Noor V. Gillani1*, Arastoo Biazar1, Yu-Ling Wu'
                              James Godowitch2, Jason Ching2, Robert Imhoff

                             1 University of Alabama in Huntsville, Huntsville, AL
                     2 Atmospheric Sciences Modeling Division, Air Resources Laboratory,
                  National Oceanic & Atmospheric Administration, Research Triangle Park, NC
                              3 Tennessee Valley Authority, Muscle Shoals, AL

    Modets-3/CMAQ (Byun  et ai.,  1998)  is  the  new
comprehensive,  multiseale,  multi-pollutant  Euierian
modeling system currently  in development under the
sponsorship of the  U.S. EPA. in regional applications,
as  with  its  predecessor  (RADM), its  coarse  grid
resolution is likely to be between 20 and 80 km.  Unlike
RADM,  however,  CMAQ   can  achieve  finer spatial
resolution by the use of nested grids in urban-industrial
areas with concentrated sources, and by plume-in-grid
treatment  (PinG)  of  major elevated  point-source
emissions (MEPSEs).  Most MEPSEs are related to
large fossil-fuel power plants. In 1993,  the utility sector
was estimated to contribute more than 72% and 33%,
respectively, of the US emissions of SOx and NOx
(EPA,  1994).  Just the largest  100  and 200 of the
thousands of "major point sources" in the US contribute
about 35% and 50%,  respectively, of the total NOx
emissions of this source category.  Thus, these largest
MEPSEs constitute an important  class, by themselves,
of the sources of ozone and aerosol precursors.  Since
these MEPSEs  do  not also emit VOC (the other  key
precursor of the ozone-aerosol chemistry), the mixing of
MEPSE NOx and background VOC is critical for plume
chemistry which is, therefore, diffusion-limited. As  this
chemistry is also very non-linear, proper simulation of
such  mixing is of critical  importance for accurate
determination of the impact of MEPSE emissions.  In
RADM and its contemporary and predecessor regional
grid models, such emissions were instantly  mixed  with
the VOC of the  entire emission  grid  cell,  thereby
completely   bypassing  the  diffusion-limited  subgrid-
scale (SGS) evolution of the plumes. Current practice is
shifting  towards  use  of PinG  to simulate  the SGS
transport and chemistry in grid models. PinG in Models-
3 is significantly different from that in other grid models.
It is aimed  particularly at  realistic  treatment of rural
MEPSE plumes  in a coarse  regional grid. This paper
provides an overview of PinG in Models-3.


PinG has been implemented in the PARIS (Seigneur et
al., 1983), UAM-V (Morris et al., 1992), SAQM (Myer et

*  Corresponding author address: NoorV. Giani, Earth
System Science Lab., Univ.  of Alabama, Huntsville, AL
35899. email: gillani@atmos.uah.edu
al., 1996) and  URM  (Kumar and  Russell,  1996) grid
models. In all of these, multiple SGS Lagrangian plumes
(or puffs) within a domain can be simulated, but they are
assumed not to interact directly with each other.  Each
plume  expands as it  advects,  entraining  the back-
ground, and the plume-background mixture then reacts.
Spatial resolution is achieved through discretization of
the plume cross-section:  PARIS used  a  rectangular
plume cross-section divided into vertically  well-mixed
pillars; the other models use an elliptical plume section
divided into concentric rings. The grid size was 4 km in
the single application of PARIS, and as high as 12 km in
UAM-V and SAQM. In all three, the outermost pillars or
elliptic rings are sequentially shed off to  the grid model
as the remaining plume attains the width of the grid cell,
and the SGS simulation stops  when  the core  of the
plume  attains grid  size.  In URM,  the  elliptic  plume
simulation  is limited to a fixed one hour, when the whole
plume is handed over.  Strictly  speaking, the  elliptic
plume model formulation is valid only as long as the
plume is fully elevated.  In  the convective  boundary
layer (CBL), this condition  typically lasts for  less than
one hour.  The common observation in all of the above
PinG applications has been that the PinG  treatment
does not make  much  difference except very locally.
This    is   not   surprising,    since   the    above
formulations/applications have been focussed on the
detailed simulation of  plume-core chemistry only in the
early phase of plume evolution.


    Models-3 is scheduled  to be released in 1998. It will
include a  preliminary developmental version of PinG
which will  be  best suited  for  rural MEPSEs within  a
coarse regional grid (A > 12 km or so) where the SGS
error in the representation of MEPSEs is largest without
PinG treatment.  Figure 1 provides a graphical perspec-
tive of the number and distribution of such MEPSEs in
eastern USA.  The "rural" cells represent likely locations
of rural MEPSEs with  total NOx emission  rate  (QNO*)
exceeding approximately  0.3 kg/s  or 10,000 tons/y
(approximately, the cut-off source strength for the top
200  MEPSEs in the  US);  "urban" ceils  are likely
locations of urban areas (with or without MEPSEs). The
urban and  rural cells together occupy only  3% of the
land mass  shown, but contribute more than 50% of the

total NOx emissions of the  region (Gillani and  Pleim,
1996).  Some of the "rural"  cells represent  suburban
MEPSEs, but many are truly rural. There  is a concen-
tration of MEPSEs in the Ohio River valley, but many of
the other rural  MEPSEs are quite isolated.    PinG
currently assumes  linear superposition of  SGS MEPSE
plumes.  Rural MEPSEs have the lowest likelihood of
interactions, not only with other MEPSEs,  but also with
the thousands of non-MEPSE "major point  sources".
    Figure 2 illustrates the typical chemical evolution of
a large isolated rural power plant plume, consisting of
three stages (Gillani and Wilson, 1980). In stage 1, the
relatively  fresh  plume  is   dominated  by  primary
emissions and a deficit of ozone due to its titration by
NO. The chemistry is mostly inorganic and VOC-limited,
In stage  3, at the other end, the plume is  substantially
diluted and has a high VOCrNQx ratio, and  is chemically
mature,  with little  NOx remaining and an excess of
ozone; the chemistry here is mostly  NOx-Iimited, much
as in the background. Stage 2 marks the transition of
the chemistry from VOC-limited (Stage 1) to NOx-Iimited
(Stage 3). It is marked by growing ozone wings at plume
edges, resulting from vigorous chemistry there fed by a
good supply of both VOC (background entrainment) and
NOx (diffusion from plume core).  This is the  diffusion-
limited chemical stage with a complex SGS structure.
Its proper simulation is the key to accurate characteri-
zation of SGS plume chemistry. Indeed, such charac-
terization requires that a chemical criterion of success-
ful completion of stage 2 chemistry be satisfied  in the
PinG solution before plume handover to the grid solution
(Gillani and Pleim, 1996),  This chemical criterion is not
observed in the PinG treatments of UAM-V, SAQM and
URM, because they are necessarily limited to near-field
and plume core chemistry (as in stage 1) as a result of
                            Figure 1. Distribution of 20 km x 20 km "rural" (D)  and
                            "urban" ( ) cells with NO,  emission flux, qNO, , greater
                            than 101J molecules cm"J s"1,  "Urban" cells are defined
                            by qvoG/QNo.3" 1 (C:N, by volume), and "rural" cells by
                            quoc/qNOl[< 1. (Figure from Gillani and Pleim, 1996).

                            their elevated  elliptic  ring formulation. As  shown in
                            Figure 2 (stage 2),  there is much fine structure  in the
                            plume even when it  is about 20 km wide.  The plumes of
                            very large MEPSEs in the eastern USA (QNOx - 3 kg
                            NOj/s or more) generally  do not  attain full chemical
                            maturation until their width is about 30 km; smaller ones
                            mature at smaller  widths. In rural Tennessee,  such
                            maturation typically requires about 100 km  and 6 h of
                            plume transport in the summer daytime CBL for large
    SO,        Stage  I

                  .A      /'-.Ozone
                                                    Stage  3
                                                                    _  Bscat
                                                                           .*' Sulfate
    Figure 2. Aircraft data of SO2 (surrogate for NOx), ozone, sulfate and Bscat  (surrogate for fine aerosol mass)
    during crosswind traverses on 23 Aug 1978 through the plume of a large MEPSE (TVA's Cumberland power plant
    in northwest Tennessee, QNOx - 3 kg/s), illustrating the three stages of plume chemical development. The three
    sets of profiles shown represent quasi-Lagrangian information at three downwind sections of the - 0700 plume
    release. Plume (age.width) at the distances shown are approx. (4 h, 10 km), (5.5 h, 20 km) and (8 h, 32 km).  The
    plume was in elevated stable layers until almost 1030, and had just completed fumigation at 80 km, where it was
    still more or less in stage 1. (Figure taken from Gillani et al., 1981).

MEPSEs, and less for smaller ones (Giilani et al, 1997),
During much of this transport, the plume is fully  mixed
over the CBL and  in contact with  the ground.   The
elevated elliptical plume formulation of UAM-V, SAQM
and URM is not the appropriate framework for such
plume transport-chemistry simulation.  The rectangular
plume geometry of  PARIS is much  more appropriate.
Unfortunately, the single application of PARIS (Seigneur
et al.,  1983)  was within a 4 km grid and the detailed
plume simulation was confined within the inner 4-km
core, dumping the critical plume edges with much SGS
structure prematurely to the coarse grid solution.
    Based on such observational evidence, our goal in
PinG of Models-3 is to perform detailed plume physical
and chemical   simulations  until  plume   chemical
maturation. Figure 3 illustrates our approach in concept,
in  keeping with  the available hourly meteorological
information, our detailed SGS plume  simulation in PinG
is designed,  in  essence,  to transform each steady
hourly emission rate  (v)  of primary species "p" from
the target MEPSE source to  a steady  hourly delivered
rate (j , excess over background)  of all modeled
species *f, at the appropriate plume handover time (two)
and location,  to the grid  model. Such transformation is
achieved by means of a  detailed  Lagrangian simulation
of plume dynamics and kinetics in which plume transport
after touchdown (within  the CBL) is expected to  last
typically about 3 to  7 hours  in the summer in eastern
USA (depending  mainly on  source strength,  plume
spread rate, background  chemical composition, and
   jre_3.  Conceptual model of a subgrid scale
         plume simulation in PinG of Models-3,
sunlight). Such simulation is facilitated by a formulation
based on a rectangular plume cross-section.  A major
limitation of this  Lagrangian  plume approach  is  that
conditions  of  high plume  shear (particularly speed
shear) cannot be modeled accurately. Such conditions
are most  likely during the nocturnal  transport  of a
vertically well-mixed daytime  plume  of  1-2  km depth.
Most  daytime  releases,  except those  of the  late
afternoon, are likely to be chemically mature before the
collapse of the CBL in the  evening. Night-time releases
of MEPSEs in the elevated stable layers typically result
in vertically  confined plumes  which  are spared much
shear because of their  small  vertical  extent.   Our
practice is to hand the plume over  to the grid model
whenever conditions prevail, e.g., high shear, which the
plume model cannot simulate with acceptable accuracy.

4.  Mode!s-3/PinG :   THE  BUILDING BLOCKS

    For each MEPSE plume treated in  PinG, its  SGS
plume dynamics and  chemistry  must be simulated for
each  release.   We have developed  a new  plume
dynamics  model  (PDM).    For  chemistry,  we have
adapted the U. of  AL Huntsville Lagrangian Reactive
Plume Model; UAH-LRPM  (Giilani, 1986), as the building
block.  The Lagrangian entity tracked is the expanding
plume section representing emissions into a thin semi-
infinite vertical slab as it passes the  target  MEPSE
source.  The new PDM simulates plume rise, plume
transport and plume crosswind spread (horizontal  and
vertical).  For releases into the nocturnal stable layer
above the shallow mixing layer, it also simulates fumiga-
tion into the rising mixing layer in the morning.  PDM
uses state-of-the-art  parameterizations of plume  rise,
and of oy and o, representing horizontal and vertical
plume spread,   or is  assumed  to consist   of  two
components representing turbulent  spread (o,,)  and
shear spread (a^), as follows:  af = oy/ + oyihz. oyjh is
due to directional shear in the mean wind field,  it is
small in coherent nocturnal plumes, but within the CBL,
its contribution can be significant after  the plume  has
attained substantial vertical depth. The implementation
of PDM, of course, requires appropriate meteorological
inputs, including dynamic 3-D temperature and  wind
fields, as well as 2-D information about mixing height, u*
and w* (friction and convective velocities). PDM outputs
(typically every  10 minutes) include,  most importantly,
the  plume cross-sectional geometry, i.e., the co-ordi-
nates of  the centroid and  of the four plume  edges,
left/right and top/bottom,
    The chemistry model (UAH-LRPM; Giilani, 1986)
simulates the complete mass balance of reactive
species within such an expanding plume section. Figure
4 illustrates the processes simulated in the LRPM. The
upper plot shows a time-height  view of night-time and
daytime plume releases. The lower plot shows the
plume cross-section, discretized into  vertically well-
mixed pillars, and its expansion from t to t+At, where At

                                                                ?H?H??&; roc nhoca H^I!H'.!*i!!(!l!Hiiii!i
                                                                iiiiiiiiii'i vaas-pnase iirssissisii^s'iisrssrmrt
                                                               Surface emissions &   depositions









~^~ at (t+At)

from sides
           jre4.  An illustration of the formulation of the UAH-Lagrangian Reactive Plume Model showing
                 the processes simulated : (a) the time-height view ; (b) the cross-sectional view.
denotes the transport (advection) time step.   In  the
figure,  the  pillar widths are shown as equal,  but  the
model   can  simulate  any   arbitrary   non-uniform
discretization also (e.g., "Gaussian" framework, as in
PARIS).  At each  plume  expansion,  horizontally and
vertically, the plume mass not  only dilutes, but there is
also entrainment from the sides (the background) and
from aloft,  as well as entrainment and  detrainment
between adjacent  pillars.  Also, there is (turbulent and
shear)  diffusion of mass  across  pillar interfaces.  In
addition, the model permits uniform surface emissions
and dry deposition to/from each pillar.  Finally, LRPM
performs detailed  gas-phase chemistry in each  pillar,
with the same options of chemical  mechanisms as
available in the Chemistry-Transport  Model (CTM)  of
Models-3.  LRPM also uses the same chemistry solvers
as  does  CTM.    It derives   background and aloft
information from the appropriate grid cell  solution.  It
receives   dynamic  plume   geometry   from  PDM

5. Models-3/PinG :   IMPLEMENTATION

    Figure 5 shows the relationship of PDM and PinG to
Models-3/CMAQ.  The main Chemistry-Transport Model
(CTM)  of CMAQ has options  to use  three levels  of
nested gridding, any one of several state-of-the-art gas-
phase  chemical mechanisms,  aqueous chemistry and
an aerosol  module. It  has two main pre-processors:
MEPPS for emissions, and MM5 for meteorology.  MEPPS
also performs the selection of MEPSEs from the "major
point sources", based on the following two criteria:
1.  QNO, > Qc, where Qc is a critical cut-off value
         (default Qc = 0.3  kg/s = 10 kT/y as NO2);
2.   h(stack) > hc, where hc is a critical cut-off value of
         stack height (default hc = 100 m).
Interface  processors ECIP  and  MCIP  process  the
outputs of MEPPS  and MM5, respectively,  to generate
appropriate inputs for CTM, PDM and PinG. PDM runs as
a pre-processor of PinG,  as shown.  It  performs  the
complete plume dynamics for each hourly release of
each MEPSE selected for  PinG treatment, generating
Figure  5.  Placement  of  PDM and  PinG  in Models-

output files containing dynamic plume geometry as well
as other related information. The actual plume transport-
chemistry simulation (as in LRPM) occurs in PinG, which
is embedded in CTM and operates interactively with it at
the level  of each CTM advection time step (= 10  min).
For a typical 5-day run of CTM including 100  MEPSEs,
PDM  and PinG  must process  a total of 12000 hourly
plume releases.  At each interaction with CTM,  PinG
initiates new plumes (if any) and expands every  "live"
plume (typically  500 - 1500  for a  100  MEPSE  run,
depending  on  time of day)  according  to the  PDM-
supplied  plume  dynamics. Consequently,  each plume
entrains mass from the grid  solution, and also receives
surface emissions (after plume touch-down). Also, gas-
phase chemistry (same as in CTM except for night-time
plume releases, for which a much simpler  chemistry is
implemented),  dry  deposition and  other  process
operations are implemented.  PinG then checks  each
live plume for the handover decision, and if necessary,
performs  plume  handover to the appropriate grid cells,
updates the grid solution and  returns  control to CTM.
The following handover criteria are used.
1. Plume reaches the boundary of the grid domain;
2. Plume is exposed to precipitation > critical level;
3. Plume enters an "urban" cell, as defined previously,
    after it has touched down. Night-time releases of
    MEPSEs in such cells may continue to remain alive
    as elevated plumes transporting above the city.
4. Major  point source emissions of NOx in the cells on
    either side of the plume exceed a cutoff  level. This
    criterion is forced because of the uni-layer plume
    limitation of the current LRPM.
5. Plume chemistry attains maturity based  on the
    criterion that plume average value of the chemical
    age parameter, NOz/NOy, exceeds a prescribed
    fraction (= 0.9) of the corresponding average value in
    plume environment.
6. Plume width exceeds grid size.
In addition, PinG also performs partial plume handover.
If along a plume trajectory in  the CBL, the mixing height
decreases by more than a prescribed fraction, then the
portion of the plume above the diminished mixing layer is
handed  over,  since  LRPM currently has only   one
vertical plume layer. According to this criterion, a well-
mixed daytime plume will be handed over piecemeal as
the mixing height collapses at the end of the solar day.
In this manner, the possibility  of high shear of such a
plume during nocturnal transport is also avoided.


    Before the release of Models-3/CMAQ in 1998, the
initial operational versions (lOVs) of  PDM and PinG will
be tested with multi-plume  data of the SOS/Nashville
Ozone Field Study of 1995.  Major areas of anticipated
future improvements of PinG include:
1.  Higher vertical resolution of the  plume to  at  least
    match that of MM5 and CTM, both of which include
   about seven layers from ground up to 800  mb (~ 2
   km).  Such an improvement will greatly improve the
   vertical distribution of surface emissions as well as
   pave  the way for better treatment of  non-MEPSE
   elevated major point-source emissions;
2. Inclusion of aqueous-phase chemistry  (same as in
3. Inclusion of an aerosol module (same as in CTM);
4. Improved   treatment   of   SGS   plume-plume

Acknowledgements/Disclaimer:  This work is supported jointly
by TVA (under Contract TV-88817V to UAH) and by EPA (under
Contract DW64937190 to TVA).  This paper has been reviewed in
accordance with EPA's  peer and administrative review policies
and approved for presentation and publication. Mention of trade
names or commercial products does not constitute endorsement or
recommendation for use.


Byun,  D.,   et al.,  1998:   Description  of Models-3
   Community Multiscale Air Quality (CMAQ) modeling
   system. 10th  Joint AMS-AWMA Conf.  on Air Poll.
   Met., Phoenix AZ, 11-16 Jan. 1998.
EPA, 1994: National Air Quality and Emissions Trends
   Report for 1993,  EPA 454/R-94-026, OAQPS, U.S.
   EPA, Res. Triangle Park, NC .
Gillani, N.V.  and  W.E. Wilson,  1980:  Formation and
   transport of ozone  and aerosols  in power plant
   plumes. Annals N. Y. Acad.  Sci. 338, 276-296.
Gillani, N.V., S. Kohli and W.E. Wilson, 1981:  Gas-to-
   particle conversion of sulfur in power plant plumes: I.
   Parameterization of the gas-phase conversion rate
   in moderately polluted ambient conditions.  Atmos.
   Environ. 15, 2293-2313.
Gillani, N.V., 1986: Ozone formation  in pollutant plumes:
   Development  and application  of a reactive plume
   model with arbitrary crosswind  resolution.  EPA-
   600/S3-86-051, U.S. EPA, Res. Triangle Park, NC.
Gillani, N.V. and J.E.  Pleim,  1996:   Subgrid-scale
   features of anthropogenic  emissions  of VOC and
   NOx in the context of regional Eulerian  models.
   Atmos. Environ. 30(12), 2043-2059.
Gillani, N.V., et al., 1997: Relative production of ozone
   and nitrates in urban and rural power plant plumes:
   I. Composite results of ten days of field measure-
     ments. Submitted, J. Geophys. Res.
Kumar, N. and A.G. Russell, 1996:  Development of a
   computationally  efficient   reactive  subgrid-scale
   plume model.  J. Geophys. Res. 101, 16737-16744.
Morris, R.E., et al., 1992:  Overview of the variable-gid
   UAM-V. 85th Annual Meeting, AWMA, Kansas City.
Myer, T., et al., 1996: The implementation of a plume-in-
   grid module in SAQM.    Report SYSAPP-96/06,
   Systems Applications Int., San Rafael, CA.
Seigneur, C., et al., 1983: On  the  treatment of point-
   source emissions  in  urban air  quality modeling.
   Atmos. Environ. 17(9),  1655-1676.

                               TECHNICAL REPORT  DATA

 The Plume-in-Grid Treatment of Major Elevated  Point-
 Source Emissions in Models-3
            S.REPORT DATE
                                                          6.PERFORMING ORGANIZATION CODE

 Noor V.  Gillani1,  Arastoo Biazar1, Yu-Ling Wu1, James
 Godowitch2,  Jason Ching2,  and Robert Imhoff3

 *Univ. Of Alabatna-Huntsville, Huntsville,  KL
 ^ame as Block  12
 3Tennessee Valley Authority, Muscle  Shoals,  AL
            10.PROGRAM ELEMENT NO.
            11. CONTRACT/GRANT NO.

U.S.  Environmental Protection Agency
Office  of Research and Development
National  Exposure Research Laboratory
Research  Triangle Park, NC 27711

            Preprint, FY-98


The plume-in-grid modeling approach for treating the subgrid scale dispersion,
transport,  and chemistry of pollutants contained in plumes  released from major point
source stacks is described.  A plume dynamics model provides the plume rise height,
plume position,  and the horizontal and vertical plume dispersions in the subgrid
scale phase when the pollutants  are to be simulated by  the  Lagrangian reactive plume
module.   The Lagrangian plume algorithm is embedded in  the  Eulerian chemistry-
transport grid model and operates  simultaneously.  A coupling exists between the
models as the grid concentrations  provide boundary conditions for entrainment to the
subgrid plume sections, and a feedback ocurrs when the  plume size and chemical
criteria  indicate a plume's pollutants are to be transferred to the grid system  for
further simulation.  The methods for simulating the plume processes are also
presented and the input/output data involved with these two components of the
Models3 plume-in-grid approach are discussed.
                               KEY WORDS AND DOCUMENT ANALYSIS

                                                                         21.NO. OF PAGES
                                              20. SECURITY CLASS (This

                                                                         22. PRICE