WRAP 2003 Emissions FinalReport v2

Document Sample
WRAP 2003 Emissions FinalReport v2 Powered By Docstoc
					                                                                     July 20, 2004




2003 Final Report:
WRAP-RMC Emissions Modeling
Support
Contract number: 30203-34




Prepared for:

Western Regional Air Partnership Modeling Forum
Western Governors’ Association


Prepared by:
Zac Adelman a
Mohammad Omary b


a
    UNC-Chapel Hill, Carolina Environmental Program
    Bank of America Plaza, CB# 6116
    137 E. Franklin St.
    Chapel Hill, NC 27599-6116
b
    UC-Riverside, Center for Environmental Research and Technology
    1084 Columbia Ave
    Riverside, CA 92507
                                                           Contents
1   Executive Summary ............................................................................................................... 2
2   Introduction ............................................................................................................................ 2
3   Final §309 Emissions Modeling............................................................................................. 3
    3.1 Task Description ............................................................................................................ 3
          3.1.1 Inventory and land use data .............................................................................. 3
          3.1.2 SMOKE ancillary inputs ................................................................................... 6
    3.2 SMOKE Modeling Details ............................................................................................. 6
    3.3 Emissions quality assurance ........................................................................................... 6
4   SMOKE Updates .................................................................................................................. 12
    4.1 Task Description .......................................................................................................... 12
    4.2 Task 1CD: Temporal, Speciation, SCC Improvements ............................................... 12
    4.3 Install SMOKEv2 at the RMC ..................................................................................... 14
    4.4 Create Automated QA Regression Tool ...................................................................... 14
5   Emissions QA Protocol ........................................................................................................ 16
    5.1 Task Description .......................................................................................................... 16
    5.2 Protocol Development .................................................................................................. 16
    5.3 Protocol Refinements ................................................................................................... 17
    5.4 Protocol Evolution ....................................................................................................... 18
6   Preliminary §308 Emissions Modeling ............................................................................... 19
    6.1 Task Description .......................................................................................................... 19
    6.2 SMOKE Input Files ..................................................................................................... 21
          6.2.1 Inventory and land use data ............................................................................ 21
          6.2.2 SMOKE ancillary inputs ................................................................................. 27
    6.3 SMOKE Modeling Details ........................................................................................... 27
          6.3.1 SMOKE Scripts ............................................................................................... 28
          6.3.2 Configuration .................................................................................................. 29
          6.3.3 Simulation description .................................................................................... 32
    6.4 SMOKE processing for WRAP source categories ....................................................... 32
          6.4.1 Area emissions processing .............................................................................. 33
          6.4.2 On-road mobile-source emissions processing ................................................ 34
          6.4.3 Point-source emissions processing ................................................................. 35
          6.4.4 Biogenic emissions processing ........................................................................ 35
    6.5 Problems encountered and remaining processing issues ............................................. 36


                                                                     ii
             6.5.1 Emissions inventory issues .............................................................................. 36
             6.5.2 SMOKE ancillary input issues ........................................................................ 37
             6.5.3 SMOKE configuration issues .......................................................................... 39
             6.5.4 Unresolved emissions issues ........................................................................... 40
    6.6      Emissions quality assurance......................................................................................... 40
             6.6.1 QA Products .................................................................................................... 41
7   Conclusions ........................................................................................................................... 46
8   References ............................................................................................................................. 47




                                                                    iii
                                                                 Tables
Table 2-1. 2003 WRAP RMC work plan emission tasks ............................................................... 3
Table 3-1. emissions inventory input files for the updated 1996 base case .................................... 4
Table 3-2. ancillary files used to process final version of 1996 inventory ..................................... 6
Table 6-1. Pre02a_36 SMOKE domain configuration ................................................................. 19
Table 6-2. Sources of emissions inventory data for the WRAP modeling domain regions. ........ 22
Table 6-3. Break down of 2002 interim emissions by source category (US emissions only) ...... 25
Table 6-4. Summary of SMOKE scripts ....................................................................................... 28
Table 6-5. Representative model days for Pre02a_36 simulation ................................................ 30
Table 6-6. 2002 modeled holidays ................................................................................................ 30
Table 6-7. Assignments of months to seasons for use of seasonal inventory files in SMOKE ... 31
Table 6-8. 36 km spatial surrogates with descriptions .................................................................. 31
Table 6-9. WRAP source categories mapped to processing approaches in SMOKE ................... 32
Table 6-10. New WRAP nonroad SCC profile mapping .............................................................. 38




                                                                Figures
Figure 3-1. Vertical daily total emission inventory. ....................................................................... 8
Figure 3-2. Daily total emission inventory for the entire domain................................................... 8
Figure 3-3. Diurnal total emission inventory for the entire domain for a week day. ...................... 9
Figure 3-4. Diurnal total emission inventory for the entire domain for a week end....................... 9
Figure 3-5. Inventory vertical distribution. Inventory vertical distribution. ................................. 10
Figure 3-6. Percentage of contribution of each source to the total CO emissions. ....................... 10
Figure 3-7. Percentage of contribution of each source to the total NOX emissions. .................... 11
Figure 3-8. Percentage of contribution of each source to the total SOX emissions ..................... 11
Figure 3-9. Percentage of contribution of each source to the total NH3 emissions...................... 12
Figure 4-1. SMOKE Regression Tool interface ........................................................................... 15
Figure 5-1. Revised emissions modeling flow chart ..................................................................... 18
Figure 6-1. Spatial plot of area source NH3 emissions; August 30, 2002 .................................... 42
Figure 6-2. Spatial plot of biogenic source isoprene emissions; August 27, 2002 ....................... 42
Figure 6-3. Diurnal time-series plot of area source NO emissions; February 1, 2002 ................. 43
Figure 6-4. Diurnal time-series plot of on-road mobile soured NO emissions; June 1, 2002 ...... 43
Figure 6-5. Vertical layer profile for point source SO2; June 14, 2002 ........................................ 44
Figure 6-6. Vertical layer fractions for point source SO2; June 14, 2002 ..................................... 44
Figure 6-7. Episode time series plot of point source NOx emissions; Winter/Summer, 2002 ..... 45
Figure 6-8. Episode time series plot of on-road mobile source NOx emissions; Winter/Summer,
     2002....................................................................................................................................... 45




                                                                        iv
1 Executive Summary
This report summarizes the emissions related tasks completed by the Western Regional
Air Partnership Regional Modeling Center (WRAP-RMC) under the WRAP 2003
Modeling Support contract (WGA no. 30203-33). The five emissions related tasks
contained in the 2003 work plan that the RMC emissions modeling team completed are
described in terms of the scope of the task, how the work was performed, deliverables,
what problems were encountered and how they were addressed. The sequence of the
tasks in this report follows the order that they are listed in the 2003 work plan. Two
modeling-related tasks that involved the generation of model-ready emissions are
described in detail as are three development tasks for improving the emissions modeling
system used by the WRAP-RMC. We provide discussion on the outstanding issues for
each of the tasks and recommendations for how they may be addressed in the future.

2 Introduction
This document is the final report for the emissions modeling tasks of the project titled
“WRAP 2003 Modeling Support” under WGA contract number 30203-33. The project
was based on a contract between the Western Governors’ Association and the University
of California at Riverside (UCR) on behalf of the Western Regional Air Partnership
(WRAP) Modeling Forum. UCR subcontracted the Carolina Environmental Program
(CEP) at the University of North Carolina (UNC) to provide emissions modeling support
to the project. We anticipate that the audience for this report is the WRAP Modeling
Forum, other Regional Planning Organizations (RPOs), and air quality scientists at the
U.S. Environmental Protection Agency (EPA) and elsewhere. This document describes
the technical results of the emissions modeling tasks of the project and summarizes the
work that we have performed with the WRAP. Our intent is to both document our work
and provide guidance and insight into performing similar analyses in the future.
The purpose of this project has been to assist the WRAP in conducting regional haze
modeling and analyses over the western United States in support of State and Tribal
Implementation Plan (SIP and TIP) development, pursuant to Title 40 of the Code of
Federal Regulations (CFR) § 51.308 and 51.309. As part of the WRAP’s overall effort, it
formed a Regional Modeling Center (RMC) at UCR to provide air quality modeling,
analysis, and technical support as needed to the WRAP participants. Initiated in 2001, the
RMC formed as a partnership between UCR, CEP, and ENVIRON Corporation. A
transition year, 2003 saw changes in both the organization at CEP and the RMC
modeling tasks. Most of the atmospheric modeling group at MCNC in Research Triangle
Park, NC moved in January, 2003 to CEP in Chapel Hill. The WRAP Project Manager at
CEP changed from Marc Houyoux to Zac Adelman in June, 2003 (McNelis, 2003). In
addition to these organizational changes, the RMC modeling tasks transitioned from
supporting modeling and analysis for §309 SIPS/TIP development to the preparation for
modeling in support of §308 SIP/TIP development.
The emissions tasks in the 2003 RMC Work Plan included finishing the §309 modeling
and developing an accompanying technical support document (TSD), improvements to
the Sparse Matrix Operator Kernel Emissions (SMOKE) model, development of an
emissions modeling quality assurance (QA) protocol, and SMOKE modeling of an


                                            2
interim 2002 emissions inventory for supporting §308 requirements. Table 2-1
summarizes the emissions modeling tasks in the Work Plan and maps these tasks to the
different sections of this report in which they are described in detail. The organization of
this document will follow the order of the tasks as they are listed in Table 2-1.

                  Table 2-1. 2003 WRAP RMC work plan emission tasks
Task        Section Description
1CD         4          Temporal allocation and speciation improvements to the WRAP
                       point and area sources
2ABC        4,5        Improved SMOKE emissions processing system
2E          6          Process and compare 1996 emissions grown to 2002
4DEFH       7          Air quality model development
No Task     3          Finish §309 modeling and compose TSD



3 Final §309 Emissions Modeling
3.1 Task Description
The purpose of task is to process the most updated emission inventories for 1996 with
updated temporal allocation, and speciation profiles, mcip22 meteorology data, and two
speciation mechanisms, CBIV and SAPRC99. Additional CMAQ input files were
generated by merging the inventory that was processed with CBIV and the wind blown
dust.

3.1.1 Inventory and land use data
Emissions inventory for 1996 was processed with mcip22 meteorology inputs. The input
files for the updated 1996 is listed in table 3-1. The major changes in the emissions
inventory input files were in the point source for WRAP states which were received from
CEP on November 11, 2003. The Agriculture no smoke management fire emissions
inventory was added to the final 1996 inventory for §309.




                                              3
                                Table 3-1. emissions inventory input files for the updated 1996 base case


                                                                              Spatial Coverage

                                                                                        East
Source        Filename                          Source: Date       WRAP     Tier1 Tier2 US Mex Can       Description


              ar96_WRAP_wrapTier1_073001.ida    PES: 7/30/01           x      x                          Annual 1996 area source inventory
                                                                                                         Annual 1996 area source
              ar96_wrapTier2.nei96.060801.ida   CEP: 6/08/01                        x                    inventory, derived from NEI96
                                                                                                         Annual 1996 area source
Area          ar96_eastUS.nei96.060801.ida      CEP: 6/08/01                               x             inventory, derived from NEI96
                                                                                                         Canadian area, on-road, non-road
                                                                                                         for 1996 National Canadian
              ar_mb96_canada.060801.ida         CEP: 6/08/01                                         x   Inventory
                                                                                                         Mexican area, on-road, non-road
                                                                                                         mobile; derived by CEP from PES
              ar96_Mexico_060701.ida            PES: 6/07/01                                     x       spreadsheet "for CMAQ Mexico"
                                                                                                         Seasonal 1996 nonroad mobile
              wrap_nr96_${season}.ida           Environ: 6/07/01       x                                 inventory
                                                                                                         Annual 1996 nonroad mobile
              nr96_wrapTier1.nei96.ida          CEP: 6/08/01                  x                          inventory, derived from NEI96
Non Road
                                                                                                         Annual 1996 nonroad mobile
              nr96_wrapTier2.nei96.060801.ida   CEP: 6/08/01                        x                    inventory, derived from NEI96
                                                                                                         Annual 1996 nonroad mobile
              nr96_east.nei96.060801.ida        CEP: 6/08/01                               x             inventory, derived from NEI96
                                                                                                         Annual 1996 road dust inventory,
Road Dust -   pvd96_wrap_env.ida                Environ: 6/19/02       x                                 derived from NEI96 by Environ
Paved                                                                                                    Annual 1996 road dust inventory,
              pvd96_wrapTier1.nei96.ida         CEP: 6/08/01                  x                          derived from NEI96




                                                                   4
                                                                                                 Annual 1996 road dust inventory,
                pvd96_wrapTier2.nei96.060801.ida   CEP: 6/08/01                      x           derived from NEI96

                                                                                                 Annual 1996 road dust inventory,
                pvd96_east.nei96.060801.ida        CEP: 6/08/01                          x       derived from NEI96
Road Dust -
Unpaved         unp_96emis${season}_env.ida        Environ: 6/19/02        x                     Seasonal 1996 unpaved road dust
                                                                                                 Annual 1996 point sources
                pt1996v3_smk.TXT                   Environ: 11/05/03       x     x   x           inventory
Point
                                                                                                 Annual 1996 point source
                ptinv_east_nwrap.nei96.ida         CEP: 6/08/01                          x       inventory, derived from NEI96
                pt96_Mexico_061101.ida                                                       x
                pthour_wrap1996_wf_102601.ems                              x                     Hourly plume distribution
Wildfire        ptday_wrap1996_wf_102601.ems       Air Sci.                x                     Daily 1996 wild fires inventory
                ptinv_wrap1996_wf_102501.ida                               x                     Lat_Long for the fire locations
                pthour_wrap1996_agnsm_073102.${m
                onth}.ems                                                  x                     Hourly plume distribution

AgFire NSM      ptday_wrap1996_agnsm_073102.${mo Air Sci.
                nth}.ems                                                   x                     Daily Inventory
                ptinv_wrap1996_agnsm_073102.${mo
                nth}.ida                                                   x                     Lat_Long for the fire locations
                ca_or96_${season}.ida                                  CA only                   Seasonal 1996 mobile inventory
                wrap_or96_${season}.ida                                    x                     Seasonal 1996 mobile inventory
                Vmt9631x_Tier1.ida                                               x               Annual miles per speed category
                Vmt9631x_Tier2.ida                                                   x           Annual miles per speed category
Mobile          Vmt9631x_east.ida                                                        x       Annual miles per speed category
                mb96h${month}_Tier1.ida            CEP: 6/08/01                  x               Monthly Inventory for HDDV
                mb96h${month}_Tier2.122101.ida                                       x           Monthly Inventory for HDDV
                mb96h${month}_east.122101.ida                                            x       Monthly Inventory for HDDV
                MOBIL5 inputs
Biogenic        BEIS2/1996 Met                     CEP: 6/08/01            x     x   x   x       Gridded inventory
Fugative Dust                                      RMC                     x     x   x   x       Processsed inventory (ntCDF)



                                                                       5
3.1.2 SMOKE ancillary inputs

The temporal profiles and special allocation files (surrogates) were the same as
in processing 2018. The inventory was processed with two speciation
mechanisms, CBIV and SAPRC99. The speciation profiles file that was
distributed with SMOKE2 was used for CBIV speciation. A new speciation file for
SAPRC99 was generated using the most updated profiles from Bill Carter.



Table 3-2. ancillary files used to process final version of 1996 inventory
            Type of
Source      ancillary       File name
            Special         agref.m3.062101.txt
Area
            allocation      surratio.cmaq_w36.072001.txt
            Special         abmgpro.m3.040402.us36+can+mex.txt
Mobile
            allocation      amgref.m3.us+mex.txt
            Temporal        amptpro.m3.121702.us+can.txt
All
            allocation      amptref.m3.032003.us+can.WRAP.txt
                            gspro.cmaq.cb4p25.txt
All         Speciation
                            gsref.cmaq.cb4p25.txt
                            gspro.EPAS99LP_ssppr.121103.dat
All         Speciation
                            gsref.cmaq.cb4p25.txt


3.2 SMOKE Modeling Details
SMOKEv2 was installed in October, 2003. For installation used the precompiled
SMOKE executables for Red Hat Linux which was downloaded from CEP SMOKE web
site. The downloaded files included SMOKE executables, and general run scripts.
New scripts were prepared for the this case to read the proper inventory input files, new
meteorology files, and ancillary files. The names of all new scripts and assign files for
this case included the label “1996.cbiv_wrap_cmaq_w36h” when we used the cbiv
speciation mechanism or “1996.saprc_wrap_cmaq_w36h” when we used SAPRC99
mechanism.

3.3 Emissions quality assurance
The RMC team has developed a post processing program which QA’s the final outputs of
SMOK. The program was run for each inventory source (are, point, etc.) individually and
for the merged inventory.
The purpose of the post processing QA is:
       i.   To make sure that all the no region or state was dropped out from the
            emissions inventory files.



                                             6
        ii.    To insure that the daily changes (week days, week ends, and holily days) are
               processed correctly, and the seasonal changes are correct.
    iii.       To insure that diurnal profiles of week days and week ends are used.
    iv.        To insure that the point source were being distributed properly into the
               vertical layers.
If any unexpected or unusual behavior of the processed inventory then a more detailed
quantitative QA was conducted using SMOKE reports. All the QA plots were posted on
the web page for the 309 modeling:
http://pah.cert.ucr.edu/rmc/1996/base_mcip22/qa.shtml
For each source category we do the following post processing QA:
a. Data output summary files:
   i.         Daily vertical total from all the layers (netCDF file)
   ii.        Domain daily total data: this is one file that has the total inventory for each
              pollutant for the entire domain for each day (text file).
   iii.       Domain hourly total data: this is hourly time series for each day for total
              domain inventory for each pollutant (text file).
   iv.        Domain daily layer total data: total domain inventory or each layer and for each
              day (text file).
b. Plots:
   i.         Daily vertical total gif plots: one plot per pollutant per day for the domain
              vertical daily total inventory (Figure 3-1).
   ii.        Domain daily total plots: one plot per pollutant for the entire simulation period
              (Figure 3-2)
   iii.       Domain hourly total plots: diurnal time series for the total domain inventory.
              One plot per day per pollutant (Figure 3-3 and Figure 3-4)
   iv.        Domain daily layer total emissions plots: vertical distribution for the domain
              total inventory. One plot per day per pollutant Note: The layer QA is done only
              for point sources and fires (Figure 3-4).
The inventory of this was intensively QA’ed for the pervious runs of 1996 emissions.
But the outputs of the post processing program was used to generate the chart pies
showing the percentage contribution of each source the various pollutants, Figures 3-6 to
Figure 3-9.




                                                  7
Figure 3-1. Vertical daily total emission inventory.




Figure 3-2. Daily total emission inventory for the entire domain.




                                             8
Figure 3-3. Diurnal total emission inventory for the entire domain for a week day.




Figure 3-4. Diurnal total emission inventory for the entire domain for a week end.




                                         9
Figure 3-5. Inventory vertical distribution. Inventory vertical distribution.




Figure 3-6. Percentage of contribution of each source to the total CO emissions.

                                 1996 CO (%)



                               0.94%



             25.95%                              27.16%
                                                                Area
                                                                point
                                                                Mobile
                                                                non Road

                                                        8.88%
                                                                W. Fire
                                                                Agf
                                                3.80%

                      33.26%




                                           10
Figure 3-7. Percentage of contribution of each source to the total NOX emissions.

                                       1996 NOX (%)



                           9.67%             7.23%
                   0.06%
                1.00%

                                                                        Area
                                                               18.63%
          14.51%
                                                                        point
                                                                        Mobile
                                                                        non Road
                                                                        W. Fire
                                                                        Agf
                                                                        Bio


                                    48.90%




Figure 3-8. Percentage of contribution of each source to the total SOX emissions

                                       1996 SOX (%)

                               1.33%      0.02%
                            5.57%            8.14%




            17.44%                                                      Area
                                                                        point
                                                                        Mobile
                                                                        non Road
                                                                        W. Fire
                                                                        Agf

                                                      67.50%




                                                     11
Figure 3-9. Percentage of contribution of each source to the total NH3 emissions.

                                           1996 NH3 (%)

                                   1.05%
                                  0.08%




                                                                       Area
                                                             41.30%
                                                                       Mobile

                                                                       non Road
                 57.44%

                                                                       W. Fire

                                                     0.14%




4 SMOKE Updates
4.1 Task Description
Tasks 1CD and 2ABC in the 2003 Work Plan contained items related to improving the
SMOKE model, SMOKE input files, installing SMOKEv2 at the RMC and automating
QA procedures for checking SMOKE output emissions. The goal for these tasks was to
improve the SMOKE modeling system based on our experiences with the §309 modeling
that we performed in 2001-2002.

4.2 Task 1CD: Temporal, Speciation, SCC Improvements
SMOKE uses tabulated look-up tables of temporal and chemical allocation profiles to
distribute anthropogenic emissions in time and to a parameterized chemical mechanism,
respectively. The basis of these look-up tables is the source classification code (SCC) for
defining which sources receive the different profiles. The purpose of task 1CD was to
review the SCCs in the WRAP area and point source emissions inventories and to explore
the temporal and chemical profiles applied to these SCCs and make any necessary
improvements. To accomplish this task, we used SMOKE to generate annual emissions
summaries for the WRAP area and point source inventories. We distilled the summaries
into the top 80% of the sources for only the significant pollutants in each source category
(Tonnesen et al., 2003). We created unique reports for each of the 13 WRAP states and
posted them to the CEP FTP site for distribution. We distributed a memorandum with the


                                              12
state reports on September 30, 2003 that explained how to interpret and review the
reports and communicate any changes back to the CEP (Adelman, 2003). We revised the
summaries to use a new version of the point source inventory and reissued a call for
comments on February 12, 2004 (Adelman, 2004).
Between the two comment periods six of the thirteen WRAP states acknowledged
receiving the request but only one, New Mexico, provided actual feedback on the
summaries. In compiling the summaries, the CEP made some recommendations for
updates to the temporal and speciation profile assignments. Supplemental
recommendations for updates to the temporal and speciation profile assignments also
comes with a new set of files produced by the U.S. EPA that are currently under review
and will be released by Summer, 2004 (EPA, 2003).
The deliverables for Task 1CD in reviewed/improved SCC assignments, improved
temporal profile assignments, and improved chemical profile assignments have not been
completed as of the writing of this report. As mentioned previously a little less than half
of the WRAP states acknowledged the request for comments on the inventory and profile
summaries. Of the six states to respond, only one, New Mexico, sent actual comments
back to the CEP. Colorado has recently been in contact with CEP to clarify what is
expected in the review and we anticipate receiving comments from Colorado in March,
2004. The CEP will make another request for comments before compiling the
information into revised SCC assignments and profiles. The combination of the
comments from CEP, comments from a few of the WRAP states, and a preliminary
version of the new EPA profiles will ultimately produce updates to the SMOKE ancillary
input files before the generation of final emissions for the §308 modeling starting in
April, 2004.
Task 1CD was difficult to complete as stated in the Work Plan for a few reasons.
   1) The reality of the staff support available at the different departments of air quality
      (DAQs) throughout the WRAP region was not clearly defined. As the level of
      expertise differs in the DAQs, the ability of these organizations to handle data
      review requests will also differ. Despite the high level of cooperation between the
      WRAP participants, many of the states and tribes do not yet have the staff to
      commit to data reviews. It was also not clear to CEP staff what states were in
      agreement that they would take part in the survey. With a better initial plan, the
      CEP could have targeted their efforts to accommodate only those states that
      committed to reviewing the data in the time frame of this part of the project.
   2) A clear communication line between CEP and the data review staffs at the DAQs
      was never established. As a large amount of communication took place between
      CEP and those states that participated in the data reviews, it became clear that
      CEP needed a list of contacts for each participating state or tribe to follow up with
      at set intervals through the review process. Using the WRAP Modeling Forum
      listserv was not effective in establishing communication between the WRAP
      participants and CEP. Direct email and occasional phone contact is required to
      complete this review. A question and answer conference call between CEP and
      all of the participating WRAP reviewers would have been an efficient means for
      initiating Task 1CD.


                                            13
   3) The way that Task 1CD was defined in the Work Plan provided insufficient point
      source inventory information for the states to complete the reviews. The Work
      Plan did not explicitly state that the reports needed to be at the county level, only
      the state and SCC level. For point sources, defining a plant ID at the state level is
      meaningless because the plant ID’s are county specific, i.e. there could be a
      facility using a plant ID of “1” in every county in a state. In addition to reporting
      the emissions at the county level, we also added plant names to the summaries to
      facilitate the identification of the individual sources by the data reviewers. CEP
      regenerated the point source reports to include county level information and
      posted them to their FTP site.
We recommend that the WRAP maintain an open task for assimilating all new
information from the WRAP states and Tribes as it becomes available. Based on the
range of experience and time frames that the different WRAP constituents are working
with, a mechanism needs to be in place to continually review the emissions data and
assimilate new information. The formation of a data review committee or forum would
be consistent with the outside review and gate keeping requirements of the WRAP
emissions QA protocol discussed in Section5.

4.3 Install SMOKEv2 at the RMC
Task 2ABC of the 2003 Work Plan called for the installation of SMOKE version 1.5 on
the RMC computing cluster and testing using the WRAP inventories. The Carolina
Environmental Program released SMOKE version 1.5 in March, 2003 and version 2.0 in
September, 2003. During 2003 CEP completed task 2ABC and extended it to also install
SMOKE version 2 for the §308 modeling tasks.
CEP completed the upgrade from SMOKE version 1.3.2 to version 1.5 on the RMC
computing cluster in July, 2003. Due to various issues with using different Fortran
compilers to build SMOKE, we moved to using precompiled SMOKE executables
created at the CEP and then installed on the RMC cluster (Houyoux, 2003). We installed
all of the source code and libraries on the RMC cluster, but did the actual compiling at
CEP.
CEP completed the upgrade from SMOKE version 1.5 to version 2.0 on the RMC
computing cluster in October, 2003. We tracked the Installation and benchmark on the
RMC machines using the web-based help ticket tracking system Bugzilla. The record of
the installation is available at the following URL:
http://bugz.unc.edu/show_bug.cgi?id=789.
SMOKE version 2.0 is now the operational version of the model that we are using for all
emissions modeling at the RMC.

4.4 Create Automated QA Regression Tool
A key step in the emissions QA process involves comparing summaries prepared by the
contractor who created the emissions inventories to reports generated by SMOKE at
different points in the modeling process. In addition to comparing inventory and
SMOKE totals, we routinely inter-compare the SMOKE reports created at each step in
the modeling (i.e. spatial allocation, temporal allocation, etc.). Our standard method for


                                            14
making these comparisons is to import the reports into Microsoft Excel spreadsheets and
create linear regression charts to graphically display the data and calculate R2 values to
quantify the comparison. Task 2ABC contained a development item for creating a tool to
automate the generation of these emissions regressions.
CEP created the SMOKE Regression Tool (SRT) to automate the creation of the
emissions regression plots (Figure 1). Developed in Python (http://www.python.org/), we
created the SRT to be platform independent, easy to install, and efficient to use. A beta
version of the SRT is available for download off the CEP ftp site:
ftp://ftp.unc.edu/pub/empd/smoke/SMOKE_RegTool.zip




                     Figure 4-1. SMOKE Regression Tool interface
Features of the SRT include,
       import and matching of state, county, and SCC level SMOKE reports.
       a probing feature to identify outliers on the regression.
       export of regression graphics in several different file formats.
       display of the regression equation.
CEP created a functional regression tool per the description in Task 2ABC in the 2003
Work Plan. The tool could now be enhanced by continuing development to add the
following features,
       configuration menu for formatting axes, fonts, titles, etc.
       script driven interface to fully automate the program.




                                              15
The alternative to a tool written in Python is a Visual Basic (VB) program that would
function as a Microsoft Excel macro. An advantage to a VB macro is that that
functionality that currently exists in Excel would be available to the tool, such as drag
and drop configuration, the Office interface, and the ability to manipulate the data before
it is plotted. The drawbacks to a VB macro are the constraints in Excel, such as the
requirement that it must be run on Windows machines, it is proprietary software, and the
limitations to the file formats that can exported from the program.
Our rationale behind developing in Python was to create a platform-independent tool that
extends the functionality of our current tool (Excel) to allow script driven plot generation.
CEP proposes to make the SRT an open source development project to which others in
the SMOKE/emissions modeling community can contribute. To fully realize the
potential of the SRT it will require additional development effort to integrate the script-
driven functionality that we envision it will use to be a fully automated QA tool. The
funding provided in Task 2ABC initialized the project and produced a working regression
tool. The next step is to complete the SRT as it was originally designed by adding the
two features described above to enhance the flexibility of the tool and automate its use.

5 Emissions QA Protocol
5.1 Task Description
The last of the SMOKE model improvements listed in Task 2ABC in the 2003 Work Plan
called for the development of an emissions modeling QA protocol. We needed a well
defined protocol for QA of the regulatory emissions modeling that established rules and
procedures for ensuring the accuracy of emissions data and for documenting the
emissions modeling process. Without predefined procedures and requirements for QA
benchmarks and reports, production-driven tasks will inevitably focus on getting results
out the door and skimp on verifying their quality. By clearly defining the expectations for
data quality and documentation at the beginning of the modeling process, a modeling
team can plan to accommodate the additional work that will be required to ensure high-
quality results.

5.2 Protocol Development
CEP developed a first draft of an emissions QA protocol based on over a decade of
experience developing and running emissions models for regulatory applications. While
the ideas for an emissions QA protocol were propagated over several years of discussion
and experience, the resources to compile these ideas into a document were not available
until the WRAP sponsored the creation of a protocol in 2003. CEP developed a
preliminary draft of the protocol in 2nd and 3rd quarters of 2003 and presented it October
8, 2003 at the WRAP Modeling Forum technical meeting at UCR (Adelman, 2003).
Based on comments received at the meeting and experience gained in implementing the
protocol during the modeling of the interim 2002 inventory (Section 6), CEP revised the
protocol several times before distributing a draft copy to the RMC modeling team on
January 8 , 2004 (Adelman, 2004). The distributed copy of the protocol has been revised
further by CEP as the interim 2002 modeling progressed. We added new QA checks and



                                             16
tables based on specific experiences in the 2002 modeling and revised the operational
analysis sections based on the reality of the modeling work.

5.3 Protocol Refinements
One of the goals for the preliminary 2002 modeling task (Task 2E) that the RMC
performed in 2003 was to test the implementation of the emissions modeling QA
protocol. While the protocol borrowed concepts from the experiences of past modeling
projects, both for the WRAP and for other agencies, the modeling team structure
proposed by the protocol was untested. When we implemented the proposed structure we
discovered that the definition of tasks and workflow between the different members of
the modeling team were not well enough defined and led to some ambiguity in the
assignment of different tasks within the protocol. The modeling team defined by the
protocol consists of a project manager, a lead emissions modeler, a QA manager, a QA
modeler, and a gatekeeper. Our experiences with Task 2E lead to combining the roles of
the QA manager and QA modeler.
We refined the roles of the QA manager and QA modeler as defined in the first draft of
the QA protocol because of ambiguity in how these two positions differed. In the actual
modeling completed under Task 2E, the lead emissions modeler performed most of the
work proposed for the QA modeler. As a result, the QA modeler ended up doing most of
the work slated for the QA manager. By combining the roles of the QA manager and
modeler into one all-encompassing QA manager position, we clarified the delineation of
the roles between the lead modeler and the QA manager positions. All of the tasks
originally assigned to the QA manager continue to be assigned to that position. The QA
manager is still responsible for ensuring that the QA checklist and products are
completed and maintained. The tasks assigned to the QA modeler have been split
between the lead modeler and the QA manager. The lead modeler is responsible for
creating all of the SMOKE QA reports and the QA manager for compiling the SMOKE
QA reports into the products defined in the emissions QA protocol. The task of report
writing has been shifted from the QA manager to the combined team of the project
manager and lead modeler, as the people in these two roles will be most familiar with the
progress of the modeling. Figure 5-1 graphically depicts the revised emissions modeling
and QA process developed through working on Task 2E.




                                           17
                  Figure 5-1. Revised emissions modeling flow chart

5.4 Protocol Evolution
The current working version of the emissions QA protocol is very useful in that it clearly
outlines a strict set of procedures for completing emissions QA using a team of modelers.
Strength of the protocol lies in its analysis of the operational aspects of emissions QA,
information flow, and communication between the modelers. The implementation of
some of these operational items still needs refinement however.
A final, standard protocol for emissions QA is somewhat of a moving target. New
emissions sources are continually being modeled, new tools become available, and new
information comes to light with time. The emissions QA protocol developed by CEP for
the WRAP is a first of its kind effort at documenting the process for evaluating the
performance of an emissions model. The protocol does not directly attempt to evaluate
the data being modeled, rather it confirms that the model is being applied correctly, that
the available data is being implemented correctly, and the results will be reproducible in
the future. The document created under Task 2ABC forms the foundation for a
continually evolving process that will require periodic reviews to account for new
techniques and the progression of the data and modeling. It is our recommendation that
the protocol not only be enforced in future modeling projects but to also have resources




                                            18
devoted to maintaining it as the seminal source of information for managing and
reviewing large emissions modeling projects.
The January, 2004 distribution (version 1.5) of the WRAP emissions QA protocol is
available on the CEP FTP site at:
ftp://ftp.unc.edu/pub/empd/wrap/.

6 Preliminary §308 Emissions Modeling
6.1 Task Description
The purpose of Task 2E in the 2003 Work Plan was to develop and QA emissions for an
interim 2002 inventory for annual regional haze modeling to prepare for §308 SIP /TIP
development. The Work Plan designated that the preliminary modeling would use an
interim 2002 inventory, assimilate the changes to the SMOKE temporal and chemical
profiles based on the reviews from Task 1CD, and implement the QA protocol described
in Section 5 of this document. Task 2E was essentially a “shakeout” simulation that
allowed the RMC to get set up for §308 modeling and to test the implementation of the
new QA protocol. We refer to Task 2E as a shakeout simulation because it provided
initial model results for confirming the setup of the model and the data, testing the
implementation of the QA protocol, and for shaking out any problems associated with
configuring a new, data intensive SMOKE simulation. The RMC emissions modeling
team that worked on the preliminary 2002 modeling consisted of modelers at CEP and
UCR. We completed a six month emissions simulation for summer/winter, 2002 on the
RMC computing cluster in January, 2004. Table 6-1 displays the SMOKE modeling
domain and the domain configuration used for the preliminary 2002 simulation. We will
refer to the preliminary 2002 simulation as Pre02a_36, the designated ID tag in the
WRAP modeling protocol.

                Table 6-1. Pre02a_36 SMOKE domain configuration




                                           19
              Grid Specification                             Value
        Columns                            148
        Rows                               112
        Layers                             19
        X-origin                           -2,736,000 m
        Y-origin                           -2,088,000 m
        X-center                           -97 degrees
        Y-center                           40 degrees
        X-cell                             36,000 m
        Y-cell                             36,000 m
        Vertical Top                       10,000 m
        Vertical Layers ()                1.0, 0.995, 0.99, 0.985, 0.98, 0.97, 0.96,
                                           0.95, 0.94, 0.92, 0.9, 0.88, 0.86, 0.82,
                                           0.77, 0.7, 0.6, 0.45, 0.25




The emissions modeling that we performed for Task 2E consisted of the following
components:
        Receive emissions inventory (EI) from WRAP EI contractors.
        Ensure all regions and inventories are covered (but not duplicated) by filling in
         gaps with base year or projected emissions from the 1996 and/or 1999 National
         Emission Inventory (NEI).
        Set up and run SMOKE for a trial period (e.g., one month)
        Perform QA on the trial period results
        Run SMOKE for the semi-annual simulation
        Perform QA on the semi-annual period results
        Deliver emissions to the WRAP RMC for use in air quality model
We address these components as they were applied to the Pre02a_36 simulation using the
following format. In Section 6.2 we describe the SMOKE input data used for this Task,
including both inventory files and SMOKE ancillary files such as cross-reference files,
temporal profiles, speciation profiles, and spatial surrogates. In Section 6.3, we describe
the overall processing strategy used in SMOKE for this project and describe the SMOKE
run scripts. In Section 6.4, we describe the processing approach and emissions results for
stationary area sources, non-road mobile sources, on-road mobile sources, road dust
sources, point sources, fire sources, and biogenic emission sources. In Section 6.5, we
describe the lessons learned from this Task and make recommendations about what can


                                             20
be done in the next round of §308 modeling to improve the data collection, modeling and
QA efforts. In Section 6.6, we provide summaries and descriptions that illustrate the
application of the QA protocol developed in 2003 to the emissions modeling for this task.
This section will conclude in Section 6.7 with a discussion about what the RMC
emissions modeling team is expecting to complete during the next phase of the 2002
modeling and how the QA protocol will be improved for the next round of modeling.

6.2 SMOKE Input Files
The WRAP tagged the emissions inventory data used in the Pre02a_36 modeling as
interim because it was the result of the initial data collection effort and did not contain all
of the inventory files that will go into the final §308 modeling. The emissions inventory
contractors to the WRAP partially delivered the inventory data for the 2002 modeling
before the deadline for locking down the SMOKE input data passed in early November,
2003. The lockdown occurred to allow sufficient time to complete the SMOKE and
CMAQ modeling and analysis before a scheduled WRAP technical meeting in Tempe,
AZ starting on January 26, 2004. CEP supplemented the missing inventory data in the
interim WRAP inventories with 1996 National Emissions Inventory (NEI96) data grown
to 2002 with SMOKE and base 1999 National Emissions Inventory (NEI99) files. We
collected the SMOKE ancillary input files, such as the temporal allocation files and
spatial surrogates from multiple sources. The details of all the SMOKE input files used
in the Pre02a_36 modeling are described below.

6.2.1 Inventory and land use data
Table A-1 in Appendix A lists details about the inventory files used in the Pre02a_36
modeling. The information contained in Table A-1 includes exact emissions inventory
file names, the source agency of the files with the delivery date to the RMC, the number
of records in the files, the spatial and temporal coverage of the files, and the pollutants
contained in the files. The inventory files used in the Pre02a_36 modeling are a
combination of files created by WRAP inventory contractors, the NEI99, and the NEI96
grown to 2002 with SMOKE. The files prepared by the WRAP contractors are actual
2002 or 2003 inventories for the WRAP region, while the inventories for the non-WRAP
states were collected and prepared by the RMC emissions modeling team to represent
2002 emissions in other parts of the modeling domain. Table 6-2 lists the inventory data
sources for the RPO’s, Mexico, and Canada in the 2002 modeling domain. In the table
“WRAP” refers to an inventory generated by a WRAP contractor, “NEI96+” is the
NEI96 inventory grown to 2002 by the RMC, “NEI99” is the base NEI99 inventory,
“BRAVO” is the 1999 Big Bend Regional Aerosol & Visibility Observations Study
emissions data for Mexico, and “Clear Skies” is the 1995 Canadian national inventory
used for the US EPA Clear Skies modeling project. Table 6-2 is also color coded to
indicate the temporal extent of the inventories with green representing annual inventories,
yellow representing seasonal inventories, and blue representing inventories derived from
daily meteorology information.




                                              21
       Table 6-2. Sources of emissions inventory data for the WRAP modeling domain regions.
Inventory WRAP             CENRAP MWRPO VISTAS MANEVU Mexico                        Canada
Area          WRAP         WRAP        NEI96+     NEI96+     NEI96+      BRAVO ClearSkies
Road          WRAP         WRAP        WRAP       WRAP       WRAP        BRAVO none
Dust
Non Road      WRAP         NEI96+      NEI96+     NEI96+     NEI96+      BRAVO ClearSkies
On Road       WRAP         NEI99       NEI99      NEI99      NEI99       BRAVO ClearSkies
On Road       WRAP         n/a         n/a        n/a        n/a         n/a        n/a
(CA)
Point         WRAP         WRAP        NEI96+     NEI96+     NEI96+      NEI96+     NEI96+
Biogenic      VISTAS VISTAS            VISTAS     VISTAS     VISTAS      VISTAS VISTAS
Ag. Fire      WRAP         NEI96+      NEI96+     NEI96+     NEI96+      none       ClearSkies
        Annual inventory         Seasonal inventory       Daily meteorology dependent emissions


    We anticipate that the 2002 WRAP area source inventory delivered to the RMC in 2003
    will be the final version of that inventory for the §308 modeling. The WRAP inventory
    contains area source data for the WRAP and CENRAP states for 2002. The RMC
    amended the WRAP inventory with the NEI96 grown to 2002 with EGAS 4.0 growth
    factors to cover the rest of the United States. We modeled the area source inventories for
    Mexico and Canada as their base years of 1999 and 1995, respectively. The Big Bend
    Regional Aerosol & Visibility Observations Study (BRAVO) produced the Mexico
    inventory used in Task 2E. The ten northern Mexican states contained in this inventory
    include San Luis Potosi, Baja California Norte, Sonora, Chihuahua, Coahuila de
    Zaragoza, Nuevo Leon, Tamaulipas, Sinaloa, Durango, and Zacatecas The Mexico
    inventory is available from the US EPA at the following URL:
    http://www.epa.gov/ttn/chief/net/mexico.html. The US EPA distributes a SMOKE ready
    version of the 1995 Canadian national inventory as part of the Clear Skies modeling
    project. This inventory is available from the US EPA at the following URL:
    ftp://ftp.epa.gov/modelingcenter/Clear_skies/CSA2003/Emissions. While we have been
    notified that updated inventories will become available, we are uncertain whether the
    amended US inventories derived from the NEI96 or the Mexico and Canada inventories
    will be replaced by updated inventories before the final 2002 modeling begins.
    The WRAP RMC received a 2003 road dust emissions inventory for paved and unpaved
    roads in late 2003 for use in the preliminary §308 modeling. The 2003 road dust
    inventory uses the same methodology to account for the transportable fraction reductions
    as was used in developing the 1996 and 2018 inventories for the WRAP §309 modeling
    performed in 2001-2003 (Pollack, 2004). The WRAP inventory contractor supplied the
    road dust inventory for all states in the CMAQ domain, and it includes PM2.5 and PM10
    emissions only. The BRAVO area source inventory contains road dust PM2.5 and PM10
    emissions for the 10 Mexican states covered in the inventory. We do not currently have a


                                                  22
road dust inventory for Canada. The US and Mexican road dust inventories that we are
currently using will be used for the final §308 modeling. We may supplement these
inventories with a Canadian road dust inventory if one becomes available.
The RMC received 2003 on-road and non-road mobile source inventories covering the
WRAP states for use in the 2002 §308 modeling from the WRAP mobile inventory
contractor in December, 2003. Both are seasonal inventories and contain pre-computed
emissions with speciated PM 2.5. The on-road mobile inventory is split between
California and the rest of the 12 WRAP states. We are supplementing the WRAP on-
road and non-road inventories with annual NEI99 activity data and NEI96 emissions
grown to 2002, respectively. For the non-WRAP on-road emissions we are using the
MOBILE6/SMOKE integration to calculate CO, NOx, VOC, and PM emissions from the
NEI99 activity data. The SO2 and NH3 emissions for on-road mobile sources come from
the NEI99 emissions inventory. For Canada and Mexico we extracted the on-road and
non-road mobile SCC’s from the Clear Skies and BRAVO area source inventories,
respectively, and grouped them with the US mobile source inventories. We expect
updates to both the non-WRAP inventories and the Canadian inventory to occur in the
next year. There is also a possibility that the 2003 WRAP mobile emissions will be
replaced by either 2002 inventories or 2003 emissions generated from more recent
emissions factor inputs.
The emissions factors used to derived the 2003 California on road mobile inventory
received by the RMC contained four vehicle types (light/heavy-duty gasoline/diesel) as
opposed to the more detailed vehicle types contained in the 1996 and 2018 inventories
used in the §309 modeling. Since the production of the interim WRAP 2003 on-road
mobile inventory, the California Air Resources Board (CARB) has released an updated
version of their on road emissions model, EMFAC2002. The WRAP mobile source
inventory contractor made a recommendation with the delivery of the on road mobile
data that a new California inventory be created that is based off the EMFAC2002 results
(Pollack, 2003). A schedule for developing and delivering an updated 2003 on road
mobile inventory for California has not been determined as of the writing of this report,
but the work may take place before the inventories are finalized for the base §308
modeling.
The RMC received the WRAP 2002 point source inventory with the area source
inventory in November, 2003 and anticipates that this will be the final version used for
the §308 modeling. The WRAP inventory contains point source data for the WRAP and
CENRAP states for 2002. The RMC amended the WRAP inventory with the NEI96
grown to 2002 with EGAS 4.0 growth factors to cover the rest of the United States. We
modeled the point source inventories for Mexico and Canada as their base years of 1999
and 1995, respectively. The Big Bend Regional Aerosol & Visibility Observations Study
(BRAVO) produced the Mexico inventory used in Task 2E. The ten northern Mexican
states contained in this inventory include San Luis Potosi, Baja California Norte, Sonora,
Chihuahua, Coahuila de Zaragoza, Nuevo Leon, Tamaulipas, Sinaloa, Durango, and
Zacatecas The Mexico inventory is available from the US EPA at the following URL:
http://www.epa.gov/ttn/chief/net/mexico.html. The US EPA distributes a SMOKE ready
version of the 1995 Canadian national inventory as part of the Clear Skies modeling
project. This inventory is available from the US EPA at the following URL:


                                            23
ftp://ftp.epa.gov/modelingcenter/Clear_skies/CSA2003/Emissions. While we have been
notified that updated inventories will become available, we are uncertain whether the
amended US inventories derived from the NEI96 or the Mexico and Canada inventories
will be replaced by updated inventories before the final 2002 modeling begins.
While most of the inventory data came either directly from WRAP inventory contractors
or from the EPA through CEP, the WRAP RMC benefited from the unified 36-km RPO
domain by using gridded biogenic land use files developed by Alpine Geophysics for the
Visibility Improvement State and Tribal Association of the Southeast (VISTAS)
modeling. The VISTAS gridded land use data came in the form of binary input files for
running BEIS3 in SMOKE. We anticipate using the VISTAS gridded biogenic land use
data for the final §308 modeling.
Fire inventory contractors developed daily point source fire inventories with hourly
plume rise information for the WRAP states for the §309 modeling (Air Sciences, 2002).
Based on recommendations from the WRAP fire forum, the RMC used the 2018 base
smoke management agricultural fire inventory to represent fires in the Pre02a_36
modeling. The decision to not use any of the wild or prescribed fire inventories
developed for the §309 modeling was made because these types of fires affect primarily
wild or forested areas that, once burned, will not burn again for several years. As the
wild and prescribed fire inventories currently available to the WRAP RMC are based off
of 1996 fire data, the fire forum deemed that the same land area that burned in 1996
would not burn again in 2002. Agricultural fires, on the other hand occur on the same
fields every year, and it is therefore justifiable to include an inventory based off of 1996
activity in the 2002 modeling. For the non-WRAP states, we used the NEI96 area source
inventory grown to 2002 to include all three fire categories in the modeling. The Canada
Clear Skies area source inventory that we used for the Pre02a_36 modeling contains wild
and agricultural fires. The Mexico BRAVO area source inventory that we used for the
Pre02a_36 modeling contains wild and prescribed fires. The RMC anticipates receiving
draft and final, actual 2002 wild and prescribed fire inventories to use for the final §308
modeling in early summer, 2004.
Table 6-3 presents emissions pie charts from the Pre02a_36 modeling for representative
weekdays in the winter and summer. These charts display the relative contribution of the
different source categories to each of the inventory pollutants. Appendix B contains bar
charts by pollutant and source category summarizing the annual inventories for all states
in the 2002 interim emissions modeling domain.




                                            24
Table 6-3. Break down of 2002 interim emissions by source category (US emissions
                                     only)

                            Area       Ag Fire      Non-Road       Point
                            Biogenic   Road Dust    On Road

              Monday January 7, 2002                        Monday July 8, 2002
                  18%
                                                            18%                   17%
                                             35%
  CO




                                                             65%
                      47%

                                       17%
                                                            49%
                                                                                   8%
                                             1%

           46%                                11%                                       8%
  NOx




                                                                                    10%

                                                                              25%
                                       25%

                      11%
                                                                                  1%
                 9%                                                                 2%
                                              37%                                    3%
            6%
  VOC




                                                                                        7%


                                                      87%



                       37%




                                              25
                  Area         Ag Fire            Non-Road         Point
                  Biogenic     Road Dust          On Road

            Monday January 7, 2002                         Monday July 8, 2002



                                     6.7%                                              4%

                                        0.3%                                           4%
NH3




        85.6%                           7.4%
                                                     92%




                                         8%

                                          1%                                            6%
SO2




                                           3%      89%
                                                                                        1%
         88%                                                                            4%




                                                                                 12%
                                     0.2%
                                        2.5%
         61.9%
                                                    53%
                                                                                        11%
PM10




                                          12.3%




                               23.0%                                             24%

                              1%                      49.4%
                                   6%                                            0.7%
                                                                                   0.1%

                                                                                        12.9%
PM2.5




         63%                              19%




                                   11%                     14.8%                 22.1%




                                        26
We anticipate supplementing as many of the NEI99 and NEI96 inventories with actual
2002 emissions inventories as the data becomes available through the other RPO’s.
Table A-4 extends Table A-1 by listing placeholders for inventory files that the WRAP
RMC anticipates receiving before the final round of 2002 modeling begins.Error!
Reference source not found.Error! Reference source not found.

6.2.2 SMOKE ancillary inputs
Tables A-2 and A-3 in Appendix A list details about the ASCII and binary SMOKE
ancillary input files used in the Pre02a_36 modeling, respectively. The RMC created the
2002 meteorology files with MM5 and preprocessed the data with MCIP to prepare them
for SMOKE. All of the ancillary emissions input files except the temporal profile/cross-
reference files and the spatial surrogates/cross-reference files originated from the
SMOKE version 2 distribution. We used the temporal profile and cross reference files
that the RMC used for the WRAP §309 modeling (Houyoux, 2003). Similar to the
gridded biogenic land use files, we used spatial surrogates developed by Alpine
Geophysics for VISTAS modeling. Contact Greg Stella at Alpine Geophysics for
additional information about the preparation of the surrogates
(gms@alpinegeophysics.com). The RMC anticipates replacing the surrogates created by
the VISTAS contractor with unified surrogates provided by the US EPA
(http://www.epa.gov/ttn/chief/emch/spatial/newsurrogate.html). The reasons behind
using the EPA surrogates are twofold. First, the EPA surrogates use more land use
categories and thus provide for higher resolution spatial allocation information. Second,
the EPA provides surrogates for three different spatial resolutions, 36-km, 12-km, and 4-
km. By starting with the 36-km unified surrogate, we can use the finer resolution data in
subsequent modeling simulations that will be on 12-km and 4-km grids within the WRAP
region.

6.3 SMOKE Modeling Details
The RMC installed SMOKEv2 on their computers in October, 2003 to use for the
Pre02a_36 modeling. The RMC installation used the precompiled SMOKE executables
for Red Hat Linux distributed from the Community Modeling and Analysis System
(CMAS) website. For a complete background on the formulation of SMOKE check the
online SMOKE manual:
http://www.cep.unc.edu/empd/products/smoke/SMOKEDOCS.shtml#manual.
Modeling for 2002 began as soon as SMOKEv2 was available on the RMC machines as
some of the emissions inventory files, like the typical year agricultural fire sources and
the NEI99 inventory for on road mobile sources were already available. As more
inventories became available they were integrated into the processing schedule. For part
of the initial setup for the Pre02a_36 modeling CEP made several changes to the
SMOKEv2 scripts to make them consistent with the QA protocol.
The QA protocol lists organization as one of the key tenets of emissions modeling. CEP
modified the SMOKEv2 scripts to break up the inventory, output file, log file, and report
directories by source category to better organize the contents of these directories. CEP
also reconfigured the SMOKEv2 scripts to be more modular so that all important
scripting processes were contained in smaller, discreet scripts that are linked together by


                                            27
a master script. This modular redesign contrasts to the original method of having
multiple functions in one large script. Smaller, targeted scripts are easier to manage and
troubleshoot and allow for greater redundancy in the modeling process.
The rest of Section 6.3 describes the scripting conventions and model configuration used
for the Pre02a_36 modeling and describes the simulation in detail.

6.3.1 SMOKE Scripts
The scripts are the interface that emissions modelers use to run SMOKE, and are
therefore the items of practical importance for anyone wanting to simply reproduce the
work performed as part of this contract. For this project, we created many SMOKE
scripts to run the required emissions modeling cases, which we describe in this
subsection. In the §309 modeling we modified the SMOKE scripting to use an
operational modeling approach that used the number of times that each SMOKE program
gets run in the course of a simulation to configure the scripts (Houyoux, 2002). For the
§308 modeling we decided to stay with the default SMOKE script set up, which is based
on source categories, to configure the scripts. We did make several modifications to the
default SMOKE scripts to modularize them, add error checking loops, and break up the
report and logs directories by source category. Now we have one script for each source
category that we are modeling that calls all of the SMOKE programs required for
simulating that source. Table 6-4 lists all of the SMOKE scripts used for the Pre02a_36
modeling and the SMOKE programs called by each script. Area, point, and non-road
mobile sources list two sets of scripts for the Pre02a_36 modeling. The scripts tagged
“wrap2002” create the 2002 emissions; the scripts tagged “nei96” import and grow the
NEI96 inventory to 2002. We combined the resulting “2002-like” inventories created by
the “nei96” scripts with the WRAP 2002 inventories to create the 2002 emissions. In
addition to the source-specific scripts listed in Table 6-4, we also listed the SMOKE
utility scripts that actually call executables, manage the log files, and manage the
configuration of the SMOKE simulations.

                        Table 6-4. Summary of SMOKE scripts
Source Category      Script Name                      SMOKE Programs/Function
Area                 smk_ar_wrap2002.csh              Smkinven, Grdmat, Spcmat, Temporal,
                                                      Smkmerge, Smkreport
                     smk_ar_nei96_import.csh          Smkinven, Cntlmat, Smkreport
                     smk_ar_nei96_growth.csh
Non-road mobile      smk_nr_wrap2002.csh              Smkinven, Grdmat, Spcmat,Temporal,
                                                      Smkmerge, Smkreport
                     smk_nr_nei96_import.csh          Smkinven,Cntlmat,Smkreport
                     smk_nr_nei96_growth.csh
Road Dust            smk_rd_wrap2002.csh              Smkinven, Grdmat, Spcmat, Temporal,
                                                      Smkmerge, Smkreport



                                            28
Source Category      Script Name                    SMOKE Programs/Function
On-road mobile       smk_mb_wrap2002.csh            Smkinven, Grdmat, Spcmat, Temporal,
(non-VMT)                                           Smkmerge, Smkreport
On-road mobile       smk_mb_wrap2002.csh_vmt        Smkinven, Premobl, Emisfac, Grdmat,
(VMT-based)                                         Spcmat, Temporal, Smkmerge, Smkreport
Point                smk_pt_wrap2002.csh            Smkinven, Grdmat, Spcmat, Elevpoint,
                                                    Laypoint, Temporal, Smkmerge, Smkreport

                     smk_pt_nei96_import.csh        Smkinven,Cntlmat,Smkreport
                     smk_nr_nei96_growth.csh
Ag Fire              smk_agfire_wrap2002.csh        Smkinven, Grdmat, Spcmat, Elevpoint,
                                                    Laypoint, Temporal, Smkmerge, Smkreport
Biogenic             smk_bg_wrap2002_beis3.csh Normbeis, Tmpbeis, Smkmerge
Merge                smk_mrgall_wrap2002.csh        Mrggrid, Smkreport
n/a                  check_smoke_out.csh            confirms the existence of SMOKE output
                                                    files following a simulation
n/a                  cntl_run.csh                   calls the SMOKE executables for running
                                                    the projection and control programs
n/a                  make_invdir.csh                creates inventory output directories by
                                                    source category
n/a                  movelog.csh                    moves existing log file to a back up file
n/a                  smk_run.csh                    calls the SMOKE executables for
                                                    everything but projection, controls, and
                                                    QA
n/a                  qa_run.csh                     calls the SMOKE executables for running
                                                    the QA programs
n/a                  smoke_calls.csh                calls smk_run.csh, cntl_run.csh and
                                                    performs configuration management; used
                                                    to clean up the source-specific run scripts
n/a                  smoke_calls_reports.csh        calls qa_run.csh; used to clean up the
                                                    source-specific run scripts

6.3.2 Configuration
The RMC emissions modeling team configured SMOKE to generate emissions for the
winter and summer months of 2002 on the 36 km unified RPO modeling domain. A full
listing of the SMOKE configuration options for this simulation is listed in Appendix C.
For the anthropogenic emissions sources that use hourly meteorology (on road mobile
and agricultural fires) we configured SMOKE to represent the daily emissions explicitly.
For the non-meteorology dependent emissions, we used a representative Saturday,


                                           29
Sunday, Monday, and weekday for each month as surrogate days for the entire month’s
emissions (we refer to this as the MWDSS processing approach). For these non-
meteorology dependent emissions sources we explicitly represented the holidays as
Sundays. Table 6-5 lists the days that we modeled as representative days in the months
that we simulated for the Pre02a_36 modeling. Table 6-6 lists the holidays in 2002 that
will be modeled.

           Table 6-5. Representative model days for Pre02a_36 simulation
         Saturday           Sunday           Monday            Weekday
         January 5          January 6        January 7         January 4
         February 2         February 3       February 4        February 5
         December 7         December 8       December 9        December 10
         June 8             June 9           June 3            June 4
         July 6             July 7           July 8            July 3
         August 3           August 4         August 5          August 6




                           Table 6-6. 2002 modeled holidays
                     Holiday                 Date
                     New Years               January 1, 2002
                     Good Friday             March 29, 2002
                     Memorial Day            May 27, 2002
                     Independence Day        July 4, 2002
                     Labor Day               September 2, 2002
                     Thanksgiving Holiday    November 28-29, 2002
                     Christmas Holiday       December 24-25, 2002


We used the designations in Table 6-7 to determine which months fell into each season
when temporally allocating the seasonal emissions inventories.




                                            30
               Table 6-7. Assignments of months to seasons for use of
                        seasonal inventory files in SMOKE

                                Month           Season
                            January            Winter
                            February           Winter
                            March              Spring
                            April              Spring
                            May                Spring
                            June               Summer
                            July               Summer
                            August             Summer
                            September          Fall
                            October            Fall
                            November           Fall
                            December           Winter

The WRAP RMC used 36-km spatial surrogates developed at Alpine Geophysics for
VISTAS modeling. The surrogate codes and descriptions are listed in Table 6-8.




               Table 6-8. 36 km spatial surrogates with descriptions
                 Surrogate ID    Description
                 1               Population
                 2               Housing
                 3               Area
                 4               Water Area
                 5               Non-Agricultural
                 6               Dryland Cropland and Pasture
                 7               Irrigated Cropland and Pasture
                 8               Cropland/grassland mosaic
                 9               Cropland/woodland mosaic
                 10              Airports
                 11              Ports
                 12              Railroads
                 13              Urban Area
                 14              Rural Area



                                         31
                  Surrogate ID     Description
                  15               Urban Primary Roads
                  16               Rural Primary Roads
                  17               Urban Secondary Roads
                  18               Rural Secondary Roads
                  19               Forest Area
                  20               Non-Forest
                  21               Sum of All Agricultural
                  22               Sum of All Roadways
                  30               Interstate highways

6.3.3 Simulation description
RMC performed the Pre02a_36 modeling simulation to test the configuration of SMOKE
and to evaluate the quality of the inventory and ancillary input data. Testing the
implementation of the new WRAP emissions QA protocol was another objective of the
preliminary modeling. The ultimate goal of the Pre02a_36 modeling was to prepare for
the final 2002 emissions simulation scheduled for late Spring, 2004. To perform these
tests we modeled a base inventory for the Winter and Summer months of 2002. While no
major stumbling blocks presented themselves during the preliminary modeling, we did
encounter a few difficulties in implementing the QA protocol. Details about the
implementation of the emissions QA protocol are contained in Section 6.6.

6.4 SMOKE processing for WRAP source categories
SMOKE has four different processing approaches: processing for area, on-road mobile,
point, and biogenic source. The WRAP source categories map to these processing
approaches as shown in Table 6-9.

 Table 6-9. WRAP source categories mapped to processing approaches in SMOKE

              WRAP Source category                       SMOKE processing approach
                  Stationary area                                 Area
                  Nonroad mobile                                  Area
                     Road dust                                    Area
                  On-road mobile                             On-road mobile
                        Point                                     Point
               Fire (agricultural fire)                           Point
                     Biogenics                                  Biogenics

Each SMOKE processing approach is for a specific SMOKE source category, which has
its own source characteristics. These correspond to the identifiers used in creating the
emission inventory (e.g. state/county FIPS code and SCC). The source categories also


                                           32
have source attributes, which are the other useful data in the emission inventories that
SMOKE uses for processing them (e.g., point-source flue gas exit height and
temperature). Source characteristics define the sources as area, mobile, or point sources
for SMOKE processing and also distinguish one source in the inventory from another.
Source attributes are additional data about the source that do not contribute to the
source’s uniqueness in SMOKE.
In SMOKE, each source category is defined by source characteristics, as follows:
      Area sources are defined by (1) country, state, and county codes, (2) SCC codes,
       or optionally (3) grid cell only.
      On-road mobile sources are defined by (1) country, state, and county codes, (2)
       SCC codes, and optionally (3) link codes.
      Point sources are defined by (1) country, state, and county codes, (2) plant codes,
       and (3) characteristics 1 through 5, one of which should be the SCC code.
      Biogenic sources are defined differently depending on the type of processing that
       you are using. They can be defined either by (1) country, state, and county codes
       and (2) land use code, or by (1) grid cell and (2) land use code.

In the following subsections, we describe the processing that we performed for each of
the WRAP source categories.

6.4.1 Area emissions processing
Area-source processing includes emissions for the stationary area inventory (including
Mexican and Canadian area, nonroad mobile, on-road mobile, and road dust), the U.S.
road dust inventory, and the U.S. nonroad mobile inventory. For the Pre02a_36
emissions modeling we grouped the area inventories by source category and applied the
MWDSS temporal allocation approach to distribute the emissions in each month. We
used the following grouping for the area inventory processing,
Group 1: Stationary Area
      WRAP 2002 area inventory: WRAP and CENRAP states, annual inventory
      NEI96_02 area inventory: MWRPO, VISTAS, and MANE-VU states, annual
       inventory, NEI96 grown to 2002 with EGAS4.0 growth factors
      BRAVO Mexico 1999 area inventory: 10 Northern Mexican states, annual
       inventory
      Clear Skies Canada 1995 area inventory: All Canadian provinces, annual
       inventory
Group 2: Non-road Mobile
      WRAP 2003 non-road inventory: WRAP states, seasonal inventory
      NEI96_02 non-road inventory: CENRAP, MWRPO, VISTAS, and MANE-VU
       states, annual inventory, NEI96 grown to 2002 with EGAS4.0 growth factors




                                            33
      BRAVO Mexico 1999 non-road inventory: 10 Northern Mexican states, annual
       inventory extracted from BRAVO area source inventory
      Clear Skies Canada 1995 non-road inventory: All Canadian provinces, annual
       inventory extracted from BRAVO area source inventory
Group 3: Road Dust
      WRAP 2002 paved road dust inventory: Entire US, seasonal inventory
      WRAP 2002 unpaved road dust inventory: Entire US, seasonal inventory
      BRAVO Mexico 1999 paved road dust inventory: 10 Northern Mexican states,
       annual inventory extracted from BRAVO area source inventory
      BRAVO Mexico 1999 unpaved road dust inventory: 10 Northern Mexican
       states, annual inventory extracted from BRAVO area source inventory
For a flow diagram on how the inventories are processed through SMOKE refer to the
WRAP Jumpstart Final Report (Houyoux, 2002).

6.4.2 On-road mobile-source emissions processing
The SMOKE processing for WRAP included two approaches for processing on-road
mobile sources. The first approach was to compute mobile emissions values prior to
providing them to SMOKE; we call this the precomputed-emissions approach. The
second approach was to provide SMOKE with VMT data, meteorology data, and
MOBILE6 inputs, and let SMOKE compute the mobile emissions based on these data.
We call this the VMT approach. These approaches are not mutually exclusive for a single
SMOKE run; therefore, we performed single SMOKE runs in which both approaches
were used.
As described in Section 6.2.1, the on-road mobile inventory contains a combination of
inventories:
      Precomputed, seasonal MOBILE6-based emissions of all pollutants for the 13
       WRAP states that included prespeciated PM2.5 data.
      Annual VMT for computing CO, NOx, VOC, and PM using MOBILE6 for the
       rest of the United States.
      Precomputed, annual 1999 NEI emissions for SO2 and NH3 for the rest of the
       United States.
      Precomputed, annual 1999 BRAVO emissions of all pollutants for Mexico
      Precomputed, annual 1995 Clear Sky emissions of all pollutants for Canada
Consequently, we used both the precomputed-emissions and VMT approaches for this
effort, with both types of processing being performed in the same program runs (and
same scripts). In SMOKE, the precomputed emissions approach is quite similar to
processing for area sources. All SMOKE on-road mobile runs were performed as a single
group using the following inventory set-up,




                                           34
      WRAP 2003 on road mobile inventory: 2003 WRAP states (without
       California) precomputed, prespeciated PM2.5 seasonal emissions
      CA 2003 on road mobile inventory: 2003 California precomputed, prespeciated
       PM2.5 seasonal emissions
      NEI99 on-road VMT: 1999 CENRAP, MWRPO, VISTAS, and MANE-VU
       states, annual NEI activities for input to MOBILE6
      NEI99 on-road inventory: 1999 CENRAP, MWRPO, VISTAS, and MANE-VU
       states, annual NEI precomputed SO2 and NH3 emissions
      BRAVO Mexico 1999 on road mobile inventory: 10 Northern Mexican states,
       annual inventory extracted from BRAVO area source inventory
      Clear Skies Canada 1995 on road mobile inventory: All Canadian provinces,
       annual inventory

6.4.3 Point-source emissions processing
Point-source processing for the WRAP Pre02a_36 modeling includes emissions for the
point-source inventory and agricultural (ag) burning inventory. The point-source
inventory included both Mexican and Canadian point sources. In SMOKE, one of the
major factors distinguishing stationary-point source inventory processing from the other
source categories is the calculation by SMOKE of plume rise and subsequent allocation
of the emissions vertically. In addition, point-source processing can include day-specific
and hour-specific emission inventories and hour-specific precomputed plume rise.
We grouped the point source emissions by stationary point and fire sources using the
following set up,
Group 1: Stationary Point
      WRAP 2002 point inventory: WRAP and CENRAP states, annual inventory
      NEI96_02 point inventory: MWRPO, VISTAS, and MANE-VU states, annual
       inventory, NEI96 grown to 2002 with EGAS4.0 growth factors
      BRAVO Mexico 1999 point inventory: 10 Northern Mexican states, annual
       inventory
      Clear Skies Canada 1995 point inventory: All Canadian provinces, annual
       inventory; version 1 of this inventory of which there have been two subsequent
       versions.
Group 2: Fire Sources
      WRAP 2018 Base Smoke Management Ag Fires: WRAP states, daily
       inventory with hourly plume rise information

6.4.4 Biogenic emissions processing
Biogenic inventories depend heavily on the meteorological conditions of each day and
hour. Temperature and solar radiation data are components of the biogenic inventory
calculation. For this modeling effort, we used the Biogenic Emission Inventory System,


                                            35
version 3.09 (BEIS3) within SMOKE and the prognostic meteorological data provided by
MCIP for input to the CMAQ. For the meteorology variables used to calculate the
biogenic emissions, we used ambient temperatures (TA) and radiation reaching the
ground (RGRND). Since the inventory is really a modeling exercise, the inventory is
computed by the SMOKE system during emissions processing, and not as a separate
inventory preparation process. The main preparation activity is creating the gridded land
use data, which in this case was done by the VISTAS inventory contractor and supplied
to the WRAP. Alpine Geophysics, LLC prepared the 36 km gridded landuse data that the
RMC used for the Pre02a_36 modeling. Contact Greg Stella at Alpine Geophysics for
additional information (gms@alpinegeophysics.com).

6.5 Problems encountered and remaining processing issues
Lessons learned during the WRAP Jumpstart project resulted in an overhaul to the RMC
emissions modeling protocol. Many of the issues that we faced during the Jumpstart
period, such as a multitude of data sources, errors in the data, delays in the data delivery
and limited resources for QA still presented themselves, but we took these issues into
account when developing the modeling protocol for the §308 modeling. The Pre02a_36
modeling was intended to test the new QA system and did present several issues that we
caught through following the draft protocol. We anticipated that we would encounter
problems during the roll out of the modeling protocol and also run into unforeseen QA
issues that we didn’t account for in the protocol. In Section 6.6 we discuss the
implementation of the QA protocol, problems that we had with rolling it out, and how we
plan on improving it for the final 2002 modeling. In this section, we describe the QA
issues that we encountered in the Pre02a_36 modeling and how they were resolved. In
detailing these problems we distinguish between inventory problems, ancillary input file
problems, and problems encountered with the SMOKE configuration. We also address
unresolved issues with the emissions modeling. Processing errors that we caught through
the procedures in the draft QA protocol are summarized in Appendix D.

6.5.1 Emissions inventory issues
Most of the inventory problems encountered in the Pre02a_36 modeling resulted from
extracting state information from the NEI-derived files for the sections of the domain not
covered by the WRAP inventories. Other problems related to redundant information in
the on-road mobile input files and an incorrect assumption used in developing the
agricultural burning fire inventory. For unknown reason we lost some columns of data
from the inventory files when extracting the non-WRAP states from the NEI total US
area source inventories. As a result of this extraction error, we lost the SO2 and NH3
inventory estimates for the non-WRAP area and non-road mobile inventories (non-road
sources are contained in the NEI area source emissions inventory). We discovered this
error through spatial visualizations of the domain totals for the affected sources. After
redoing the extraction and running the corrected inventories through SMOKE this
problem was resolved. In making the switch from MOBILE5b, used for the §309
modeling, to MOBILE6 we erroneously double counted particulate matter (PM)
emissions because we supplemented the MOBILE6 emissions with precomputed PM
from the NEI.



                                            36
A new feature of MOBILE6 not contained in MOBILE5b is the capability for the model
to calculate PM emissions from on road mobile source activities. MOBILE6 calculates
CO, NOx, VOC, and PM emissions, where MOBILE5b only generated CO, NOx, and
VOC. It’s necessary to supplement the output from MOBILE6 with pre computed NH3
and SO2 emissions from the NEI. When running MOBILE5b it was also necessary to
supplement the activity-derived emissions with pre computed PM. Initially we included
pre computed PM as a supplement to the MOBILE6 output. After discovering this error
we removed PM from the supplemental mobile inventory for the WRAP states and
reprocessed the inventory. The EPA brought another error in the default MOBILE6 files
to our attention during the Pre02a_36 modeling.
MOBILE6 is distributed with a process file called MEPROC that defines the on road
mobile source activities that produce emissions. The MEPROC file in the SMOKEv2
distribution defined refueling as a process for which MOBILE6 would compute
emissions. Madeline Strum of the U.S. EPA notified CEP on January 20, 2004 that the
refueling processes needs to be removed from MOBILE6 as it is already covered in the
non-road mobile emissions inventory (Seppanen, 2004). By including refueling in the
MEPROC file, we double counted emissions from this process; once in MOBILE6 and
once in the non-road mobile inventory. This issue affected only the non-WRAP states in
the modeling domain. To fix this issue we reran MOBILE6 after removing the refueling
emissions from the MEPROC file.
We learned after the Pre02a_36 modeling was completed that the WRAP agricultural
burning fire inventory assumed a layer 1 depth of 80 m, where the actual height of layer 1
in our modeling is about 38 m. As the WRAP fire inventories use pre-computed plume
rise and include the fraction of the fires emitted into layer 1 as a variable, overestimating
the height of layer 1 leads to improper vertical distribution of the emissions. The
magnitude of the emissions effects resulting from this error is undetermined but we
speculate that we are placing too large a fraction of the emissions in layer 1 thus
overestimating the local effects from the fires and underestimating the transport. This
error with the agricultural burning inventory is unresolved as of the writing of this report.

6.5.2 SMOKE ancillary input issues
Some of the emissions modeling issues that we encountered during the Pre02a_36
simulation resulted from problems with the temporal profiles, spatial surrogates,
speciation profiles and biogenic input files. In comparing seasonal trends between
different regions of the modeling domain we noticed very large hour to hour variability in
CO emissions from area sources in the Southeast US and in parts of Canada. Time series
of the emissions modeling results in these regions of the domain showed very sporadic
emissions patterns which contrast with the actual temporal profiles that tend to be
smoother from hour to hour. Upon investigation of the SMOKE temporal profiles we
discovered a formatting problem with the diurnal profiles for fire sources that affected the
entire modeling domain. It was more pronounced in the Southeast U.S. and Canada
because the fires in these regions are treated as area sources. Treating fires as area
sources as opposed to point sources is significant because area emissions are spread over
larger portions of the domain (by county). By contrast, the WRAP fire inventories are
treated as point sources and are located in the specific grid cell where the fire occurs.



                                             37
The erroneous diurnal patterns were easier to spot in the parts of the domain where the
fires were treated as area sources because on the spatial plots of the emissions results
these patterns occurred in multiple grid cells as opposed to single cells scattered
throughout the domain. Another reason why the erroneous temporal profiles were
pronounced in some regions of the domain more than others is because of regional
agricultural practices. One of the affected fire sources was agricultural field burning. In
those regions of the domain where agricultural field burning is treated as an area source,
the diurnal variability in the emissions is very apparent because of the way that the
emissions are spread out over several grid cells. We reformatted the temporal profiles for
the fire sources and regenerated the area and fire emissions to correct this problem. The
recreated emissions were more consistent with the designated diurnal patterns for fire
sources.
A second error in the temporal profiles caused the non-road mobile emissions in the
WRAP states to have flat diurnal profiles relative to more pronounced diurnal variability
in the rest of the domain. Investigation of this error revealed other problems with the
speciation profiles and gridding surrogates. The reason for this error was an
inconsistency between the WRAP non-road mobile inventory and the temporal, chemical,
and gridding cross-reference files. The non-road mobile inventory contractor created
new source classification codes (SCC’s) to represent lumped vehicle classes. A series of
new SCC’s specific to the WRAP inventory represented the combination of emissions
from 2-stroke and 4-stroke gasoline engines, liquid petroleum gas (LPG), and
compressed natural gas (CNG) engines. Without specific entries in the temporal cross-
reference files, these lumped SCC’s received the default temporal profiles, which are flat
for the diurnal, weekly, and monthly profiles. These new SCC’s also did not have entries
in the speciation and gridding cross reference files, forcing the application of the default
profiles and surrogates, respectively. We discovered the problem because animations of
the emissions results showed diurnal variability in the non-WRAP states and relatively
constant emissions in the WRAP states. To fix the problem, we created a mapping of the
new SCC’s to the profiles used by the component SCC’s that make up the new sources
(i.e. 2-stroke gasoline, 4-stroke gasoline, LPG, and CNG). Table 6-10 lists the new
SCC’s and the temporal, speciation, and spatial surrogates assigned to them.

                Table 6-10. New WRAP nonroad SCC profile mapping
                     Temporal                          Speciation
      SCC                                                                 Surrogate
                     Monthly      Weekly     Daily      VOC      PM2.5
      2200005000 262              7          26        1186      35700    6
      2200008005 262              7          26        1186      35700    10
      2200006000 262              7          26        1186      35700    1
      2200002000 262              7          26        1186      35700    1
      2200003000 262              7          26        1186      35700    1
      2200004000 262              7          26        1186      35700    2



                                            38
                     Temporal                          Speciation
      SCC                                                                 Surrogate
                     Monthly      Weekly     Daily      VOC      PM2.5
      2200007000 262              7          26        1186      35700    14
      2200010000 262              7          26        1186      35700    1
      2200001000 262              7          26        1186      35700    1
      2200009000 262              7          26        1186      35700    1
      2200008000 262              7          26        1186      35700    10


The last problem with the SMOKE inputs that we encountered was with an offset in the
grid in the gridded land use files used for deriving biogenic emissions. The contractor
who created these files actually discovered the error which caused the biogenic emissions
to be offset one cell to the South and West in the domain. Recalculating the land use
after fixing the definition of the domain origin fixed the problem. A new set of gridded
land use files were received by the RMC on March 26, 2004 and immediately integrated
into the modeling. The Pre02a_36 emissions referred to here, however, used the
erroneous land use files.

6.5.3 SMOKE configuration issues
A few configuration problems with SMOKE affected the point and biogenic emissions in
the Pre02a_36 simulation. The issue affecting the point sources is not as much a problem
as it is a configuration option. The convention for processing elevated point source
emissions is to constrain the vertical extent of the emissions to the mixed layer, or around
2500-3000 m. The primary reason for limiting the number of layers that point sources
emit into is that the file size and run time savings gained by using fewer emissions layers
far outweigh the small differences in the distribution of the emissions at the top of the
model. In the subsequent round of modeling we plan on limiting the number of vertical
layers for the point source emissions from 19 to 15, roughly corresponding to the top of
the mixed layer. While these top 4 layers of the model contain no more than 1% of the
total emissions, by limiting the number of emissions layers we gain a 25% speed up in
processing time and reduce our intermediate and output file sizes by about 25%.
SMOKE allocates the emissions in layers 16-19 into layer 15 when the vertical
configuration is changed. Another configuration issues that we addressed in the
Pre02a_36 modeling affects the biogenic emissions.
SMOKE uses a supplemental freeze date file to flag when the last freeze date in the
spring and the first freeze date in the fall occur throughout the modeling domain. These
freeze dates are used by SMOKE to determine when to switch back and forth between
winter and summer biogenic emissions factors. Without the freeze dates file SMOKE
defaults to using summer emissions factors through the entire year. The initial Pre02a_36
simulation did not use the freeze dates file and thus used summer biogenic emissions
factors through the winter months in the simulation. We created the freeze dates file
using the SMOKE utility METSCAN and implemented the freeze dates switch in the
final Pre02a_36 simulation.


                                            39
6.5.4 Unresolved emissions issues
A few issues with the SMOKE emissions remained unresolved at the completion of the
Pre02a_36 simulation. A listing of these issues follows.
          Applying US holidays to Canadian and Mexican emissions – in processing the
           foreign inventories with the US inventories we are applying the US holidays
           to these sources. Canada and Mexico observe different holidays than the US
           and we are misrepresenting the temporal patterns on the days in these parts of
           the domain that are classified as US holidays.
          Friday emissions slightly higher than the rest of the week for sources using
           MWSS temporal processing – in the MWSS approach the emissions for
           Tuesday-Friday should be exactly the same in a given month. For some
           reason the emissions on Fridays are slightly (~1%) higher than the rest of the
           week.
          Canada and US temporal profiles define seasons differently – US seasons are
           defined as Winter = January, February, December; Spring = March, April,
           May, etc. Canadian seasons are defined as Winter = January, February,
           March; Spring = April, May, June, etc. This inconsistency in season
           definitions leads to an offset in the timing of how the seasonal temporal
           patterns are applied between the two countries. Annual time series of the
           emissions show intra-seasonal variability where we expect to season constant
           emissions within each season. This intra-seasonal variability is due to the
           offset in the season definitions between the US and Canada.
          The Canadian and Mexico inventories are missing some pollutants – the
           BRAVO Mexico inventory does not contain NH3 emissions from point, non-
           road and on-road mobile sources. The Canadian inventory does not contain
           PM emissions for point sources.
          Similar to the error in double counting on-road mobile PM emissions in the
           non-WRAP states, we also discovered that we double counted NH3 and SO2
           emissions from on-road mobile sources in the non-WRAP states. MOBILE6
           generates all standard criteria pollutants (CO, NOx, VOC, NH3, SO2, PM2.5,
           and PM10). We originally supplemented the MOBILE6 emissions with NEI99
           data for NH3, SO2, and PM. During the Pre02a_36 modeling we caught the
           problem with double counting PM; it was not until much later that we caught
           the issue with NH3 and SO2. The final Pre02a_36 emissions double counted
           NH3 and SO2 from on-road mobile sources in the non-WRAP states.

6.6 Emissions quality assurance
The Pre02a_36 emissions simulation not only tested the configuration and input files to
be used for the WRAP 2002 §308 modeling, it was also used to test the implementation
of the emissions QA protocol that CEP developed for the WRAP modeling. CEP created
an entirely new approach to managing the emissions modeling project for the WRAP by
employing multiple modelers to oversee, QA, and model the emissions. Complete details
on the modeling protocol are contained in Adelman, 2004. The different types of



                                           40
analysis used in the QA process are listed in the QA protocol. The QA products for
tracking and documenting the QA processes are contained in the Appendices of this
report. Explanations and examples of the different QA products are contained in this
section.
All of the input files used for the Pre02a_36 modeling are detailed in Appendix A.
Appendix B contains state bar chart summaries of the major source categories by
pollutant (not including fires or biogenics). Appendix C lists the SMOKE configuration
options for all of the SMOKE programs by source category. Appendix D contains the
QA tracking worksheet listing the major issues encountered in the Pre02a_36 modeling
and how the issues were resolved. Appendix E contains a QA checklist documenting the
QA steps performed per the emissions modeling QA protocol developed for the WRAP
(Adelman, 2004).
The WRAP RMC maintains a website with documentation and QA graphics for the
Pre02a_36 simulation. The base page for the §308 emissions modeling documentation is
here: http://pah.cert.ucr.edu/aqm/308/emissions.shtml. Documents with the same
information contained in the Appendices of this report are linked to off of this page. The
set of emissions QA regression spreadsheets are located here:
http://pah.cert.ucr.edu/rmc/2002/QA_pre02a36.reports/sheets and the web page
containing links to the various QA graphics is here:
http://pah.cert.ucr.edu/aqm/308/qa_pre02a36.shtml. Brief explanations of the QA
products contained in this report and on the RMC pages follows.

6.6.1 QA Products
The QA products contained in the Appendices are described at the beginning of each
Appendix section. Along with the information in the Appendices, spreadsheets and
documents on the RMC website contain tabulated information about the input and output
data for the Pre02a_36 simulation. Descriptions of the source and contents of the input
files and the location and sizes of the output files are tabulated in these documents. A
suite of QA graphics are also presented on the RMC website. Brief descriptions and
examples of these QA graphics follow.

6.6.1.1   Spatial Plots
Generated with the Package for Analysis and Visualization of Environmental data
(PAVE), 2-D spatial plots for each day for every pollutant in each major source category
(e.g. area, point, non-road mobile, etc) show the extent of the spatial coverage of the
emissions data. These plots also indicate the relative magnitude of the emissions data for
the different regions of the domain at different times of the year. We use these plots
primarily to confirm that emissions are not located incorrectly (e.g. over water cells),
there are no unexpected areas of the domain that are missing emissions data, and that
there aren’t areas of the domain with extremely high emissions relative to the
surrounding cells. Figure 6-1 and Figure 6-2 are examples of the 2-D spatial plots used
to QA the Pre02a_36 emissions modeling results.




                                            41
          Figure 6-1. Spatial plot of area source NH3 emissions; August 30, 2002




   Figure 6-2. Spatial plot of biogenic source isoprene emissions; August 27, 2002


6.6.1.2    Diurnal Time Series Plots
Time series-plots for each day for every pollutant in each major source category (e.g.
area, point, non-road mobile) show the diurnal emissions distribution averaged across the
entire domain. We use the diurnal time-series graphics to look for signature emissions


                                           42
patterns in the different source categories, differences in weekday and weekend diurnal
patterns, and a smooth distribution of the emissions from hour to hour. Spikes or
sporadic diurnal patterns indicate problems in the temporal distribution of the emissions.
Figure 6-3 and Figure 6-4 are examples of the diurnal time-series plots used to QA the
Pre02a_36 emissions simulation results.




Figure 6-3. Diurnal time-series plot of area source NO emissions; February 1, 2002




Figure 6-4. Diurnal time-series plot of on-road mobile soured NO emissions; June 1,
                                         2002




                                            43
6.6.1.3   Vertical Layer Profiles and Vertical Layer Fraction Plots
To QA the vertical distribution of point source emissions we use vertical layer profiles
and fraction plots averaged over the entire domain to illustrate how the emissions are
allocated to the model layers. We create these vertical distribution plots for each model
day for every pollutant for emissions sources treated as point sources (e.g. stationary
point and fires). The vertical distribution plots show how much of the emissions are
allocated to the different model layers, how high the emissions are injected into the
model, and how different pollutants have different vertical distributions. Figure 6-5 and
Figure 6-6 are examples of vertical layer profile and fraction plots used for the QA of the
Pre02a_36 point source emissions.




Figure 6-5. Vertical layer profile for point        Figure 6-6. Vertical layer fractions for
        source SO2; June 14, 2002                      point source SO2; June 14, 2002


6.6.1.4   Episode Time Series Plots
Episode time series-plots for every pollutant in each major source category (e.g. area,
point, non-road mobile, etc) show the temporal emissions distribution for the modeling
episode averaged across the entire domain. We use the episode time-series graphics to
look for seasonal and monthly temporal patterns, holiday signatures, and
weekday/weekend effects. Spikes or sporadic diurnal patterns indicate problems in the
temporal distribution of the emissions. Figure 6-7 and Figure 6-8 are examples of the
episode time-series plots used to QA the Pre02a_36 emissions simulation results. As
case Pre02a_36 only covered the winter and summer months of 2002, these plots have
missing data in the spring and fall months.




                                              44
    Figure 6-7. Episode time series plot of point source NOx emissions; Winter/Summer, 2002




Figure 6-8. Episode time series plot of on-road mobile source NOx emissions; Winter/Summer, 2002




                                              45
7 Conclusions
2003 was a transition year for the WRAP-RMC emissions modeling team where we
finished the modeling started as part of the Jumpstart phase of the project for §309
SIP/TIP development and began to prepare for the §308 modeling. This report
documented the work completed to finalize the 1996 base case emissions for the §309
modeling. We engaged in tool and capacity development to prepare for the §308
modeling and to position us to avoid the problems that we encountered during the
Jumpstart modeling. A summary of the tasks that the RMC emissions team worked on to
prepare for the §308 modeling follows.
          We initiated a review of the temporal and speciation profiles applied to the
           WRAP emissions. While the completion of this task is still pending, we have
           better defined the requirements for completing the review. We anticipate
           continuing the SMOKE input data review process and working on putting a
           mechanism in place to facilitate the reviews by the WRAP states.
          The emissions regressions tool prototype developed in 2003 is in beta form
           and performs the basic functions as stated in the 2003 work plan. With
           additional resources we would like to further develop the regressions tool to
           work with scripts to automate the creation of the emissions regressions that we
           use for doing QA on the SMOKE output.
          We successfully developed, rolled out, and tested a quality assurance protocol
           for emissions modeling and are now actively using it during the §308
           modeling. The QA protocol defined a strict set of procedures for ensuring that
           emissions modeling projects are well documented and organized. Templates
           for tracking the progress and producing documentation of SMOKE modeling
           facilitate the QA process. We made some refinements to the QA protocol
           based on our experience with the initial implementation
          We completed a six month (winter/summer) emissions shake out simulation
           for 2002 to test the installation of SMOKEv2 at the RMC, collect and process
           the basic inventory and SMOKE ancillary input files for the 2002 modeling,
           shake out any problems with the run scripts, and test the implementation of
           the QA protocol.
As of the writing of this report we have completed two subsequent annual simulations for
the preliminary §308 modeling. The tools and procedures developed in 2003 well
prepared the RMC emissions modeling team to produce, QA, and deliver SMOKE output
efficiently and on schedule. The project management infrastructure that we developed in
the form of the Bugzilla issue tracking system, weekly emissions QA conference calls,
and well-defined roles for each member of the modeling team has produced a system that
is working as well as we had originally envisioned when first discussing the concept of a
comprehensive emissions QA protocol over a year ago. We anticipate continuing to
make progress on the emissions modeling infrastructure, the modeling tools, and in
reviewing the data throughout 2004 and beyond.




                                           46
8 References
Tonnesen, G., M. Houyoux, R. Morris, “WRAP Regional Modeling Center 2003 Work
   Plan,” University of California at Riverside – Center for Environmental Research and
   Technology, addressed to Tom Moore – Western Regional Air Partnership, April 15,
   2003.
Adelman, Z, “Memorandum: Temporal and Speciation SCC Improvements”, University
   of North Carolina – Carolina Environmental Program, addressed to the WRAP
   Emissions Forum and WRAP States, September 19, 2003
U.S. EPA, “Operation Shakeout,” Contract no. 68D-02-066, Assignment no. 1-01, Tasks
   4, 5, 6, March 19, 2003.
Houyoux, M., Z. Adelman, U. Shankar, R. Morris. “Final Report: WRAP Regional
   Modeling Center – Short-Term Modeling Analysis,” University of North Carolina –
   Carolina Environmental Program and ENVIRON, addressed to the WRAP Modeling
   Forum, March 31, 2003
Adelman, Z. “Quality Assurance Protocol: WRAP RMC Emissions Modeling with
   SMOKE,” University of North Carolina – Carolina Environmental Program,
   addressed to WRAP Modeling Forum, January 7, 2004.
Adelman, Z. “Quality Assurance Protocol: WRAP RMC Emissions Modeling with
   SMOKE,” University of North Carolina – Carolina Environmental Program,
   presented to the WRAP Modeling Forum at UC Riverside, October 8, 2003.
McNelis, D., “Re: The WRAP-UNC Regional Modeling Center Emissions Support
  Contract (WGA Contract No. 30203-33),” University of North Carolina – Carolina
  Environmental Program, addressed to Mr. Richard Halvey, Western Governors
  Association, September 26, 2003.
Pollack, A. et al. “Final Report: Development of WRAP Mobile Source Emission
    Inventories,” presented to the Western Governors’ Association, February 9, 2004,
    http://www.wrapair.org/forums/ef/inventories/mobile/040209Final_MSEI.pdf.
Pollack, A. “2003 On-Road SMOKE Files,” email to Zachariah Adelman and
    Mohammad Omary, cc to Tom Moore, Gail Tonnesen, and Ralph Morris, December
    9, 2003, email: APollack@environcorp.com.
Air Sciences, Inc., “Draft Final Report – 1996 Fire Emissions Inventory,” addressed to
    the WGA/WRAP, December, 2002,
    http://www.wrapair.org/forums/fejf/documents/emissions/FEJF1996EIReport_02120
    8.PDF.
Seppanen, C. “Remove refueling emissions from MEPROC file,” submitted as a ticket to
   the CEP Bugzilla email tracking system, http://bugz.unc.edu/show_bug.cgi?id=867,
   January 20, 2004.




                                          47

				
DOCUMENT INFO
Shared By:
Categories:
Tags:
Stats:
views:35
posted:6/14/2012
language:English
pages:51