table 4-2; the locations of the sources are shown in Figure 4-1.
   Non-gridded Receptors for the Ring Source Analysis

   The ring sources were arranged in concentric ovals around Shenandoah NP.  Thus, only the
   SNP non-gridded receptors, and not those at JRFW, were used to assess impacts. For the July
   and October ring source simulations, only the 55 SNP receptors were included in the
   MESOPUFF II input files. The locations of the SNP receptors were the same as those
   described in Section 3. For the January and April ring source simulations, the JRFW
   receptors were inadvertently included in the input files, but were excluded in the analysis of
   results.
   APPLICATION OF MESOPUFF II FOR RING SOURCES

   A total of 24 MESOPUFF II input files were constructed for the ring source analysis,
   corresponding to six distances for each of the four months.  MESOPUFF II ring source
   simulations were performed in batches of six over weekends on a 386/25 MHz PC with a 100
   MB disk drive. The MESOPAC output file for each month, needed as input to MESOPUFF
   II, required 53 MB of disk storage space, the MESOPUFF II executable required 3.3 MB, and
   the output from each simulation required 1.5 MB.  Thus, the total disk space requirements for
   one set of six ring simulations was 65 MB.

   The large size of the MESOPUFF U executable file also required some adjustments to the
   PC. Although the PC used for this exercise was equipped with 4 MB of memory,
   applications such as Windows and network software typically consume a large portion of
   the 4 MB.  In order to run MESOPUFF II, the PC was re-booted from a floppy disk in a
94030rl.40
                                           4-3

-------
    TABLE 4-1. Emission rates and stack parameters for idealized sources.
Stack Parameters
SO2 Emission Rate (g/s)
SO4 Emission Rate (g/s)
NOX Emission Rate (g/s)
Stack Height (m)
Stack Diameter (m)
Stack Exit Velocity (m/s)
Stack Exit Temperature (K)
Value for Idealized Source
181.0
8.1
135.0
81.0
3.97
20.7
425.0
TABLE 4-2. Source ring characteristics
Ring Distance Number of Distance
from SNP Sources Sources
50 km 4 138 km
100 km 6 144 km
125 km 7 146 km
150km 8 148km
175 km 9 149 km
200km 10 150km
Total Emissions (g/s)
Between
S02 NOX
724 540
1086 810
1267 945
1448 1080
1629 1215
1810 1350
94030rl.40
                                            4-4

-------
   stripped-down mode without applications. The consequences of this were that it was not
   possible to run MESOPUFF II from Windows, which would allow the model to run in a
   background mode. Instead, the MESOPUFF II runs were performed overnight and over
   weekends when dedicated use of the PC was possible.  The size of the MESOPUFF n
   executable varies with parameters such as the number of surface and upper air meteorological
   stations, ozone stations, sources, and receptors. The parameters used for the applications
   described here, both the PSD and ring applications, pushed the executable close to the
   maximum size that could be executed on a machine with 4 MB of memory.

   For the months of January, April, and October, CPU times on the 386 were approximately one
   hour per source per month simulated. For July, the CPU times were approximately twice as
   large as the other months.  The reason for this increase in CPU was not investigated, but may
   be due to lower wind speeds in July which kept puffs in the modeling domain for a longer
   period of time. Based on CPU times for the PSD source analysis (Section 3), CPU times
   would be decreased by a factor of 5 to  10 on a 486 PC.

   MESOPUFF II output for the ring source analysis consisted of 3-hour average concentrations
   and wet and dry fluxes of all species.  The results were processed using the MESOFILE
   postprocessor.
   APPLICATION OF MESOFILE FOR RING SOURCES

   The MESOFILE postprocessor was used to obtain 24-hour average and monthly average
   concentrations and monthly cumulative deposition for the ring source simulations. Because
   MESOFILE only processes a single species during each run, batch files were set up to
   execute a series often consecutive MESOFILE runs for each ring scenario. As a result, a
   total of 240 MESOFILE input and output files were generated for the ring source analysis.

   MESOFILE requires the input files to be named infilel.dat and infile2.dat, and creates a
   single output file called file.1st. Thus, batch files are needed to rename the MESOPUFF II
   output files, execute MESOFILE, and rename the MESOFILE output.

   MESOFILE can perform two basic operations: averaging and summing. Averages can be
   specified over any integral number of output intervals. Values can also be scaled by a
   constant prior to averaging or summing. This feature is useful for converting sulfate and
   nitrate concentrations to ammonium sulfate and ammonium nitrate in the modeled PM10
   calculation.

   MESOFILE also creates a binary file called file25.dat. Results from a previous MESOFILE
   run can be  accessed through this file.  This allows for the calculation of parameters involving
   more than one species at a time, such as the combination of sulfate and nitrate into modeled
   PM10 and the calculation of total S and total N deposition.
94030rl.40
                                          4-5

-------
   The values calculated for the ring source analysis were based on the PSD and AQRV
   indicators in Table 3-4. For SO2, MESOFILE was set up to report 3-hour, 24-hour, and
   monthly averages.  For NOX, nitric acid, sulfate (SO4=), nitrate (NO3") and modeled PM10, 24-
   hour averages and monthly averages were produced.  SO4= and NO3" concentrations output by
   MESOPUFF II were multiplied by factors of 1.38 and 1.29 to convert to ammonium sulfate
   ((NH4)2SO4) and ammonium nitrate (NH4NO3), respectively.  Following the Phase 1
   recommendations, the sum of ammonium sulfate and ammonium nitrate was reported as
   modeled PM10.  Total sulfur deposition was calculated by summing wet and dry deposition
   fluxes for SO2 and  SO4= over the month.  Conversion factors were applied to convert SO2 and
   SO4= to a sulfur basis, and to convert 3-hour average fluxes to 3-hour cumulative deposition.
   Total nitrogen deposition was calculated by summing wet and dry deposition fluxes for NOX,
   HNO3, and NO3" over each month. Conversion factors were applied to convert NOX, HNO3,
   and NO3" to a nitrogen basis, and to convert 3-hour average fluxes to 3-hour cumulative
   deposition.

   Other input files required by MESOFILE are the MESOPUFF II output files for
   concentration and wet and dry deposition. The total disk space required for MESOPUFF II
   output files from the ring simulations is 24 MB (1 MB per simulation).  MESOFILE output
   files also occupied  1 MB for each of the 24 scenarios. Individual MESOFILE runs were very
   fast; each set often runs required only about 5 minutes of elapsed time on the 386 PC. Also,
   the MESOFILE executable code occupies only 1.3 MB, and thus can be run without any of
   the special set-up needed for the MESOPUFF II runs.

   MESOFILE output is in the form of text files that contain the averaged concentrations or
   cumulative fluxes for each receptor.  Note that MESOFILE can produce isopleth plots for
   gridded receptors, but not for non-gridded receptors.  MESOFILE also identifies the highest
   value among all receptors for a given averaging period. However, it does not calculate the
   highest value among a set of highest values. For example, in this application, MESOFILE
   provided 30 or 31 sets of 24-hour averages for each species (per month modeled). In order to
   find the highest 24-hour average that occurred during the month, it was necessary to scan the
   output file manually to locate the highest value from  among  the 30 or 31 daily maximum
   values. This was somewhat time-consuming.  It is virtually impossible to use MESOFILE to
   identify the highest second-high value, as defined for PSD analyses.

   The calculation of 4-month averages from the monthly averages was also very time-
   consuming. This was accomplished by importing the sections of the MESOFILE output text
   files containing monthly averages for each receptor into PC spreadsheets, re-arranging the
   data into columns, averaging, and locating the maximum. For each species, 24 blocks of text
   had to be imported and re-arranged.
   RESULTS FROM RING SOURCE SIMULATIONS

   The results obtained with MESOFILE were input into PC spreadsheets to produce the graphs
   shown in Figures 4-2 through 4-22.  The results are presented as the highest value among all
   SNP non-gridded receptors, as a function of source distance from SNP. In the following
94030rl.40
                                           4-6

-------
   discussion, for comparison purposes, the impacts are examined relative to the allowable Class
   I PSD increments listed in Table 3-4. In these comparisons, it is important to bear in mind
   that the sources do not represent actual sources.
   Concentration Impacts as a Function of Distance

   SO2 Concentrations

   Figure 4-2 shows the highest 3-hour average SO2 concentration for each of the four months as
   a function of source distance from SNP.  As might be expected for a primary pollutant, the
   highest impacts occur for the 50-km source ring. The highest 3-hour average concentration is
   20 |ig/m3, close to the allowable Class I PSD increment of 25 |ig/m3.  The simulated 3-hour
   average SO2 concentrations decrease by up to a factor of four between the 50 and 100 km
   rings, but then increase between the 100 and 125 km rings for two of the months. This
   increase occurs because the sources on each ring are located in different positions relative to
   SNP (see Figure 4-1). If a larger number of smaller sources were used, a smoother decrease in
   concentration with source distance would likely result.

   There is no consistent trend for the highest 3-hour average SO2 concentrations to be higher in
   any particular month of the year, although the lowest values were simulated for July for five
   of the six rings.  Modeled conversion of SO2 to sulfate and SO2 deposition are both
   maximized in July. These two effects may account for the lower SO2 concentrations
   simulated for July.

   The ratio of concentration to source strength is commonly termed x/Q, and represents a tool
   for estimating impacts based on source strengths. In the ring source application, 3-hr impacts
   are likely to be primarily the result of a single source, rather than the total  source strength of
   the source ring.  Dividing the maximum 3-hr SO2 concentrations by the single-source SO2
   emission rate of 181 g/s (1437 Ibs/hr) gives x/Q values of 0.11, 0.038, and 0.014 for the 50
   km, 125 km, and 200 km rings,  respectively, where the units are (|ig/m3)/(g/s).  In units of
   (|ig/m3)/(lbs/hr) the corresponding x/Q values are 0.88, 0.30, and 0.11.

   Figure 4-3 shows the highest 24-hour average SO2  concentrations for each month as a function
   of source distance from SNP. Again, the greatest impacts occur for the 50-km  ring, although
   the decrease from the 50-km to the 100-km ring is not as pronounced as for the 3-hour
   impacts.  The highest 24-hour average concentration of 2.8  |ig/m3 is  slightly more than half
   the allowable Class I PSD increment of 5  |ig/m3. The highest concentrations at each source
   distance generally are simulated for October and January, with the lowest concentrations
   generally simulated for July.

   For the 24-hour impacts, it is not possible to determine whether the 24-hour impacts are the
   result of a single source, or the confluence of puffs from two or more sources.  Thus, it is
   difficult to select the most appropriate source strength to use in calculating x/Q
94030rl.40
                                            4-7

-------
    M
 CM


10


€B
oj


I

&wğ.

 I
CO

IS

JC
         25
        20
        10'
         5^
         0-
                    D
                                      C     g

                                      *     *
                                             I
             25   SO   7B   100   125   150   175   200
                "  Jan  A  Apr   *   Jul

           ftan nan     ftt           JWlf, and
94030rl.40
                                     4-8

-------
c*>

 £  25

 OJ

 3

  

 |p  1 Ğ5!


 8j
 *_


 4    !•
 cy
 
-------
   Using the total source strength for the ring would seriously underestimate x/Q, whereas using
   the source strength from a single source might result in an overestimate.

   Figure 4-4 shows the highest monthly average SO2 concentrations. Note that because each
   point represents a different receptor, the highest 4-month average is not determined as the
   average of the four points for each source distance in Figure 4-4.  Figure 4-5 shows the
   highest 4-month average SO2 concentrations. Since one month from each season was
   modeled, the 4-month averages may be used to approximate an annual average.  The highest
   4-month average  SO2 concentration is 0.34 |ig/m3, about one-sixth of the allowable PSD
   increment for the SO2 annual average.

   Comparison of Figures 4-2 through 4-5 shows that the decreasing trend in peak SO2
   concentration with distance is less pronounced as averaging time increases.  Recall that the
   total emissions from each source are constant, but that the total emissions for each ring
   increase proportionally to ring circumference. Short-term impacts are generally the result of
   only one source at a time.  Since the ring sources are all the same size, the 3-hour impacts
   would be expected to decrease as the inverse of the distance from SNP (I/distance).
   However, the longer-term impacts are the cumulative impacts of all the sources. For longer-
   term impacts, therefore, the concentrations normalized by the number of sources would be
   expected to decrease as the inverse of the distance.

   Figure 4-6 shows the per-source highest 4-month average SO2 impacts. Whereas the absolute
   impact decreased by a factor of two from the 50 km ring to the 200 km ring, the normalized
   impact decreased by more than a factor of four. Thus, the normalized impact decreased at
   greater rate than I/distance. This is likely to reflect additional losses of SO2 due to reaction
   and deposition at the greater distances.

   These results for  SO2 suggest that 3-hour impacts are the limiting consideration for sources
   located 50 km from SNP.  For sources of the size used in this analysis, sources located further
   than 200 km  are likely to have small 3-hour SO2  impacts by comparison to sources located
   closer in. However, it is not possible from this analysis to make a blanket statement that  all
   sources located further than 200 km can be considered negligible in terms of 3-hour SO2
   impacts. The largest source in the PSD analysis (see Table 3-2) has SO2 emissions that are ten
   times as  large as the per-source emissions used in the ring analysis. For sources that size or
   larger, 3-hour SO2 impacts at 200 km or greater distances may approach the allowable
   increment.

   For sources located 100 km or further from SNP, these results suggest that 24-hour impacts
   may be the limiting consideration.  Sources located at 100 km from SNP with emissions
   double those used in this analysis would  produce maximum simulated 24-hour SO2
   concentrations near the allowable increment. Simulated 4-month average SO2 concentrations
   were sufficiently  low that the annual average PSD increment for SO2 does not appear to be a
   limiting consideration at any distance for the range of source strengths addressed in this
   analysis.
94030rl.40
                                           4-10

-------
w
E
™%H
CD -n M
a 0,4'"
ğ—
o"
§> 0.3-
aver
o p
L* fo
X
n-
*

*
D
Ğ A
fe
_,„„, 	 *f -„„„„ 	 	
D 4
• m
n I
* 4 i


                50    75   100  125   150   175   200
               Jan   A   Apr   *   M   o   Get
          ring      for          July, aoJ October,
94030rl.40
                                    4-11

-------
   0,36
75   100
Distance
                           125  150  175
                          from SNP (km)
200
         nog     far
94030rl.40
                                4-12

-------
                      75   100  128  150  175  200
             nog     for
94030rl.40
                                   4-13

-------
   Nox Concentrations

   Figure 4-7 shows the highest monthly average concentrations for NOX.  (Plots of shorter-term
   NOX concentrations were not prepared because there are no short-term PSD standards for
   NOX.) Comparison of Figure 4-7 to Figure 4-4 for monthly average SO2 illustrates the faster
   rate of chemical decay for NOX. At 50 km, NOX concentrations are generally on the order of
   60 percent of the SO2 concentration for January, April, and October, reflecting the ratio of the
   emission strengths of SO2 and NOX in the ring source input files. NOX concentrations at 50 km
   are 40 percent of SO2 concentrations for July, because NOX reacts much more rapidly than
   SO2. At 200 km, NOX concentrations are on the order of 50 percent of the SO2 concentration
   for January, April, and October, and 30 percent of SO2 concentrations for July.

   The highest monthly average NOX concentration was slightly less than 0.3 |ig/m3, or about 11
   percent of the allowable increment of 2.5 |ig/m3 for NOX annual average in Class I areas.  The
   annual average will be lower than the highest monthly average. Thus, much greater NOX
   source strengths than those used in this analysis would be needed for NOX concentrations to
   approach the allowable increment.
   PM10 Concentrations

   As discussed in Section 3, the only PM10 constituents modeled by MESOPUFF II are sulfate
   and nitrate.  In the remainder of this section, modeled PM10 refers to the sum of ammonium
   sulfate and ammonium nitrate, and includes both primary and secondary sulfate. Typical
   sources may emit other primary PM10 components that are not included in this analysis.

   Figure 4-8 provides the highest 24-hour average modeled PM10 concentrations as a function of
   source distance from SNP.  These impacts show a completely different pattern than SO2 or
   NOX.  The highest 24-hour average modeled PM10 concentration of 0.7 |ig/m3 occurs for July
   for the 175 km ring.  This represents less than one-tenth of the allowable PSD increment of 8
   |ig/m3 for PM10 in Class I areas. There are no clear trends in the 24-hour average modeled
   PM10 concentrations with source distance from SNP.  For the months of January, April, and
   October, a weak decrease in maximum concentration with distance was simulated.  July
   modeled PM10 concentrations show a minimum at 125 km and are roughly equal for the 50 km
   and 200 km rings. At the greater distances, July modeled PM10 concentrations are much
   higher than those from the other seasons.

   Figure 4-9 presents the highest monthly average modeled PM10 concentrations as a function of
   source distance from SNP.  The highest impact occurred for July for the 150 km ring.  The
   seasonal effect is very pronounced for the monthly average modeled PM10, with July
   concentrations more than twice those in January, and April and October falling somewhere
   between the two.
94030rl.40
                                           4-14

-------
       0.3
 CT
  I  „Ğ
  05 0.25s
   X
  O    ^Ğ
  Z    0.2'
0)
OQ
ĞB
lğ
     0.1S
  g    0.1
  E
  CO
 .5?

        0-
                   n
                   if
                              °
                                          D
                                              m
                                                     -*-
            25   50    75   100  125  150  175  200
                Jan   A   Apr  *  Jul   n  CM
                       acinulauWl
           ibg      for JgHBuy,         and
94030rl.40
                                     4-15

-------
1
eğT"
i
""*** *t
o

°-

I 1
| 0.4-
3
Ğ****
£ 0,2-
.0*
ft.



*

A

j(?
E A
	 n_ _ • 	 £ 	 	
DC *
C "
• Q



            25    50    75    100  125   160   175   200
                                      SNP (km)
              "   Jan  A  Apr   *   Jul    o   Oet
                        sonibmd 24-tow avenge FM10 ooncenaMiniB
            ring gouceet fa ftauajy, April, My, uri October,      for tfau
        FM10 is define.! as [be 100 of        sulfelc and anuaofdum
94030rl.40
                                      4-16

-------
u.ir
ğUğ'ftuğ
n
E
*••Ğğ
B
g 0,15-
S*
Q-
Ğ
o^^
OS
Si 0.1-
dS*
Ğ0
i
o
E O.OS-
1
x:
cğ
in
n-
Hi sfe

ffc"



A A


A
A


*
o o * e
D g " " •
•






                                     t      i      i
                    50   75  100  125  150  175   200
Jan  A  Apr  *   Jui
                                              * oct
                4-f,                         AIu
          tern ni|                  July, nri           For ttti
              is      a tint                  Mad
94030rl.40
                                    4-17

-------
   In Figure 4-10, the seasonal dependence of modeled PM10 concentrations is illustrated for the
   125 km ring. Figure 4-10 clearly shows that sulfate formation is maximized in July in the
   MESOPUFF U model, whereas modeled nitrate concentrations are relatively constant year
   round. Modeled sulfate and nitrate concentrations were nearly equal for all months but July.
   This is contrary to measured PM10 data for SNP and other locations in the eastern U.S., where
   sulfate concentrations are typically three times as large as nitrate concentrations.  Two factors
   are likely to be responsible for this apparent inconsistency. The high background ammonia
   concentration of 10 ppb used as a default value in MESOPUFF II will result in an
   overestimate of aerosol nitrate concentrations. However, the lack of a parameterization of
   rapid in-cloud sulfate formation in MESOPUFF U may lead to an underestimate of sulfate
   concentrations.

   Figure 4-11 presents the highest 4-month average modeled PM10 concentrations.  The 4-month
   average peaks for the 125 km ring. The peak value of 0.11 |ig/m3 is only 3 percent of the
   allowable PSD increment for annual average PM10 concentrations in  Class I areas.
   Comparison of Figure 4-11  to Figure 4-5 for SO2 shows that, at a distance of 50 km,  9 percent
   of the modeled total sulfur is in the form of sulfate.  At 200 km, 16 percent of the modeled
   total sulfur is in the form of sulfate.

   It is important to recognize that there is a primary component to the modeled PM10
   concentrations  shown in Figures 4-8 through 4-11, since primary SO4= emissions equal to 3
   percent of SO2  emissions (as sulfur) were assumed.  Although it is not possible to distinguish
   the primary sulfate from the secondary sulfate, an estimate of the magnitude of the primary
   contribution can be obtained from the SO2 concentration at the same  receptor over the same
   averaging period. For the highest monthly average modeled PM10 concentrations (Figure 4-9),
   up to 50 percent of the sulfate at 50 km is estimated to be primary. At 200 km, approximately
   25 percent of the monthly average sulfate  is primary. Primary sulfate is less of a contributor
   to the 24-hour averages, contributing 15 percent or less of the highest 24-hour average sulfate
   concentrations  at 50 km.

   Figure 4-12 shows the normalized, per-source, 4-month average modeled PM10
   concentrations. When the modeled PM10 concentrations are normalized in this manner, a
   decreasing trend in modeled PM10 concentration with source  distance is observed.
   Normalized 4-month average modeled PM10 concentrations at 200 km are 37 percent of those
   at 50 km.

   These results suggest that 24-hour average PM10 impacts, rather than the annual averages, are
   the limiting consideration for PSD purposes.  However, considerably greater source strengths
   than those used here would be needed for modeled PM10 concentrations to approach the
   allowable increment for Class I areas. At 175 km, roughly ten times  the source strength used
   here would produce 24-hour average impacts  near the allowable increment. (According to
   Table 3-2, at least one source of that size is located at roughly 175 km from SNP). Because
   distances of only 200 km were used in this analysis, it is difficult to draw firm conclusions
   about the modeled PM10 impacts.  The cumulative impacts from numerous sources both closer
   and further than 200 km from SNP might well be  considerably greater than the ring source
   impacts modeled in this analysis.
94030rl.40
                                           4-18

-------
  X
                Jan
                                 Month
Oct


                                                      For ttus
              ii       • the MM                and
94030rl.40
                                       4-19

-------

     E
5
Q_
 ii
 O>
 S
 (D

 i


 c
 o


4
*#•*
 IP1
 01
JC

X
          0.1

        0.06
        0,04'
        0,02'
                      75   100  125  150  175
                                                        200



                     n      aod amtuunium
                                       4-nĞaij      FM, c,
                                       t^_ .	_J^. 
-------
    s
    Q.


    9
    m

    4~ğ
    C
    Q
   13
    o
    c
    EP

   X
         0,03

0.02

    i    0.01

                          75   100  125   150  175
                                             200

                                         njtrntii
                                      few     lor tfais       PMj,3 is

                    HI tho    of             and
94030rl.40
                                       4-21

-------
   An important AQRV associated with PM10 is visibility.  MESOFILE is not capable of
   producing visibility estimates as outlined in the Phase 1  recommendations, where extinction is
   a function of relative humidity.  Thus, visibility impacts were not assessed for the ring source
   analysis.
   Deposition Impacts as a Function of Distance

   S Deposition

   Figure 4-13 shows the highest monthly cumulative S deposition. The highest monthly total S
   deposition of nearly 0.05 kg/hectare occurred for July for the 50 km ring.  Modeled S
   deposition decreased with source distance from SNP, although the decreasing trend was very
   weak for October. At all distances, the highest S deposition occurred in July, and the lowest
   in January.

   Figures 4-14 through 4-17 show the contributions of the four components of total S deposition
   (wet and dry deposition of SO2 and SO4=) for each month. Dry SO2 deposition accounts for
   most of the modeled S deposition in all months. Deposition of SO4= is small for all months
   except July. The main reason for this is that most of the total sulfur remains in the form SO2
   at all distances modeled.  Thus, although deposition velocities  for SO4= may be higher than
   those for SO2, deposition fluxes are higher for SO2.

   Wet deposition shows greater random variability than dry deposition. This makes sense, as
   wet deposition requires both a puff and precipitation to be present at the same time at a given
   receptor. Thus, trends in wet deposition with source distance are somewhat obscured by the
   random variability.
   N Deposition

   Figure 4-18 shows the highest monthly cumulative N deposition. The highest monthly total N
   deposition of 0.013 kg/hectare occurred for the 50 km ring for July.  Modeled N deposition
   decreased slowly with source distance from SNP. At all distances, the highest N deposition
   occurred in July, and the lowest in January. Total N deposition was  less than half the modeled
   total S deposition for  all months and distances.

   Figures 4-19 through  4-22 show the contributions of the five components of total N deposition
   (wet and dry deposition of HNO3 and NO3", and dry deposition of NOX) for each month.
   MESOPUFF II assumes zero wet deposition for NOX. The majority  of modeled N deposition
   in July is due to dry deposition of HNO3.  In the other months, NOX dry deposition is
   important, and NO3" deposition can be important as well. Considering the high  aqueous
   solubility of HNO3, the modeled wet deposition of HNO3 appears surprisingly low.
94030rl.40
                                           4-22

-------
.5

"5
o
a.
9
~O
 1




'I
 §

 E

 15
 0)
      0.03-
      0.01
         o-
                               .
                              -e-
                                               i
                                                   0
            2S   50    7S   100   125  150  175  200
             ğ   Jan   A   Apr  *  Jul   °   Oct
                      iia§      rĞ              and
94030rl.40
                                 4-23

-------

               50   75   100   125  150   175   200
        Wet SO4 jf Dry SO4 ill Wet SO2      Dry SO2
         4-lC                    198K BBiJ S      ftoei

94030rl.40
                               4-24

-------
         o.os
      m
           o
              50    75   100  125   ISO   175   200
        Wet     ^B Dry SO4      Wet SO2


94030rl.40
                                4-25

-------
                  0,05
                         5O    78   100   125   150   175   200
                                                SNP
                  Wet
Dry            Wet
Dry
              HKH3RB 4-1 ft,          Of       July 1*88    S
94030rl.40
                                        4-26

-------

              50
           75   100  125   150  175  200
WİtSO4
                                 Wet SO2 Egg Dry SO2


94030rl.40
                             4-27

-------
s
§
8-
•o
Z
a
2 0.01 -
>H
f
o
£
1
X
ft-


Ğc
*
*


sK
X
A
M
	 Q t
HD ffl A
D n

                    25
50   75   100   125  150  175  200

                        Jan   A  Apr   *   Jul    °   Oct
                      4-JI,  Highest sLmulaiid moothly ctimulafrvc     N
                            nnfl NUBMH for Janmry. April, My, and
94030rl.40
                                       4-28

-------
   9

   a
   CO
    g
   32
   "(ft
    O
    CL

   13

   Z
   J


   18
        0.01

                    75    100  125  150  175
      200
            Wet NO3
                           Wet
Dry

94030rl.40
                                 4-29

-------

             50     75   100  125   150   175
      200

           DryNOjc

Dry
     sources
94030rl.40
                                   4-30

-------


                50   76   100   125   150  17S
200
                             Wet
                             DryNO3

                                                      rine
94030rl.40
                                 4-31

-------

      1
      o
      *
      en

      if
      Ğ"!••ğ




      .1
      •*•



      I

      QL

      0}

      T3
      I
      *•*
      *-/

      a
          0,01


                50
75   100   125   150   175

                   (km)
                                                  200
         IHI Dry
             Wet NO3
       DryNOS
                      Dry


94030rl.40
                                4-32

-------
   The equilibrium between HNO3 and NO3" affects the results presented here, both for the
   modeled PM10 concentrations and the nitrogen deposition. The ring source analysis utilized
   the default value for background ammonia of 10 ppb. This value is likely to be too high,
   especially for winter. When ammonia concentrations are high, the nitrate equilibrium favors
   the formation of aerosol nitrate. As a result, modeled aerosol nitrate values may be too high.
   For N deposition, it is less clear what the effect of high background ammonia would be. Dry
   deposition is faster for HNO3 than it is for NO3", but wet deposition is faster for NO3". In
   particular,  snow is assumed in MESOPUFF II to scavenge particles but not gases.  Therefore,
   if background ammonia is high, NO3" deposition will be overestimated and HNO3 deposition
   will be underestimated.  The net effect on N deposition may be small.
   PROBLEMS ENCOUNTERED

   Although MESOFILE is flexible, it cannot provide results in the form needed for the
   PSD/AQRV impact analyses as described in Table 3-4.  It is not set up to identify second-
   highest impacts, and cannot calculate visibility impacts according to the formula outline in the
   Phase 1 Recommendations.

   A minor error in the MESOFILE code was noted when we attempted to obtain average
   concentrations for a time period other than the full month simulated.  The problem occurs
   with the specification of the day and hour. Apparently, the start day must be specified as the
   start day of the MESOPUFF II simulation, regardless of what day MESOFILE is to be run
   for. For example, in order to obtain an average concentration for day 3 of a simulation that
   began on day 1, hour 0, and continued for one month, one must specify the day as 1 and the
   hour as 48.  Specifying the MESOFILE start date as 3 and the start hour as 0 will result in an
   error message.  The MESOFILE code was not modified to alleviate this problem.
   CONCLUSIONS

   The ring source analysis illustrates some of the effects of source-receptor distance on air
   quality and deposition impacts. For the primary species, SO2 and NOX, peak impacts drop off
   rapidly with distance.  MESOPUFF n results suggest that sources of the size used for this
   analysis (183 g/s SO2), located 50 km from SNP, are capable of producing 3-hour SO2 impacts
   close to the allowable PSD Class I increment.  For the secondary species, SO4= and NO3",
   impacts did not show a decreasing trend for sources between 50 km and 200 km from SNP.
   The 4-month average modeled PM10 peaked at 125 km and was slightly lower at 200 km than
   it was at 50 km.  Although modeled PM10 concentrations were well below allowable PSD
   Class I increments for all rings, the lack of a clear trend suggests that sources beyond 200 km
   may need to be considered in some cases in order to assess the impact upon PM10 and related
   parameters, such as visibility.
94030rl.40
                                           4-33

-------
                                     5  SUMMARY
   STATUS OF THE MESOPUFF II MODELING SYSTEM

   This project began with the retrieval of the MESOPUFF n modeling system from the EPA
   SCRAM electronic bulletin board.  As discussed in Section 2, two minor changes to these
   files were made as a result of execution of the test case. The updated files have been released
   to the SCRAM bulletin board.

   In order to run MESOPUFF II for the demonstration application, several other modifications
   were made to the modeling system.  These changes primarily relate to formats and the ability
   of the models to read and process input data.  None of these changes are required to execute
   the test case, however these MESOPUFF n modifications are necessary for future applications
   with the SNP modeling domain. These model changes were all presented and discussed in
   Section 3; they are summarized below:

         PMERGE

             Several  changes were made to accommodate two additional data quality flags that
             were present in the data but were not understood by PMERGE.

         MESOPAC

             The format designator for the statement reading the upper air height field was
             changed from a length of 5 characters to 6 characters, as all preprocessing carried
             6 characters.

             A quality assurance routine was changed to check for X characters rather than
             dash (-)  characters to indicate unlimited ceiling heights, because standard CD 144
             format supplies X characters rather than dashes.

         MESOPUFF II

             The format designator for the statement reading ozone data was  changed to
             accommodate 200 ozone stations rather than 50.

         •   In the PARAMS.PUF file, the maximum number of ozone stations was increased
             to 200, and several other changes were made to optimize the parameters for this
             application.  Changes to the PARAMS file require recompilation of the model
             source code.
94030rl.50
                                           5-1

-------
   DEMONSTRATION APPLICATION

   A demonstration application of the MESOPUFF II modeling system was conducted for
   Shenandoah National Park and James River Face Wilderness.  For this application, a five-year
   data set (1988-1992) of MESOPUFF H meteorological inputs was developed. MESOPUFF H
   was applied for the years 1988-1990 for a condensed set of existing PSD sources located
   within 200 km of SNP.

   The primary purpose of this exercise was to document the implementation process rather than
   to obtain a set of modeling results. As described in detail in Section 3, a number of problems
   were encountered and resolved during the model application process. Additional processors
   were developed to streamline data preparation, modeling, and postprocessing systems.

   Resources Consumed

   For this project, five years of raw meteorological data, including surface, upper air, and
   precipitation data, were acquired. Five years of ozone data were also needed, but much of that
   data were already available in-house. Identifying data suppliers and ordering the needed data
   consumed about 20 labor hours.  The total cost of the data was under $3,000. The data were
   supplied on 8-mm tape cartridges within 2 weeks of placing the order. These tapes were read
   onto our TRACE/Unix mainframe computer.

   Development of the five-year set of meteorological inputs consumed 460 labor hours over
   four months. During much of this time, two 486 PCs were dedicated for use in the
   meteorological data processing. In addition, several data processing steps were not feasible on
   a PC and were conducted on a mainframe computer. The mainframe computer was also used
   for temporary storage of the 3.2 gigabytes of output produced by MESOPAC and for
   transferring the output to 8 mm tape cartridges.

   Application of MESOPUFF II was much less resource-intensive than application of
   MESOPAC. Exercise of MESOPUFF U for three years required 45 labor hours and three
   weeks of dedicated usage of one 486 PC.  Again, the mainframe computer was needed for
   temporary storage of the MESOPAC output files.
   Processors Developed

   Six pre-processors and two post-processors were developed for this project.  The names and
   functions of each are summarized below.

   Two pre-processors extract information from raw data files:

         •   RESFC reads TD-3280 hourly surface data and extracts data for a given
             space/time window.
94030rl.50
                                           5-2

-------
             REUPR2 reads TD-6201 upper air data and extracts data for a given space/time
             window.

   Additional pre-processors are used to prepare data:

             TOCD144 reformats TD-3280 surface data into CD144 format and fills in
             missing data.

             FILLUPR reads the upper air data output by REUPR2 and fills in missing data
             according to a specified set of rules (see Section 3).

             PARSE splits the surface and upper air data into separate files for each station, as
             required by MESOPAC.

         •   GRIDDIT calculates and formats station grid coordinates for surface, upper air,
             and precipitation stations.

   Two post-processors generate air quality summaries from MESOPUFF n results:

             RHPOST extracts relative humidity data from the MESOPAC output file and
             prepares 3-hour averages for input to PSDPOST.

             PSDPOST  calculates specific PSD and AQRV parameters for one year of
             MESOPUFF II output for non-gridded receptors.

   These processors will be made available to future MESOPUFF n users via the SCRAM BBS.
   A description of each program and a guide for application are presented in Appendix C. The
   use of these processors, combined with review of the demonstration MESOPUFF n
   application, will greatly facilitate future MESOPUFF n applications.
   Data Files Created

   For future applications utilizing the same modeling domain, none of the meteorological pre-
   processing steps described here need be repeated.  MESOPUFF n can be run with different
   sources and/or receptors using the MESOPAC output files and the ozone data files developed
   for this demonstration modeling exercise.  However, the MESOPAC output files are rather
   large, so it may be a simpler process to start with the MESOPAC input files (processed
   surface, upper air, and precipitation data) constructed for this domain, and then run
   MESOPAC and MESOPUFF II for each modeling period.
   Deviations From Phase 1 Recommendations

   A number of deviations from the IWAQM recommended approach were used for this
   demonstration MESOPUFF II application. Many of these deviations were made due to the
   limited resources available for conducting the demonstration modeling effort. For example,
   since the MESOPUFF n model was applied on a PC, it was necessary to perform monthly
94030rl.50
                                          5-3

-------
   runs (for a subset of the PSD source inventory), for which data handling was cumbersome. If
   a faster computer system were used, the runtime could be reduced, enabling multiple runs,
   more sources to be modeled, or a higher puff release rate. If a larger computer system were
   used, storage concerns could be minimized, which might eliminate the need to perform
   monthly runs (i.e., an entire year could be run at once). A PC-based system with one to two
   gigabyte hard drives and a removable optical drive may lessen many of the obstacles that were
   encountered.

   The deviations from the interim IWAQM recommendations are identified below:

   In order to reduce the volume of MESOPUFF II output, 3-hour average concentrations and
   fluxes were output rather than 1-hour averages, as required  by the IWAQM interim
   recommendations. However, the minimum averaging time  necessary to compute the desired
   PSD and AQRV impact measures (see Table 3-4) is three hours.

   To reduce CPU time and the size of the MESOPUFF II executable code, a puff release rate of
   1 puff per hour rather than 4 puffs per hour was used. If possible, a realistic assessment
   (following the interim IWAQM recommendations) should use 4 puffs per hour. However, if
   CPU time is limited, a release rate of one puff per hour should be adequate (i.e., will produce
   similar results), especially for the long travel times associated with secondary air pollutants.

   Both of these parameters (output averaging period and puff release rate) are specified in the
   MESOPUFF II user input file and can be changed by future users. However, the RHPOST
   and PSDPOST processors are configured for 3-hour averages; PSDPOST cannot be used
   with any other averaging interval without modification. In  addition, if the puff release rate is
   increased beyond 1 puff per hour, the maximum allowable number of puffs may be exceeded.
   This would  necessitate  changes to the PARAMS.PUF file, which would in turn require
   recompilation of the MESOPUFF II executable code.

   The collected source data represents only a subset of the actual set of sources that would be
   required in order to properly assess the PSD impacts on SNP. As was discovered, multiple
   triggering dates apply to various types of sources in each of the states surrounding a particular
   receptor.  In addition, different inventories are needed to assess PSD increments than those
   required for AQRV impact analysis (see Appendix B). For future MESOPUFF II
   applications, it will be necessary to identify the multiple inventories necessary for assessment,
   and then apply the MESOPUFF n model for each inventory. In addition, for the
   demonstration modeling exercise, many sources were consolidated (to reduce computer time).
   In order to follow the interim IWAQM recommendations, all sources must be modeled
   individually (i.e.,  consolidating sources would not be an acceptable approach for future
   regulatory MESOPUFF II applications).

   A large set of receptors was obtained from the state of Virginia for SNP and JRFW. A subset
   of these receptors were used for the demonstration modeling exercise. A number of the
   receptors  were spatially close together, so there is arguably justification for removing some of
   these, however, this reduction of receptors was done largely to make the size of output files
94030rl.50
                                           5-4

-------
   more manageable. All sensitive receptors (as defined by the FLMs responsible for the Class I
   area in question) should be included in future regulatory MESOPUFF II applications.

   For assessments of multiple sources, the interim IWAQM recommendations require that two
   five-year MESOPUFF II modeling exercises be conducted; one run using all sources to
   determine impacts to secondary NAAQS pollutants (secondary particulate matter) and
   AQRVs (visibility and deposition), and a second model run using only sources beyond 50 km
   from a receptor for SO2 and NOX.  To simplify the analysis (and reduce resources required),
   only one run was  performed for the demonstration modeling exercise (for three years) using
   sources beyond 50 km of Shenandoah National Park for all pollutants.

   Primary sulfate emissions are allowed in the MESOPUFF II modeling system, however these
   emission rates are generally not well documented and most emission inventories do not
   include primary sulfate emission rates. For the demonstration modeling exercise, it was
   assumed that primary sulfate emissions were 3 percent of the SO2 emissions.  The interim
   IWAQM recommendations do not specifically address this issue.

   Another issue that is not specifically addressed in the interim IWAQM recommendations is
   the procedure for filling in missing meteorological data.  The approach developed for the
   demonstration modeling exercise (and applied in the preprocessing programs;  see Appendix
   C) represents sound engineering practice for filling in missing data.  Other approaches are
   possible, including various other interpolation schemes.

   For the demonstration modeling exercise, background ozone data were prepared to represent
   the seasonal variation in observed background concentrations. MESOPUFF n uses a default
   concentration (80 ppb), however users may supply their own data. The interim IWAQM
   recommendations do not specify whether the default is acceptable, however it was our
   determination that seasonal observations should be used instead.  This decision was supported
   by IWAQM.

   As a practical consideration, relative humidities (used to estimate extinction) were limited to
   95 percent.  The extinction efficiency/relative humidity relationships that were employed are
   not valid at such high relative humidities.  Relative humidity is not typically measured
   accurately above 95 percent. In addition, such high relative humidity conditions are likely to
   involve sufficient water condensation for cloud formation, and light extinction (and hence,
   visibility) in the presence of obscuring clouds is generally not regulated. This adjustment was
   made to improve the relationship between concentration and visibility impacts, however the
   interim IWAQM recommendations do not include this adjustment.

   Finally, the ISCST2 model was only exercised for one month in order to demonstrate the
   process of integrating the ISCST2 results with MESOPUFF H results. The IWAQM interim
   recommendations require that 5 years be simulated with MESOPUFF n, and that the Gaussian
   model (ISCST2)  be modeled for the same 5-year period.  The ISCST2 and MESOPUFF H
   model results for  all periods simulated are to be integrated.
94030rl.50
                                           5-5

-------
   Process Review

   During the course of performing the demonstration application, the project team received
   considerable guidance and technical review from members of the IWAQM. An initial work
   assignment was developed by EPA OAQPS that called for review and testing of the
   MESOPUFF II system, followed by development of a demonstration modeling exercise. A
   modeling protocol was developed after consultation with the IWAQM team, in which issues
   such as modeling domain, period to be modeled, receptor locations, and sources to be
   modeled, were presented, discussed and resolved. The IWAQM members who represent
   Federal Land Managers (NPS and NFS) provided technical guidance with respect to
   interpreting model results, i.e., determining the parameters to be used to measure PSD and
   AQRV impacts at Class I areas.  The IWAQM representatives from EPA Region HI were
   contacted for assistance in collecting source data. Each of the individual states within Region
   in were asked by Region HI to supply emission source data (we were able to receive source
   data from all but one state within about six weeks). The IWAQM representatives from the
   Virginia Department of Environmental Quality provided the receptor locations for
   Shenandoah National Park and James River Face Wilderness (and the source data for
   Virginia).

   The IWAQM group assisted the project team in defining the  scope of the demonstration
   analysis. It was the primary objective of the work assignment to carry out an analysis using
   the IWAQM Phase 1 Recommendations in order to demonstrate and improve the modeling
   system.  The IWAQM group's original intent was to exercise the model in the most realistic
   manner as possible in order to discover any and (hopefully) all the problems that would be
   faced by potential future users of the system.  Therefore it was decided that we would attempt
   to perform  a meaningful analysis with the demonstration modeling exercise. An analysis that
   would consider the impacts from all PSD sources impacting SNP and JRFW would be
   informative, in that the results would indicate the total amount of the allowable PSD
   increment that has been consumed by sources that were either constructed or modified since
   the enactment of the PSD rules in 1977. However, it was soon readily apparent, after
   discussions with officials responsible for PSD regulations, that the determination of PSD
   sources is not straightforward; there are multiple triggering dates that are applicable for
   various classes of sources, and furthermore, AQRV analyses should consider a separate group
   of sources.  The scope of the analysis was modified so that the demonstration modeling
   exercise would evaluate the PSD and AQRV impacts of an arbitrary group of PSD sources,
   consisting of most or all sources for which data were made available.

   Because the demonstration modeling scope was altered in this way, the IWAQM group
   welcomed an additional change in project scope whereby the 60 monthly MESOPUFF II
   simulations were reduced to 36 simulations.  Since the resources had been allocated for 60
   months of simulations, the remaining 24 monthly simulations were used in an additional
   analysis to  examine the potential for PSD and AQRV impacts as a function of source distance.

   Numerous technical memoranda were distributed to the IWAQM members seeking input and
   approval regarding modeling issues, including results  of the review of the example problem,
94030rl.50
                                           5-6

-------
   recommendations for improvements to example problem distribution files, plans for
   meteorological data collection and preparation, receptor selection, source selection,
   development of the modeling plan for the distance vs. impacts analysis, and processing of the
   model output. The IWAQM representative from EPA OAQPS acted as the EPA work
   assignment manager for the project, overseeing the technical and administrative details of the
   work, and coordinating the communication between the project team and the IWAQM
   membership.

   For future applications of MESOPUFF n using the IWAQM Phase 1 recommendations, the
   first step is to define the scope of modeling work by specifying the sources to be analyzed and
   the receptors of interest. It is necessary to determine what technical  questions are being asked
   of the regional air quality model and how regulations might dictate the objective of the
   model's application.  It will be necessary to consult with the FLMs and/or other applicable
   regulatory authorities, such as local State or regional EPA representatives, to establish the
   recommended procedure for selecting sources and receptors for the given application (i.e.,
   PSD permit application, AQRV analysis, etc.). The allowable cumulative PSD increments for
   Class I areas have been established (see Table 3-4, for example), however the acceptable
   levels for AQRV impacts have not been identified. It may be sufficient to assess the
   incremental impacts from a single new (or modified) source in the context  of current
   conditions, or it may be necessary to evaluate the cumulative impact of all sources.  Guidance
   from the FLMs will therefore be required to determine how to evaluate the AQRV impacts
   from a new (or modified) source in Class I areas.

   If the modeling domain coincides with the domain constructed for this demonstration
   modeling exercise, then it would be advisable, if possible, to utilize the five-year (1988-92)
   meteorological data set prepared for the current analysis because development of those data
   was particularly resource intensive and difficult.  Once the sources, receptors, modeling
   domain and modeling period have been selected, the regional impact analysis should be
   conducted by following the IWAQM Phase 1 Recommendations, and through the use of this
   report as a demonstration of the modeling process.
   IMPACTS AS A FUNCTION OF DISTANCE

   In addition to the PSD analysis, a separate investigation of the impacts of hypothetical sources
   as a function of distance was conducted on the same modeling domain. In this investigation,
   uniform hypothetical sources were placed in rings of constant distance from SNP, ranging
   from 50 km to 200 km. Four months were simulated, each representing one season of the year
   1988.

   The results from this investigation showed that impacts from primary species such as SO2 and
   NOX decreased with source distance from SNP.  Short-term impacts decayed with source
   distance in a  manner consistent with the hypothesis that they are the result of a single source.
   Long-term impacts decayed with source distance as though they were the result of the entire
   ring of sources.  Specifically, maximum 3-hour SO2 impacts decreased by 86 percent between
   the 50 km and 200 km source rings.  For the maximum 24-hour SO2 impacts, the decrease was
94030rl.50
                                           5-7

-------
   68 percent.  Maximum monthly average SO2 levels decreased by 58 percent; when the
   concentrations are normalized by the number of sources on each ring, the decrease was 83
   percent.  For NOX, the maximum monthly average concentration decreased by 65 percent
   between the 50 km and 200 km source rings, and the normalized maximum monthly average
   concentration decreased by 86 percent.

   The trend in simulated PM10 (defined here as the sum of ammonium sulfate and ammonium
   nitrate) with source distance was very different from the trend in SO2 and NOX.  Maximum
   modeled PM10 impacts occurred at the intermediate distances (125 to 175 km).

   Modeled sulfur deposition was dominated by SO2 deposition at all distances and for all
   seasons, although sulfate deposition peaked in July.  Therefore, sulfur deposition showed a
   similar trend with distance as did the long-term SO2 concentrations.  Modeled nitrogen
   deposition was dominated by HNO3 deposition in summer and NOX deposition in the other
   seasons.
94030rl.50
                                           5-8

-------
                                    REFERENCES
   Browder, J.  1993.  Personal communication. Transferred discrete receptor coordinates for
         Shenandoah National Park and James River Face Wilderness.  October, 1993.

   Browder,!.  1994.  Personal communication. Transferred ISCMPF computer code.
         February, 1994.

   Cass, G.R. 1980. Methods for Sulfate Air Quality Management. EQL Report No. 16-2.
         Environmental Quality Laboratory, California Institute of Technology, Pasadena,
         California.

   Cimorelli, A. 1993. Personal communication. November, 1993.

   EPA.  1992a. A Modeling Protocol for Applying MESOPUFF II to Long Range Transport
         Problems. EPA-454/R-92-021. U.S. Environmental Protection Agency, Office of Air
         Quality Planning and Standards, Research Triangle Park, North Carolina.

   EPA.  1992b. User's Guide for the Industrial Source Complex (ISC2) Dispersion Models.
         EPA-450/4-92-008.  U.S. Environmental Protection Agency, Office of Air Quality
         Planning and Standards, Research Triangle Park, North Carolina.

   EPA.  1993. Interagency Workgroup on Air Quality Modeling (IWAQM) Phase 1 Report:
         Interim Recommendation for Modeling Long Range Transport and Impacts on
         Regional Visibility.  EPA-454/R-93-015.  U.S. Environmental Protection Agency,
         Office of Air Quality Planning and Standards, Research Triangle Park, North Carolina.

   EPA.  1994. A Revised User's Guide to MESOPUFF II (VS. 1). EPA-454/B-94-025. U.S.
         Environmental Protection Agency, Office of Air Quality Planning and Standards,
         Research Triangle Park, North Carolina.

   Morris, R.E. and C.C.  Chang.  1992.  User's Guide to the National Park Service's Air Quality
         Modeling System. SYSAPP-92/009. Prepared for National Park Service, Lakewood,
         Colorado, by Systems Applications International, San Rafael, California.

   Olsen, A.R.  1988.  1986 Wet Deposition Temporal and Spatial Patterns in North America.
         EPA DE-AC06-76RLO 1830, U.S. Environmental Protection Agency, Research
         Triangle Park, North Carolina.
94030rl.09
                                          R-l

-------
   Scire, J.S., F. Lurmann, A. Bass, and S. Hanna.  1984.  User's Guide to the MESOPUFF II
         Model and Related Processor Programs.  EPA-600/8-84-013. U.S. Environmental
         Protection Agency, Atmospheric Research and Exposure Assessment Laboratory,
         Research Triangle Park, North Carolina.

   Scire, J.S. 1993. Personal communication. December, 1993.

   Sisler, J.F., D. Huffman, and D.A. Latimer. 1993. Spatial and Temporal Patterns and the
         Chemical Composition of the Haze in the United States:  An Analysis of Data from
         the IMPROVE Network. 1988-1991. ISSN No. 0737-5352-26. Report prepared for
         National Park Service, Fort Collins, Colorado and EPA/EMSL, Las Vegas, Nevada.
94030rl.09
                                          R-2

-------
                                  APPENDIX A

    INTERAGENCY WORKGROUP ON AIR QUALITY MODELING (IWAQM)
The following IWAQM members participated on the assessment of the Phase 1
Recommendations for MESOPUFF n application:
      Group Chairman
      Mark Scruggs (IWAQM chairman)
      National Park Service
      Air Quality Division
      P.O. Box 25287
      Denver, CO  80225-0287
      National Park Service
      John Vimont
      National Park Service
      Air Quality Division
      P.O. Box 25287
      Denver, CO  80225-0287
      U.S. Environmental Protection Agency
      Jason Ching
      Atmospheric Research and Exposure Assessment Laboratory (MD-80)
      U.S. Environmental Protection Agency
      Research Triangle Park, NC 27711
      Alan Cimorelli
      U.S. Environmental Protection Agency
      Region HI (3AM12)
      841 Chester Street
      Philadelphia, PA 19107
      John S. Irwin (work assignment manager)
      SRAB, TSD (MD-14)
      Office of Air Quality Planning and Standards
      U.S. Environmental Protection Agency
      Research Triangle Park, NC 27711
 94030rl.07
                                      A-l

-------
      U.S. Forest Service
      Richard W. Fisher
      U.S. Forest Service
      Rocky Mountain Station
      240 West Prospect Road
      Ft. Collins CO  80526
      U.S. Fish And Wildlife Service
      Bud Rolofson
      National Park Service
      P.O. Box 25287
      Denver, Colorado 80225-0287
      State Participants
      James Browder
      Virginia Department of Environmental Quality
      4900 Cox Road
      Glen Allen, VA  23060
      Pat Hanrahan
      State of Oregon
      Department of Environmental Quality
      811 SW 6th Avenue
      Portland, OR 97204
      Ken McBee
      Virginia Department of Environmental Quality
      4900 Cox Road
      Glen Allen, VA  23060
94030rl.07
                                       A-2

-------
                                      APPENDIX B

                             BASELINE DETERMINATION
   The following notes are a summary of a telephone conference held on November 2, 1993,
   between the project team, IWAQM members, and Dan deRoeck (responsible for new source
   review at EPA OAQPS).  The topics of the conversation were (1) how to select sources that
   should be included as PSD sources for the demonstration modeling analysis, (2) the
   relationship between PSD increments and AQRVs, and (3) a summary  of emission inventory
   needs for each proper analysis. An important issue for PSD source determination was the
   triggering dates for various classes of emission source.

   As a result of this conversation, it was discovered that more than one emission inventory
   would be needed and that the MESOPUFF n model would have to be exercised for each
   inventory separately to determine the impacts from each grouping of sources.  Because these
   additional model runs were not possible within our resources, we decided to restrict the scope
   of the demonstration modeling exercise to determine the PSD and AQRV impacts at Class I
   areas (Shenandoah NP and James River Face W) from to an arbitrary group of PSD sources.

   The following notes were received from Al Cimorelli, EPA Region in, on November 3,  1993:
   BASELINE DETERMINATION

         •  Baseline determination is no different for Class I areas than it is for Class II.

            •  Baseline is determined individually for Section 107 areas (i.e., attainment and
                unclassified areas). This means that for Class I areas as large as Shenandoah
                baseline dates may be different for different parts of the park.

            •  There are 2 dates that are important:  MAJOR source baseline date and
                MINOR source baseline date.

                •  Major Source Baseline Date:  Established by the Statute (1/6/75 for SO2
                   and Particulate) - All major sources whose emissions have changed, as a
                   result of construction activities, after 1/6/75, effect increment. However,
                   only the emission changes effect increment (major source is defined as 100
                   tons/yr for the 28 categories listed in the PSD regs and 250 tons/yr for all
                   other sources.
94030rl.04
                                           B-l

-------
                •  Minor Source Baseline Date: Established by first complete PSD
                   application either locating in 107 area or whose emissions significantly
                   impact the 107 area. After the Minor Source Baseline date all actual
                   emission changes from any source for any reason effect increment.

             •  Therefore, proper construction of the increment effecting inventory must
                consider both dates.

             •  For the Demo project we did not ask for an inventory of this type.  We asked
                simply for the increment effecting inventory.  We decided not to go back to the
                states and clarify, we will see what we get back and discuss at the 11/30
                meeting. At that meeting we may request the states to develop the appropriate
                inventory by say the end of the year.
   AQRV's

          •  We discussed AQRVs and their relationship to Increment:

             •  Dan explained that he believed that AQRVs should be viewed in a way similar
                to NAAQS not Increment.  That is, they should be viewed from an absolute air
                quality point of view not relative to some baseline.  However, he indicated that
                there may be others in the program with a different view. For example,
                AQRVs could be evaluated in the same way as increment.

             •  However adverse impacts on AQRVs, unlike NAAQS, should represent what
                is actually occurring not what could occur if all sources were operating at their
                allowable emissions.

             •  Therefore, to properly evaluate AQRVs we either must rely on measured data
                or model using an inventory of actual emissions from ALL presently existing
                sources which have a meaningful impact in the Class I area. It is against this
                measurement or prediction which we compare the modeled incremental
                impacts of an applicant source.

          •  In designing the Demo Project we did not follow this interpretation of AQRVs.
             We plan to calculate the AQRVs using the increment effecting inventory we
             requested from the states.  If we continue along this line we must be careful in
             explaining what we have done, what it means & how it relates to properly
             constructed analysis.


   EMISSION INVENTORIES

   In summary, a proper Class I area analysis would require the development of three separate
   inventories.

             i.     For PSD Increment:
94030rl.04
                                           B-2

-------
                (1)    The change in emissions that have occurred as a result of construction
                       at any major sources, if that construction occurred after the Major
                       Source Baseline date.

                (2)    All actual emission changes, from any source, which occurred after the
                       Minor Source Baseline Date.

             ii.     ForAQRVs:

                    Actual present day emissions from all sources

             iii.     ForNAAQS:

                    Allowable emissions from all sources.
94030rl.04
                                            B-3

-------
                                     APPENDIX C

                       DESCRIPTION AND USER'S GUIDE FOR
                  NEW MESOPUFF II PRE- AND POSTPROCESSORS
   The MESOPUFF II program system is divided into three basic processors: the MESOPAC II
   meteorological preprocessor, the MESOPUFF n Lagrangian puff model, and the MESOFILE
   receptor concentration postprocessor. The MESOPUFF n system also contains several
   programs that prepare raw meteorological data for the MESOPAC preprocessor, including
   READ62, PMERGE, and PXTRACT.  These programs, as well as MESOPAC, require raw
   data in specific formats, which are available through the National Climatic Data Center.

   Before the demonstration study, internal array dimensions within all programs were set to
   values that were  assumed to allow for a large  degree of flexibility over the range of potential
   model applications. During the demonstration study, however, we found that several array
   dimensions were inadequately small, and that others were too large (leading to unnecessarily
   large memory requirements).  The table below presents a comparison between the original
   array dimensions and those currently set in the programs. The new values were selected to
   tailor the codes for the quantity of raw data, emission sources, and receptors required in the
   demonstration study, and are adequate for most future applications on this domain.  It is by no
   means meant to be optimum for all future applications in other areas.  We recommend,
   however, that dimensions for the number of point sources and total puffs should not be
   increased significantly above the current  settings, as MESOPUFF n speed is non-linearly
   dependent on the number of sources and puffs modeled.

         Table C-l.  Comparison of array dimension parameters between "original" and
         "demonstration" versions of the MESOPUFF II modeling system.
Parameter
MXNX
MXNY
MXSS
MXUS
MXPS
MXPUFF
MXREC
MXPTS
MXARS
MXOZ
Original Value
100
100
100
20
100
10,000
1,000
1,000
200
50
Demonstration
Study Value
50
35
50
10
200
20,000
100
20
1
200
Description
Maximum cells in x direction
Maximum cells in y direction
Maximum surface stations
Maximum upper air stations
Maximum precipitation stations
Maximum active puffs on the grid
Maximum non-gridded receptors
Maximum point sources
Maximum area sources
Maximum ozone stations
94030rl.lO
                                          C-l

-------
   Besides meteorological data requirements, MESOPAC and MESOPUFF n require gridded
   land use information, and optionally, hourly observed ozone concentrations at an array of
   monitoring sites within the modeling domain.  The original MESOPUFF II system does not
   contain programs to prepare these input data.  Further, we were not able to procure raw
   meteorological data in the exact formats required by MESOPAC, and were not able to use
   MESOFILE to calculate concentration, deposition, and visibility increments on seasonal to
   annual time scales.  Therefore, additional pre- and postprocessors were developed to prepare
   the input and output data for the demonstration modeling study.  These processors have been
   packaged for the SCRAM BBS, where they will be available for public use. The processors
   are grouped into the following categories:

         Surface data processing
   •      Upper air data processing
         Meteorological  station programs
   •      Gridded land use data processing
         Ozone data processing
   •      Postprocessing programs

   A description of each program and a guide for application are presented on the following
   pages.
94030rl.lO
                                           C-2

-------
   SURFACE DATA PROCESSING
   GETSTN/RESFC Tape Extraction Programs

   GETSTN and RESFC are used together to extract hourly surface data for a given time and
   space window from a single raw TD-3280 file.  The product of GETSTN and RESFC
   programs is an intermediate monthly surface data file for stations located within the
   meteorological modeling domain. These programs were run for all 60 months of 1988-1992,
   extracting data for a spatial window within the latitude/longitude ranges 34-42 °N and 71-
   86°W, which resulted in hourly surface observations for up to 48 stations.

   RESFC can be run on each single raw TD-3280 file and its output combined into a single
   monthly intermediate file (a rather cumbersome process if many raw files exist for a particular
   year), or, as is recommended in Section 3, the raw data can be first concatenated and RESFC
   run on the combined raw file.  Although RESFC requires substantially more mainframe CPU
   time to process the larger annual concatenated files, significant labor can be saved by
   removing RESFC setup time for all subfiles and removing the post-RESFC combining
   procedures.

   Because of file size and time required to operate the tape extraction programs, GETSTN and
   RESFC were run on our Trace Multiflow 14/300 Unix mainframe system. On the Unix
   system, processing of each month took between 20-30 minutes of mainframe CPU time (6
   hours per year); about an hour of labor per year was spent setting up and executing RESFC,
   and inspecting output diagnostics and data files.

   Raw TD-3280 data does not contain information about each station's geodetic coordinates, so
   location information must be obtained from a separate source. The ASCII WBAN master list
   file, obtained from DRI along with the raw TD-3280 surface data, catalogs (in numerical
   order) all WBAN weather station identification numbers within the U.S. and its territories
   along with station names, latitude/longitude coordinates, period of operation, and types of
   services each station offered. Often, one WBAN station number is repeated for several sites if
   a particular station has moved; dates of operation at each site location are given in the file
   with the most current coordinates for active stations reported first (current locations are valid
   as of 1987). Usually, station sites are moved by no more than 1 kilometer, but there are some
   instances when an old WBAN number for a non-operational site has been recycled for a new
   station much farther away.  Under these circumstances, the most current location for a
   particular WBAN number may not be reported first.

   GETSTN reads the WBAN master list file and extracts all station information for those sites
   within a specified latitude/longitude range. The output file generated by GETSTN is used by
   RESFC to assign latitude/longitude coordinates to the surface station data.  Once GETSTN
   has been run, the  output file should be checked to ensure that the most current position for
   each active WBAN number is reported first.  GETSTN reads input parameters from the
   standard input unit.  This can be accomplished using a free-format run stream, an example of
   which is presented below:

   %/bin/time -i getstn Ğ ieof


94030rl.lO                                     C-3

-------
   wbanmstl .1st    (master WBAN station file name
   all.station.lst     (GETSTN output file name
   34.,42.,71.,86.    (latitude/longitude minima & maxima
   ieof

   After running GETSTN, RESFC can be run on a concatenated TD-3280 file or individual
   sub-file.  As the program processes the raw surface data and comes upon a new station, it will
   search through the output of GETSTN for the first record containing the current WBAN
   number and assign those coordinates. The following is an example Unix run command for
   RESFC:

   %/bin/time -i resfc

   RESFC requires an input parameter file labelled "resfc.in", which contains information about
   the input file names, period over which to extract surface meteorological data, coordinate
   type, and time zone.  A sample free-format RESFC parameter file for December 1990 is
   displayed below:

   all.station.lst     (GETSTN output file name
   1990sfc.all       (Concatenated raw TD-3280 file name
   sfcdec.90        (monthly RESFC output data file name
   1201 1231      (Start month/day, end month/day
   0 5              (Coordinates (lat/lon), time zone (EST)

   where coordinate and time zone codes are given by:

         Code  Coordinates
         0  Lat/lon
         >0 UTM
         <0 Local grid units

         Code  Time zone (standard time)
         5  Eastern
         6  Central
         7  Mountain
         8  Pacific
94030rl.lO
                                          C-4

-------
   TOCD144 Data Reformatting/Filling Program

   TOCD144 reads the monthly intermediate files, fills missing values, and reformats the data
   into monthly files in CD144 format. The program was written to fill 1-6 hour intervals of
   missing data using linear interpolation over time. For longer periods of missing values, data
   from surrounding surface stations are used to spatially interpolate information to the station
   using an inverse-distance-squared weighting technique. A minimum of 2 stations and a
   maximum of 4 stations are used within 200 km of a station needing spatial interpolation.  If all
   data for a particular station is missing for a significant portion of a month, the station should
   be deleted from the database for that month.

   TOCD144 was compiled using Microsoft (MS) FORTRAN 5.0; processing of each month
   took about 5 minutes of PC CPU time using an 486/66 MHz processor (this PC was used
   throughout all raw meteorological data preprocessing).  Total labor and computer time
   required to run TOCD144 averaged about 1.5 hours per year.

   Since MESOPAC uses input temperature to determine liquid/frozen precipitation states if
   present weather fields are missing from the CD144 data, all present weather fields are set to
   missing values in the processed CD144 files. Sea level pressure in millibars is calculated
   from station pressure in inches of mercury using the hypsometric equation (via station
   temperature and elevation) within TOCD144. TOCD144 does not interpolate for missing
   cloud ceiling height fields, either in time or space, due to the often  large temporal and spatial
   variations observed in cloud heights, and the potentially large and uncertain impacts on
   MESOPUFF II applications. For the most part, missing cloud heights occur at night, as these
   observations are typically estimated visually; they only periodically occur during daylight
   hours.  In TOCD144, sky cover percentage is used to set cloud height: if sky cover is less than
   50%, cloud height is set to an "unlimited"  ceiling; if cloud cover is greater than 50%, cloud
   height is set to 5000 ft, a height which alters the classification of solar insolation within
   MESOPUFF II to "low cloud" values (refer to page 5-13 of the MESOPUFF H User's Guide;
   EPA 1994).

   TOCD144 is run at the DOS prompt by typing the program name and the name of the input
   parameter file as an argument:

   C:\>TOCD144 APR92.PRM

   The parameter file contains the names of the monthly surface data file generated by RESFC,
   the new "filled" and reformatted CD 144 output file, and a diagnostic message file.  A sample
   for April 1992 is given below:

   apr92.sfc     [monthly surface data file name from RESFC
   sf92_apr.fll   (monthly filled CD144 file  name
   apr92.msg     message file name
94030rl.lO
                                           C-5

-------
   UPPER AIR DATA PROCESSING

   REUPR2 Tape Extraction Program

   REUPR2 extracts 12-hourly upper air data for a given time and space window from a single
   raw nationwide/annual TD-6201 file. To reduce file sizes and expedite filling of missing data,
   REUPR2 writes out sounding data between the surface and about 7000 m, and deletes non-
   mandatory levels if they have missing pressure. The resulting TD-6200 file contains
   intermediate monthly upper air data for stations located within the meteorological modeling
   domain. Because of file size and time required to operate the tape extraction program,
   REUPR2 was also run on our Unix mainframe system. The program was run for all 60
   months of 1988-1992, extracting data for a spatial window within the latitude/longitude
   ranges 34-42°N and 71-86°W, which resulted in upper air data from as many as 8 rawinsonde
   stations. REUPR2 required about 15-20 minutes of mainframe CPU per month, yet very little
   labor was involved.

   A sample Unix run command for REUPR2 is provided below:

   %/bin/time -i reupr2

   REUPR2 requires an input parameter file called "reupr.in" that contains the longitude/latitude
   range, date range, name of the raw nationwide TD-6201 data file, and name of the REUPR2
   output file. A sample for December 1992 is presented below:

      71.0   86.0   34.0    42.0
   1130 1231       [Beginning month/day, ending month/day
   1992upr.all          (Input raw TD-6201 file name
   up92_dec.dat         (Output upper air data file

   The first record format is (4flO. 1); the second record is free-format. It is generally good
   practice to specify an extra day before and after the month of interest when extracting  12-
   hourly upper air data, particularly if time-interpolation is required in later processing (which
   was not done in the current study).
94030rl.lO
                                           C-6

-------
   FILLUPR Data Filling Program

   FILLUPR reads the intermediate TD-6200 file and scans the sounding data for missing
   values. The program linearly interpolates between pressure levels for data gaps less than 200
   mb deep. If a significant portion of the sounding is missing (i.e., > 200 mb), or if the
   sounding is missing altogether, the program spatially interpolates data from nearby sites to the
   station location using a distance weighted average (at mandatory levels only). FILLUPR
   follows the processing scheme outlined below:

   (1)    In a first pass through the data, the program finds the first level where there is a
         missing value (either height, temperature, or wind). The program proceeds up the
         sounding until a non-missing value is found. If this gap is greater than 200 mb deep,
         the mandatory levels within the missing block are flagged for spatial interpolation of
         this particular variable. If the gap is less than 200 mb deep, linear interpolation (using
         height or log-pressure) is performed for the variable at all levels (mandatory or
         otherwise) within the gap. For winds, interpolation is done for vector components.

   (2)    The program then proceeds farther up the sounding and repeats the procedure for data
         gaps aloft. If no valid data is found up to the top of the 7000 m sounding, and the data
         gap is greater than 200 mb, all mandatory levels above the level of good data are
         flagged for spatial interpolation.  If the data gap to the top of the sounding is less than
         200 mb, the data are extrapolated using the highest two levels with valid data (this
         saves on spatial interpolation time).

   (3)    Data for all soundings needing spatial interpolation are written to a temporary direct
         access file to ease memory requirements and quicken run time. In the second pass
         through the data, a matrix containing the top 5 closest stations to each of the stations in
         the file is computed. Then the flags for missing data at mandatory levels are checked
         for spatial interpolation.

   (4)    When a mandatory level requiring spatial interpolation for a particular variable is
         identified, data from the closest 2 to 4 upper air stations within a radius of 500 km are
         used in an inverse-distance weighted average. Stations containing spatially
         interpolated values at the same mandatory level are not used in this calculation.

   (5)    After a temperature sounding has been completely filled via spatial interpolation, the
         resulting temperature gradients are checked against a slightly super-adiabatic lapse  rate
         to insure thermodynamically realistic values. The limiting lapse rate specified was  -
         0.15 K/m; this limit was violated only infrequently. When the violation occurs, the
         interpolated temperatures are adjusted to adhere to an adiabatic lapse rate (-0.01 K/m)

   FILLUPR was compiled using MS FORTRAN 5.0, and took about 5 minutes PC CPU time
   per month.  Labor and CPU time together required about two hours per year to process upper
   air data through FILLUPR.

   Like TOCD144, FILLUPR is run at the DOS prompt; it is supplied with the name of the
   input parameter file as an argument:
94030rl.lO
                                            C-7

-------
   C:\>FILLUPR APR.PRM

   The parameter file contains the name of the extracted TD-6200 data from REUPR2, the name
   of a temporary direct access file for intermediate processing, the name of the "filled" output
   TD-6200 data file, the name of a message file, and the date/time of the sounding preceding the
   first sounding of the month. A sample parameter file for April 1992 is presented below:

   up92_apr.dat         (Input monthly upper air data file name
   dummy.out          (Temporary direct access file name
   up92_apr.fll          ("Filled" output file name
   apr92.msg           (Message file name
   3  31 12              (Month/day/hour of preceding sounding

   In the last record,  the free format string "3 31 12" represents the March 31 1200Z sounding,
   whereas the first sounding in the "up92_apr.dat" file is April 1 OOOOZ.
94030rl.lO

-------
   METEOROLOGICAL STATION PROGRAMS

   PARSE Station Splitting Program

   PARSE was written to split monthly surface CD 144 files and TD-6200 upper air files into
   numerous monthly station files, which are ready to be used directly by MESOPAC.
   Processing of single monthly files and splitting into separate station files just before running
   MESOPAC for each month simplified intermediate file management and provided faster
   meteorological data processing.  Typically, data for 45-48 surface stations needed to be split
   into separate station files each month, taking just about 1 minute of PC CPU using MS
   FORTRAN 5.0, and virtually zero labor. Up to eight separate upper air station files were
   generated by PARSE, taking only a few seconds of PC CPU time.

   PARSE is run separately for surface and upper air data files.  The program must be told which
   type of file it is supplied by specifying a flag as an argument.  For example, the following
   command at the DOS prompt will split surface data for April 1992 into 45 separate station
   files:

   C:\>PARSE SF92_APR.FLL S

   where the "S" denotes surface data, and the file name is the "filled" CD 144 format  data file
   generated by TOCD144.  Similarly for upper air data in the same month, PARSE will
   produce 7 upper air station data files:

   C:\>PARSE R6APR_92.DAT U

   where the "U" denotes upper air data, and the file  name is the file generated by READ62,
   which operates on output from FILLUPR.
94030rl.lO
                                          C-9

-------
   GRIDDIT Station Location Processing Program

   GRIDDIT calculates surface, upper air, and precipitation station meteorological grid
   coordinates from latitude/longitude data. The program generates one file containing grid
   coordinates and other information for all stations on the MESOPAC grid, which is then ready
   to be concatenated into a MESOPAC input control file (PAC.INP). Since the number of
   stations reporting each month vary, GRIDDIT must be run for each month to ensure that the
   proper station locations are supplied to MESOPAC.

   For surface and upper air stations, the latitude/longitude position coordinates are stripped from
   their respective monthly intermediate files (i.e., those files output by RESFC and REUPR2,
   respectively).  The MESOPAC input file PAC.INP also requires that surface station grid
   locations include representative surface roughness values. Therefore, GRIDDIT must be
   supplied with the MESOPUFF II land use file so that once surface station locations are
   determined within the MESOPAC domain, appropriate surface roughnesses can be assigned
   to each station. Since no coordinate information is available with the TD-3240 precipitation
   data, the program cross-references identification numbers within each month's MESOPAC-
   ready precipitation data file with station numbers in a "precipitation station history file" that
   SAI obtained from DRI.

   For the upper air data, we extracted the first 18 columns from each record of the intermediate
   monthly TD-6200 files output by REUPR2 (this file is formatted such that one record
   contains an entire sounding) and wrote them to monthly radiosonde station location files. An
   example of these files is given below:

     38603822  8233
    137233605  7957
         Continued for all radiosonde stations

   The order of each record is station WBAN number (column 1-8), latitude degrees (column 9-
   10), latitude minutes (columns 11-12), longitude degrees (column 13-16), and longitude
   minutes (column 17-18).  The format is (18,212,14,12).

   For the surface data, the station header records were stripped from the monthly RESFC output
   files and written to monthly surface station location files.  An example of these files is given
   below :
94030rl.lO
                                           C-10

-------
    3812 ASHEVILLE/    82.483  35.433  638.0  0768  5
    3860HUNTINGTON  82.550  38.367  255.0  0768  5
         Continued for all surface stations

   The order of each record is station WBAN number (columns 1-6), longitude in decimal
   degrees (columns 20-30), latitude in decimal degrees (columns 30-40), and station elevation
   in meters (column 40-50).  The format is (i6,14x,f7.3,3x,f7.3,2x,f6.1).

   GRIDDIT requires an input parameter file,  which is read from the standard input file unit,
   and writes diagnostic messages to the standard output file unit. It is easiest to develop an
   input parameter file and execute GRIDDIT using the following DOS command:

   C:\>GRIDDIT < GRIDDIT.INP > GRIDDIT.OUT

   In the example above, GRIDDIT writes diagnostic output to the file GRIDDIT.OUT instead
   of the screen. The input parameter file (GRIDDIT.INP) contains the MESOPAC station
   location file name to be generated by GRIDDIT, the input surface  station location file name,
   the input MESOPUFF II land use file name, the input radiosonde station location file name,
   the MESOPAC-ready precipitation file name, the precipitation station history file name, and
   some grid parameters. An example input file for December 1988 is given below:

   decSS.stn              |MESOPAC station location file name
   decSS.sfc              (Surface station location file name
   landuse.mesopuff       |MESOPUFF n landuse file name
   decSS.upr              |Upper station location file name
   pmerge.dec            |MESOPAC precipitation file name
   shfmst92.cat           (Station history file name
   100. 3900. 17 20. 50 35  (Grid parameters

   In this example, the grid parameters are read in free format,  and include the UTM easting
   coordinate for the meteorological grid  origin (southwest corner, km), the UTM northing for
   the grid origin (km), the UTM zone, the grid cell size (km), and the number of grid cells  in the
   east-west and north-south directions, respectively.
94030rl.lO
                                          C-ll

-------
   GRIDDED LAND USE PROCESSING

   PRELND Data Extraction Program

   Land use data are supplied in the form of Geographical Information System (GIS) database
   files. The GIS file that PRELND requires contains the distribution of 11 land use categories
   over the entire U.S. at 1/4 degree longitude by 1/6 degree latitude resolution.  This file (called
   "gis.dat") was made available to the National Park Service with delivery of the National Park
   Service Air Quality Modeling System. PRELND maps land use data from this file to the
   meteorological grid in terms of the percentage of each of 11 UAM categories in each cell.

   PRELND was run on our Unix mainframes. A sample Unix run command is shown below:

   %/bin/time -i prelnd2

   PRELND requires an input parameter file called "plnd.in" that contains information about I/O
   file names and grid information:

   gis.dat          |Raw GIS database file name
   prelnd2.out     (Output ASCII message file name
   landuse.gis      (Output binary landuse file name
      100.   3900.    17    20     50     35

   The last record of "plnd.in" contains the grid UTM origin (km), UTM zone, grid cell  size
   (km), and the number of cells in the east-west and north-south direction, respectively. The
   format is (2fl0.0,4110).


   UAM2MESO Landuse Mapping Program

   The UAM land use categories output by PRELND do not match those required by
   MESOPUFF II. The program UAM2MESO identifies the dominant GIS land use category in
   each cell and maps it into an appropriate MESOPUFF II category.  The assumed mapping
   arrangement between UAM and MESOPUFF n land use categories is displayed in Table 3-3
   (Section 3). All steps in developing a MESOPUFF II land use field were again executed on
   our Unix mainframe because of the necessity to read binary output from PRELND.

   UAM2MESO is designed to read input information from the standard input file unit; an
   example of the Unix run stream for UAM2MESO is given below:

   %/bin/time -i uam2meso Ğ EOF
   landuse.gis             (Binary landuse file name from PRELND
   landuse.mesopuff                              (ASCII MESOPUFF H-ready landuse file
                                                name
   EOF
94030rl.lO
                                         C-12

-------
   OZONE DATA PROCESSING

   EXTRACT Data Extraction Program

   EXTRACT withdraws hourly ozone data from files written in EPA Aerometric Information
   Retrieval System (AIRS) AMP-350 data work file format.  The program produces monthly
   files for a spatial window covering the entire MESOPAC meteorological domain.

   EXTRACT was run on our Unix mainframe. A sample Unix run stream for June 1992 is
   provided below:

   %/bin/time -i extract Ğ ieof
   rawairs.meso92.o3.dat  |Raw AIRS AMP-350 work file name
   1                     (Old/new data flag: should be " 1"
   o3jun92.raw           (Output work file name
   o3jun92.out           (Message file name
   6 92                  (Month/Year of data to extract
   17                    lUTMzone
    100. 1100.            |Min/Max UTM easting coordinates
   3900. 4600.            |Min/Max UTM northing coordinates
   ieof

   AIR2MESO Reformatting Program

   AIR2MESO reformats the hourly ozone data into MESOPUFF H OZONE.DAT format, and
   produces a separate file containing grid coordinates for each AIRS monitor location. The
   coordinates file is then easily inserted into the MESOPUFF II input file (PUFF.INP) as
   necessary when supplying an OZONE.DAT file to the model. In order to simplify processing,
   AIR2MESO does not average multiple stations within the same grid cell. Also, the program
   does not add pseudo stations with background ozone in areas lacking data coverage. The total
   number of ozone monitoring sites within the meteorological domain depends highly on
   season, ranging from about 50 or 60 in the winter, to around  150 in the summer.

   MESOPUFF H identifies missing data within the OZONE.DAT file and fills it with either the
   default (80 ppb) or user-specified background ozone concentration.  The AIR2MESO
   program was written to calculate daytime/domain average ozone concentrations for each
   month to assist the user in supplying MESOPUFF U with a specific background ozone
   concentration for each application.

   A sample Unix run stream for the June 1992 AIR2MESO  run is shown below:
94030rl.lO
                                         C-13

-------
   %/bin/time -i air2meso Ğ EOF
   92 6               (Year/Month
   100. 3900. 20.       (Grid UTM origin, cell size (km)
   o3jun92.raw         (Input work file name from EXTRACT
   o3jun92.dat         (Output OZONE.DAT file name
   o3jun92.inp         (Output monitor location file name
   EOF
94030rl.lO
                                        C-14

-------
   POSTPROCESSING PROGRAMS

   RHPOST Humidity Extraction Program

   AQRV computations include the calculation of aerosol extinction, which in turn is dependent
   on relative humidity. Relative humidity data is contained within the large 50 MB MESOPAC
   output files, only a few of which can fit onto the PC hard drive. The objective of PSDPOST,
   however, was to operate on all 12 months of each year continuously. Therefore, RHPOST
   was developed to extract hourly relative humidity data at each surface station from
   MESOPAC files, and output 3-hour average relative humidity at each receptor location to
   much smaller files for subsequent input to PSDPOST.

   Since MESOPAC was compiled using Lahey FORTRAN 5.2, RHPOST had to be compiled
   with the same in order to read the binary MESOPAC output files. RHPOST required just a
   few minutes of CPU time per month (on the 486/50 MHz PC) to extract relative humidity data
   from MESOPAC output files. Again, humidity files could only be extracted from 3 or 4
   MESOPAC output files before more MESOPAC files could be transferred from offline
   storage.

   RHPOST reads input parameters from the standard input file unit and outputs diagnostic
   messages to the standard output file unit.  The easiest way to supply the program with the
   required input parameters is to edit an input file (e.g., called RHPOST.IMP) and type the
   following command at the DOS prompt:

   C:\>RHPOST < RHPOST.INP >RHPOST.OUT

   In the above example, diagnostic messages will be sent to the file RHPOST.OUT instead of
   the screen. The input parameter file contains the name of the monthly MESOPAC output file
   from which to extract humidity data, the name of the monthly output humidity file, the
   number of receptors for which to assign humidity values, and a list of all non-gridded receptor
   locations (in terms of grid units).  An example RHPOST.INP file is shown below:

   ju!90.out            MESOPAC binary output file name
   rhjul90.dat         (Output binary humidity file name
   64                 (Number of non-gridded receptors
     27.715   14.049   |Grid coordinates of receptors
     27.744   13.878   |(Free format)
     27.877   13.992
     27.901   13.831
     27.910   14.189
     27.995   13.969
         Continued for 64 receptor locations
94030rl.lO
                                         C-15

-------
   PSDPOST Postprocessing Program

   PSDPOST allows for quick and efficient calculations of the specific PSD and AQRV impacts
   (concentration averages and deposition sums), as described in Table 3-4 (Section 3).
   PSDPOST requires 12 months of binary MESOPUFF II output files and 12 monthly
   RHPOST binary output files for each year. PSDPOST then determines humidity-dependent
   extinction coefficients resulting solely from modeled PM10 (ammonium sulfate and
   ammonium nitrate). This calculation specifically follows the procedure outlined in Appendix
   B of the IWAQM Phase 1 recommendations. Visibility impacts are reported as the amount of
   time (both in terms of the absolute number of 3-hour periods and percent of each year) the
   calculated maximum extinction over all receptors is more than 10% higher than the cleanest
   observed extinction.  The cleanest observed extinction was taken to be the 90th percentile
   extinction over all years in which statistical analyses of IMPROVE monitoring data are
   available at SNP (1987 through 1991). From Sisler et al. (1993), the 90th percentile
   particulate extinction for this period at the Shenandoah monitor was 0.005 km"1.  This was
   combined with a Rayleigh (pure air) scattering extinction of 0.010  km"1 to obtain a total
   observed cleanest extinction of 0.015 km"1.

   Since MESOPUFF II and RHPOST were compiled using Lahey FORTRAN 5.2, PSDPOST
   had to be compiled with the same in order to read the binary model output files.  PSDPOST
   required about 10 minutes of PC CPU per month (on the 486/50 MHz PC).

   Like RHPOST, PSDPOST reads an input parameter file from the  standard input file unit, and
   outputs diagnostic messages to the standard output file unit. PSDPOST will generate a file
   called "PSDPOST.DAT", which will contain PSD increments and  statistics for each non-
   gridded receptor and an overall summary.  Again, it is easiest to execute PSDPOST using the
   following DOS command:

   C:\>PSDPOST < PSDPOST.INP > PSDPOST.LST

   In this example, diagnostic messages will be routed to PSDPOST.LST rather than to the
   screen. PSDPOST is meant to calculate certain measures for an entire  year's worth of
   MESOPUFF II output.  The input parameter file (PSDPOST.INP in this example) is therefore
   rather lengthy. For each run, the input file specifies the number of non-gridded receptor
   groups over which to calculate PSD increments, and the receptor index ranges for each group.
   Then for each month, the input file specifies the names of the binary MESOPUFF II output
   concentration, dry deposition, and wet deposition files, along with  the name of the monthly
   relative humidity files generated by RHPOST. A sample input file for the 1989 MESOPUFF
   II run is presented below:

   2                  |Numb er of receptor group s
   1 9 10 64            (Start/end indices (1-9,10-64)
   jan89.dat            (Concentration file name
   jan89.dry            |Dry dep file name
   jan89.wet           |Wet dep file name
   rhjan89.dat         (Humidity file name
   1                  (Humidity flag
   feb89.dat
94030rl.lO
                                         C-16

-------
   feb89.dry
   feb89.wet
   rh_feb89.dat
   1
   mar89.dat
   mar89.dry
   mar89.wet
   rh_mar89.dat
   1
   apr89.dat
   apr89.dry
   apr89.wet
   rh_apr89.dat
   1
   may89_a.dat
   may89_a.dry
   may89_a.wet
   rh_may89.dat
   0
   may89_b.dat
   may89_b.dry
   may89_b.wet
   1
   jun89.dat
   jun89.dry
   jun89.wet
   rhjun89.dat
   1
   jul89.dat
   ju!89.dry
   ju!89.wet
   rhjul89.dat
   1
   aug89.dat
   aug89.dry
   aug89.wet
   rh_aug89.dat
   1
   sep89.dat
   sep89.dry
   sep89.wet
   rh_sep89.dat
   1
   oct89.dat
   oct89.dry
   oct89.wet
   rh_oct89.dat
   1
   nov89.dat

94030rl.lO
C-17

-------
   nov89.dry
   nov89.wet
   rh_nov89.dat
   1
   dec89.dat
   dec89.dry
   dec89.wet
   rh_dec89.dat

   Note that beginning with the second group of input file names (February), a flag must be
   given to signal whether a relative humidity file is to be opened for that group. A value of" 1"
   indicates that a humidity file is to be opened, while a value of "0" will keep the current
   humidity file open for subsequent processing. This allows for cases in which a particular
   month was split into 2 or more MESOPUFF II runs (i.e., due to program crash, system
   problem, or planned separation). This was the case for May 1989, where in the above
   example, the MESOPUFF n application was split into "a" and "b" runs; for the "b" group, no
   humidity file is specified.  PSDPOST attempts to resynchronize the humidity file with the
   current date/time that the program reads from the MESOPUFF n output files.
94030rl.lO
                                          C-18

-------
                                 APPENDIX D

                        ISCST2 MODEL RESULTS AND
                INTEGRATION WITH MESOPUFF II RESULTS
The tables on the following pages are portions of output files from the ISCST2 model runs
and the ISCMPF integration program:

      (1) page 9 of ISCST2 output file for SO2

      (2) page 9 of ISCST2 output file for NOX

      (3) page 1 of ISCMPF output file (MESOPUFF II header)

      (4) ISCMPF output file table corresponding to maximum 3-hour average SO2
         concentrations

      (5) ISCMPF output file table corresponding to second high 3-hour average SO2
         concentrations

      (6) ISCMPF output file table corresponding to top 10 3-hour average SO2
         concentrations
                                     D-l

-------
 *** ISCST2  -  VERSION 93109  ***     *** WAMPLER-LONGACRE:  SOx Emissions;   Sept. 27, 94                         ***         09/27/94
                                     ***                                                                          ***         12:04:26
                                                                                                                             PAGE    9
 *** MODELING  OPTIONS USED:   CONG   RURAL                 DFAULT

                                                  ***  THE SUMMARY OF HIGHEST  3-HR RESULTS  ***


                                         ** CONG  OF SO2       IN  (MICROGRAMS/CUBIC-METER)                  **

                                                       DATE                                                                  NETWORK
GROUP  ID                          AVERAGE CONG      (YYMMDDHH)               RECEPTOR   (XR, YR,  ZELEV,  ZFLAG)      OF  TYPE  GRID-ID


ALL     HIGH  1ST HIGH VALUE  IS       11.41345   ON 90070503:  AT  (  703580.00,   4270240.00,       0.00,       0.00)   DC
        HIGH  2ND HIGH VALUE  IS        8.16474   ON 90072624:  AT  (  703580.00,   4270240.00,       0.00,       0.00)   DC


 *** RECEPTOR  TYPES:  GC = GRIDCART
                       GP = GRIDPOLR
                       DC = DISCCART
                       DP = DISCPOLR
                       BD = BOUNDARY
                                                                   D-2

-------
 *** ISCST2  -  VERSION 93109  ***     *** WAMPLER-LONGACRE:  NOx Emissions;   Sept. 27, 94                         ***         09/27/94
                                     ***                                                                          ***         12:18:34
                                                                                                                             PAGE    9
 *** MODELING  OPTIONS USED:   CONG   RURAL                 DFAULT

                                                  ***  THE SUMMARY OF HIGHEST  3-HR RESULTS  ***


                                         ** CONG  OF NOX      IN  (MICROGRAMS/CUBIC-METER)                  **

                                                       DATE                                                                  NETWORK
GROUP  ID                          AVERAGE CONG      (YYMMDDHH)              RECEPTOR   (XR, YR,  ZELEV,  ZFLAG)      OF  TYPE  GRID-ID


ALL     HIGH  1ST HIGH VALUE  IS        8.38784   ON 90070503: AT  (  703580.00,   4270240.00,       0.00,       0.00)   DC
        HIGH  2ND HIGH VALUE  IS        6.00033   ON 90072624: AT  (  703580.00,   4270240.00,       0.00,       0.00)   DC


 *** RECEPTOR  TYPES:  GC = GRIDCART
                       GP = GRIDPOLR
                       DC = DISCCART
                       DP = DISCPOLR
                       BD = BOUNDARY
                                                                   D-3

-------
     JULY  1990 MESOPUFF II/  ISCST2 MERGING
VERSON=   5.1   LEVEL= 93181  NSYR=90  NSDAY=182   NSHR= 0  NADVTS=   744   IAVG=    3
DGRID=   20000.0  IASTAR=17  IASTOP=46  JASTAR=  4  JASTOP=33  ISASTR=28   ISASTP=35
NAREAS=    0   NREC=  64  IPRINF=   36  LGAUSS=T   LCHEM=T  LDRY=T  LWET=T   LPRINT=T
NSPEC= 5
LWETG=F   LWETNG=T  LDRYG=F  LDRYNG=T  LPRFLX=T
 NPUF=  1  NSAMAD=  1  IELMET=50   JELMET=35
 JSASTR=15   JSASTP=22  MESHDN=  1   NPTS=  10
L3VL=T  LVSAMP=T  WSAMP=   2.00   LSGRID=F
XREC=27.72  27.74 27.88 27.90  27.91 28.00 28.05  28.09 28.19 30.52  30.61 30.65 30.74  30.80  30.84 30.94 30.96  31.01 31.08 31.18
     31.24  31.27 31.28 31.37  31.42 31.51 31.57  31.65 31.67 31.70  31.83 31.85 31.91  31.99  32.02 32.11 32.17  32.23 32.25 32.31
     32.37  32.37 32.38 32.45  32.51 32.56 32.60  32.61 32.68 32.78  32.82 32.86 32.87  32.93  32.94 32.96 33.19  33.20 33.23 33.30
     33.33  33.50 33.52 33.56


YREC=14.05  13.88 13.99 13.83  14.19 13.97 13.85  14.10 14.01 16.93  17.40 17.97 17.62  16.88  18.05 17.71 17.19  18.26 17.45 19.51
     18.08  17.61 18.47 19.30  18.78 17.86 18.41  18.13 18.87 19.11  18.68 19.41 18.38  19.62  18.92 19.23 19.89  18.99 20.58 18.74
     20.09  21.12 19.41 20.86  21.31 20.22 19.22  20.49 19.88 19.59  20.31 21.26 19.84  20.70  20.06 21.02 20.82  21.83 21.48 22.02
     21.16  21.58 20.81 21.26
                                                              D-4

-------
                            THE
REC NO    CONG   (YYMMDDHH)  AT
                                 FIRST HIGHEST OF 3-HR AVERAGE  CONCENTRATION VALUES AT EACH  NON-GRID RECEPTOR FOR
                                    CONCENTRATIONS IN MICROGRAMS/CUBIC-METER, RECEPTOR LOCATIONS IN METERS
                                                                                                                       SO2
RECEPTOR  (X,Y)
REC NO    CONC   (YYMMDDHH)  AT
RECEPTOR  (X,Y)
1
3
5
7
9
11
13
15
17
19
21
23
25
27
29
31
33
35
37
39
41
43
45
47
49
51
53
55
57
59
61
63
4 .
4 .
5.
3 .
4 .
4 .
5.
8 .
5.
5.
5.
6 .
9.
7.
9.
9.
7.
7.
6 .
7.
6 .
7.
7.
10 .
8 .
7.
9.
8 .
8 .
9.
10 .
7.
.8388
.5770
.1683
.9992
.3328
.7055
.2547
.6506
.8416
.1203
.8279
.3105
.7763
.1578
.4161
.3962
.6521
.2901
.4944
.2647
.9826
.6686
.7897
.1468
.3982
.2208
.3857
.8118
.3657
.5684
.1119
.9922
(90072818)
(90072818)
(90072818)
(90072818)
(90072818)
(90072812)
(90072206)
(90072506)
(90072506)
(90072812)
(90072903)
(90072712)
(90072606)
(90072703)
(90072703)
(90072703)
(90072709)
(90072712)
(90072903)
(90072903)
(90072903)
(90072706)
(90071412)
(90072706)
(90072703)
(90072903)
(90072703)
(90072703)
(90071412)
(90072900)
(90072900)
(90071412)
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
( 554300.
( 557540.
( 558200.
( 560960.
( 563740.
( 612200.
( 614760.
( 616860.
( 619260.
( 621660.
( 624740.
( 625700.
( 628400.
( 631360.
( 633480.
( 636560.
( 638280.
( 640339.
( 643440.
( 645059.
( 647340.
( 647540.
( 650260.
( 651940.
( 653660.
( 656400.
( 657300.
( 658840.
( 663820.
( 664620.
( 666540.
( 670420.
.00,
.00,
.00,
.00,
.00,
.00,
.00,
.00,
.00,
.00,
.00,
.00,
.00,
.00,
.00,
.00,
.00,
.94,
.00,
.94,
.00,
.00,
.00,
.00,
.00,
.00,
.06,
.00,
.06,
.00,
.00,
.00,
280980 .
279840 .
283780 .
277060 .
280100 .
348060 .
352420 .
360980 .
343840 .
349020 .
361559.
369320 .
375660 .
368220 .
377320 .
373560 .
367640 .
378320 .
397800 .
411500 .
401720 .
388160 .
426180 .
384420 .
397620 .
406160 .
396740 .
401100 .
416320 .
429620 .
423140 .
416100 .
.00)
.00)
.00)
.00)
.00)
.00)
.00)
.00)
.00)
.00)
.97)
.00)
.03)
.00)
.00)
.00)
.00)
.00)
.00)
.00)
.00)
.03)
.00)
.00)
.00)
.00)
.00)
.00)
.00)
.03)
.00)
.00)
2
4
6
8
10
12
14
16
18
20
22
24
26
28
30
32
34
36
38
40
42
44
46
48
50
52
54
56
58
60
62
64
4 .
4 .
4 .
4 .
5.
6 .
6 .
4 .
8 .
11 .
4 .
6 .
6 .
11 .
7.
6 .
6 .
7.
7.
8 .
9.
8 .
7.
8 .
9.
8 .
8 .
10 .
6 .
4 .
10 .
10 .
.3077
.0606
.4119
.7420
.5231
.5819
.5571
.6043
.3091
.4135
.9670
.1868
.0722
.0045
.8607
.8602
.6451
.4563
.4549
.1755
.1971
.2775
.2484
.0164
.5046
.5742
.7174
.3587
.0817
.8063
.3305
.5191
(90072818)
(90072818)
(90072818)
(90072818)
(90072815)
(90072206)
(90072815)
(90072506)
(90072706)
(90070503)
(90072809)
(90072712)
(90072709)
(90072703)
(90072700)
(90072712)
(90072712)
(90072712)
(90072706)
(90072709)
(90071412)
(90071412)
(90072903)
(90072903)
(90072706)
(90071412)
(90072903)
(90071412)
(90072900)
(90072715)
(90072900)
(90072900)
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
( 554880.
( 558020.
( 559900.
( 561720.
( 610340.
( 612940.
( 615920.
( 618880.
( 620140.
( 623580.
( 625500.
( 627380.
( 630180.
( 633000.
( 634020.
( 637020.
( 639740.
( 642260.
( 644600.
( 646200.
( 647480.
( 648960.
( 651300.
( 652260.
( 655560.
( 657220.
( 658600.
( 659240.
( 664080.
( 666039.
( 669980.
( 671160.
.00,
.00,
.00,
.00,
.00,
.00,
.00,
.00,
.00,
.00,
.00,
.00,
.00,
.00,
.00,
.00,
.00,
.00,
.00,
.00,
.00,
.06,
.00,
.00,
.00,
.00,
.00,
.06,
.00,
.94,
.00,
.00,
277560 .
276620 .
279380 .
281980 .
338580 .
359360 .
337520 .
354199.
365260 .
390239.
352240 .
385940 .
357260 .
362640 .
382180 .
388200 .
392380 .
384640 .
379780 .
374740 .
422440 .
417120 .
404400 .
409880 .
391760 .
425140 .
414100 .
420500 .
436660 .
440379.
431640 .
425180 .
.00)
.00)
.00)
.00)
.03)
.00)
.00)
.97)
.00)
.97)
.00)
.03)
.03)
.00)
.00)
.00)
.00)
.00)
.00)
.00)
.00)
.03)
.00)
.00)
.00)
.00)
.00)
.00)
.00)
.97)
.00)
.00)
                                                                 D-5

-------
                        *** THE SECOND HIGHEST OF 3-HR AVERAGE CONCENTRATION VALUES AT EACH NON-GRID RECEPTOR FOR
                                    CONCENTRATIONS IN MICROGRAMS/CUBIC-METER, RECEPTOR LOCATIONS  IN METERS
                                                                                                                      SO2
REC NO    CONC   (YYMMDDHH)  AT
RECEPTOR  (X,Y)
REC NO    CONC   (YYMMDDHH)  AT
RECEPTOR  (X,Y)
1
3
5
7
9
11
13
15
17
19
21
23
25
27
29
31
33
35
37
39
41
43
45
47
49
51
53
55
57
59
61
63
3 .
3 .
3 .
3 .
3 .
4 .
4 .
5.
5.
4 .
5.
5.
6 .
6 .
7.
6 .
7.
6 .
6 .
5.
6 .
7.
5.
7.
7.
6 .
7.
7.
8 .
5.
9.
7.
.2351
.2662
.6546
.1087
.5060
.3656
.4683
.2634
.6303
.5533
.5488
.4169
.7027
.3759
.0137
.9202
.5083
.2126
.2030
.8111
.0245
.6504
.7656
.7295
.3849
.6103
.8440
.4880
.3175
.4192
.5550
.3090
(90072815)
(90072815)
(90072815)
(90072815)
(90072815)
(90072615)
(90072615)
(90072712)
(90072812)
(90072818)
(90072712)
(90072809)
(90072712)
(90072712)
(90072712)
(90072712)
(90070203)
(90072709)
(90072712)
(90072900)
(90072703)
(90072712)
(90072709)
(90072712)
(90072712)
(90072703)
(90072712)
(90072712)
(90072903)
(90072903)
(90071412)
(90072900)
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
( 554300.
( 557540.
( 558200.
( 560960.
( 563740.
( 612200.
( 614760.
( 616860.
( 619260.
( 621660.
( 624740.
( 625700.
( 628400.
( 631360.
( 633480.
( 636560.
( 638280.
( 640339.
( 643440.
( 645059.
( 647340.
( 647540.
( 650260.
( 651940.
( 653660.
( 656400.
( 657300.
( 658840.
( 663820.
( 664620.
( 666540.
( 670420.
.00,
.00,
.00,
.00,
.00,
.00,
.00,
.00,
.00,
.00,
.00,
.00,
.00,
.00,
.00,
.00,
.00,
.94,
.00,
.94,
.00,
.00,
.00,
.00,
.00,
.00,
.06,
.00,
.06,
.00,
.00,
.00,
280980 .
279840 .
283780 .
277060 .
280100 .
348060 .
352420 .
360980 .
343840 .
349020 .
361559.
369320 .
375660 .
368220 .
377320 .
373560 .
367640 .
378320 .
397800 .
411500 .
401720 .
388160 .
426180 .
384420 .
397620 .
406160 .
396740 .
401100 .
416320 .
429620 .
423140 .
416100 .
.00)
.00)
.00)
.00)
.00)
.00)
.00)
.00)
.00)
.00)
.97)
.00)
.03)
.00)
.00)
.00)
.00)
.00)
.00)
.00)
.00)
.03)
.00)
.00)
.00)
.00)
.00)
.00)
.00)
.03)
.00)
.00)
2
4
6
8
10
12
14
16
18
20
22
24
26
28
30
32
34
36
38
40
42
44
46
48
50
52
54
56
58
60
62
64
2 .
2 .
3 .
3 .
5.
4 .
5.
4 .
5.
10 .
4 .
5.
5.
6 .
7.
5.
6 .
6 .
7.
7.
6 .
7.
6 .
6 .
9.
7.
7.
9.
4 .
4 .
4 .
8 .
.9588
.9756
.3077
.6305
.4423
.9327
.9293
.5566
.7724
.4414
.8869
.4463
.8470
.7959
.0098
.9444
.1067
.4216
.4279
.4911
.3134
.3222
.1225
.8271
.1159
.9734
.9002
.2875
.7874
.6170
.7610
.1059
(90072815)
(90072815)
(90072815)
(90072815)
(90072812)
(90072712)
(90072812)
(90072712)
(90072712)
(90072700)
(90072812)
(90072600)
(90072809)
(90072709)
(90072712)
(90072703)
(90072703)
(90072706)
(90072712)
(90072706)
(90072709)
(90072903)
(90072703)
(90072900)
(90072703)
(90072900)
(90072900)
(90072903)
(90072803)
(90072900)
(90072621)
(90071412)
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
AT
( 554880.
( 558020.
( 559900.
( 561720.
( 610340.
( 612940.
( 615920.
( 618880.
( 620140.
( 623580.
( 625500.
( 627380.
( 630180.
( 633000.
( 634020.
( 637020.
( 639740.
( 642260.
( 644600.
( 646200.
( 647480.
( 648960.
( 651300.
( 652260.
( 655560.
( 657220.
( 658600.
( 659240.
( 664080.
( 666039.
( 669980.
( 671160.
.00,
.00,
.00,
.00,
.00,
.00,
.00,
.00,
.00,
.00,
.00,
.00,
.00,
.00,
.00,
.00,
.00,
.00,
.00,
.00,
.00,
.06,
.00,
.00,
.00,
.00,
.00,
.06,
.00,
.94,
.00,
.00,
277560 .
276620 .
279380 .
281980 .
338580 .
359360 .
337520 .
354199.
365260 .
390239.
352240 .
385940 .
357260 .
362640 .
382180 .
388200 .
392380 .
384640 .
379780 .
374740 .
422440 .
417120 .
404400 .
409880 .
391760 .
425140 .
414100 .
420500 .
436660 .
440379.
431640 .
425180 .
.00)
.00)
.00)
.00)
.03)
.00)
.00)
.97)
.00)
.97)
.00)
.03)
.03)
.00)
.00)
.00)
.00)
.00)
.00)
.00)
.00)
.03)
.00)
.00)
.00)
.00)
.00)
.00)
.00)
.97)
.00)
.00)
                                                                D-6

-------
                               THE
                                     10 MAXIMUM OF 3-HR AVERAGE CONCENTRATION VALUES AT EACH NON-GRID RECEPTOR FOR
                                       CONCENTRATIONS IN  MICROGRAMS/CUBIC-METER,  RECEPTOR  LOCATIONS IN METERS
                                                                                                                         SO2
    RANK    CONC   (YYMMDDHH)  AT
          (X,Y)
                                                        RECEPTOR NO.
                                                                             RANK
                                                                                    CONC
               (YYMMDDHH) AT
       (X,Y)
                                                                                                                                 RECEPTOR
NO.
     1    11.4135  (90070503)  AT
     3    10.5191  (90072900)  AT
     5    10.3587  (90071412)  AT
     7    10.1468  (90072706)  AT
     9     9.7763  (90072606)  AT
(  623580.00, 390239.97)     20
(  671160.00, 425180.00)     64
(  659240.06, 420500.00)     56
(  651940.00, 384420.00)     47
(  628400.00, 375660.03)     25
 2   11.0045  (90072703) AT
 4   10.4414  (90072700) AT
 6   10.3305  (90072900) AT
 8   10.1119  (90072900) AT
10    9.5684  (90072900) AT
(  633000.00, 362640.00)     28
(  623580.00, 390239.97)     20
(  669980.00, 431640.00)     62
(  666540.00, 423140.00)     61
(  664620.00, 429620.03)     59
                                                                    D-7

-------