United States
Environmental Protection
Agency
Office of Air Quality
Planning and Standards
Research Triangle Park, NC 27711
EPA-454/R-98-020
December 1998
AIR
& EPA
A Comparison of CALPUFF with ISC3
-------
ACKNOWLEDGMENTS
Special credit and thanks are due John Irwin, NOAA for his
technical assistance and advice through all phases of the project,
from study design and meteorological data selection to analysis and
presentation of results. In the model comparisons, credit is due to
Tom Coulter, EPA for his work on the steady state analyses and to
Pete Eckhoff for the variable meteorology analyses. Thanks are
also due Dave Strimaitis and Joe Scire of Earth Tech for their
cooperation and technical assistance with the CALPUFF runs.
DISCLAIMER
This report was reviewed by the Office of Air Quality Planning and
Standards, EPA for approval for publication. Mention of trade
names or commercial products is not intended to constitute
endorsement or recommendation for use.
-------
PREFACE
In this report a comparison is made of two different dispersion
models, CALPUFF and ISC3. CALPUFF is a Lagrangian puff
model which simulates continuous puffs of pollutants released into
the ambient flow, whereas ISC3 is a Gaussian plume model that
treats emissions from a source as a contiguous mass. CALPUFF
may be configured to treat emissions as integrated puffs or as slugs.
ISC3 is currently recommended for routine use in assessing source
impacts involving transport distances of less than 50km. This
report is being released to establish part of the basis for review of
the consequences resulting from use of CALPUFF in routine
dispersion modeling of air pollution impacts.
-------
TABLE OF CONTENTS
1. Introduction 1
2. Technical Background 1
3. Results 3
3.1 Steady State (screening) Meteorological Conditions 3
3.1.1 Residual Analysis 4
3.1.2 Point Sources (surface and elevated) 6
3.1.3 Area Source 7
3.1.4 Volume source 7
3.2 Variable Meteorological Conditions 8
3.2.1 Scenarios for Sensitivity Study 8
3.2.2 Preliminary Studies 9
3.2.3 Sensitivity Study 14
4. Summary and Conclusions 17
4.1 Steady State Meteorological Conditions 17
4.2 Variable Meteorological Conditions 18
4.3 Conclusion 19
5. References 20
Appendices
A. Switch settings for CALPUFF input file to emulate ISC3's "Regulatory Default" mode
B. Meteorological conditions for the steady state CALPUFF/ISC3 comparisons
C. Characteristics for sources used in the CALPUFF/ISC3 comparisons
D. Receptor array used in the CALPUFF/ISC3 comparisons
E. Puffs versus Slugs: CALPUFF's Two Simulation Modes
F. Summary statistics from performance matrix - point sources (Z; = 3000m)
G. Summary statistics from performance matrix - area source (emissions simulated as slugs)
H. ISCST3's treatment of virtual sources
I. Summary statistics from performance matrix - volume source
J. Wind rose patterns
K. Puff and slug model concentrations
L. Additional figures illustrating results with variable meteorological conditions
-------
1. Introduction
With the initial use of models such as CALPUFF for regulatory applications, there is the
question of how the model will behave with respect to more widely used models like the
Industrial Source Complex Short Term (ISC3ST) model, hereafter ISC3. Several sensitivity and
comparison studies were designed and performed to determine how CALPUFF would behave
when set to emulate ISC3. The results of those runs were analyzed and are discussed here.
This evaluation features a systematic, phased series of implementation modes. Section 3.1
involves simple screening modes in which conditions are extremely limited and controlled.
Section 3.2 addresses the more general mode in which meteorological conditions are allowed to
vary hourly. Section 4 provides a summary and conclusions from this investigation. References
are listed in Section 5, followed by the appendices.
2. Technical Background
CALPUFF is a Lagrangian puff model. The model is programmed to simulate continuous
puffs of pollutants being emitted from a source into the ambient wind flow. As the wind flow
changes from hour to hour, the path each puff takes changes to the new wind flow direction.
Puff diffusion is Gaussian and concentrations are based on the contributions of each puff as it
passes over or near a receptor point. For these tests, CALPUFF was set to emit 99 puffs per hour
(default). A sufficiently large number of puffs is necessary to adequately reproduce the plume
solution at near-field receptors.
CALPUFF was originally designed for mesoscale applications and treated emissions as
integrated puffs. As features were added to the model for handling local-scale applications, it
was realized that use of the integrated puff approach was inefficient. A more efficient approach
was developed to treat the emissions as a slug, in which the slug is stretched so as to better
characterize local source impacts. The slug can be visualized as a group of overlapping circular
puffs having very small separation distances. When run in the slug mode, the hourly averaged
pollutant mass is spread evenly throughout the slug. For a given hour, if all of the hourly slug
has not passed over a receptor, concentrations are reduced by the mass that has not passed over
the receptor (Appendix E; Section 2.1 of Reference #2). Note that when run in a slug mode,
once the slug's lateral dispersion (oy) approaches the length of the slug itself (as eventually
happens with downwind distance), CALPUFF samples the pollutant mass as apuffto improve
computational efficiency. At sufficient downwind distance, there becomes no benefit or
advantage for the slug simulation.
-------
In the comparison studies described in this report, CALPUFF was run in both the puff mode
(emissions simulated as integrated puffs) and the slug mode (emissions simulated as slugs).
When the distinction between puffs and slugs is important or significant, they will appear in
italics (i.e., slugs or puffs; see Appendix E). In the generic sense, the use of "puffs" will be used
to connote the characterization of a continuous release of a series of overlapping averaged puffs,
in which the transport and dispersion of each puff is treated independently, based on local (time
and space varying) meteorological conditions. Whereas, the use of "plume" will be used to
connote the characterization of a continuous release, in which the release and sampling times are
long compared with the travel time from source to receptor, and the meteorological conditions
are steady state over the travel time.
-------
3. Results
In this comparison, CALPUFF (Version 4.0, level 960612) was compared with the latest
version of ISC3 (dated 96113). CALPUFF was run in a mode that enabled ISC3-type
meteorological data as input, and therefore winds are horizontally homogeneous for each hour.
ISC3 was implemented in the "Regulatory Default" mode and the input file for CALPUFF was
configured so as to emulate this to the best extent possible (see Appendix A). Both surface and
elevated sources were simulated for rural environments in flat terrain, free of obstacles.
3.1 Steady State (screening) Meteorological Conditions
In this approach to the comparison, meteorological conditions were held constant (as in
SCREENS) so as to express true model differences, i.e., without the bias of a varying (temporally
and spatially) meteorological regime. Meteorological data sets were synthesized with fixed
meteorological conditions (Pasquill-Gifford stability category, wind speed, and mixing height)
and were of duration estimated to be sufficient to advect CALPUFF's puffs to the edge of domain
(generally 24 - 48 hours). (Of course, ISC3's steady state plume reaches the edge of the domain
instantaneously.) For Pasquill-Gifford (P-G) stability category^, 5 wind speeds were used, for
B, there were 9 wind speeds, for C, 11 wind speeds, for D, 13 wind speeds, for E, 9 wind speeds,
and for F, 1 wind speeds. A matrix describing the basis for the 54 meteorological conditions
used is provided in Appendix B.
The elevated point sources were 35m, 100m and 200m, respectively. Surface releases were
simulated with a 2m point source, a 500m X 500m area source, and a typical volume source.
Characteristics for each source type are described in Appendix C. Sources were placed at the
center of a 2 X 2 grid cell domain, with grid spacing set to 150km. While effects within the first
50km are of most interest and significance, straight-line receptors were located with decreasing
density out to 100km (Appendix D). The 62 receptors were placed along a radial aligned at
360°, coincident with the bearing used for transport winds.
Unique model runs were made for each combination of source type and meteorological
condition (i.e., Pasquill-Gifford stability category, wind speed, and mixing height). Each model
was configured to output the highest hourly average concentration for SO2 (no deposition or
chemical transformation).
-------
3.1.1 Residual Analysis
For each pair of model runs (CALPUFF and ISC3), a signed residual (7?,, XCALPUFF -
s, H8m"3) was computed at each of the 62 receptors. From the 62 residuals, a mean (R,
jigm"3), standard deviation (OR, |igm"3), and sum of residuals squared (S Rt. ) were computed.
The statistic R provides an indication (sign) of bias along the receptor radial. The statistic OR
provides general indication of the variance along the receptor radial. Because many of the
absolute residuals \;
the receptor radial.
absolute residuals were quite small, S Rt provides a relatively robust indicator of accord along
Another robust statistic was envisioned in which the absolute residual at each receptor
was related to, say, ISCS's predicted concentration value at that receptor. Because of the
mathematical problem posed by zero values (can't divide by zero), the statistic %R;(% residual)
was defined in terms of the maximum concentration predicted by ISC3 for each run:
R
) 100
%ISC3max
The mean % residual follows as:
%R =
62
As with R, the statistic %R provides an indication (sign) of bias along the receptor radial.
Another statistic of interest was the Fractional Bias (FB):
R.
FB =
%CALPUFF
Having by definition a distribution from -2 to +2, a value of zero indicates no bias between
XcALPUFF and XlSC3-
-------
One problem that arises with the FB statistic is when the mean paired concentration is
very close to zero: the FB statistic can be artificially inflated to a value close to ±2. Since cases
in which the mean is close to zero are of little interest in this comparison, a filter was applied:
// (*C4LP^ + 1iscs} < o.ooi^ -3, then FB = 0.0
For each run pair (i.e., CALPUFF versus ISC3), a mean fractional bias was computed as:
FB = - - (62 receptors)
62
As with R and %R, FB provides an indication (sign) of bias along the receptor radial. While a
value of zero would be ideal for FB, the following was established as a "goal":
-0.10 < FB < 0.10
Specific instances for which this goal was not met were noted.
There are some caveats to the interpretation of FB. Its behavior is closely related to its
structure. Its value is influenced not only by the absolute difference of the paired concentrations,
but by their relative magnitude as well. Thus, modest R/s related to "large" x's (e.g., from a low
level release) yield modest FB;'s (and a modest FB). Such a scenario can include a fairly large
variance (OR = 56 |igm"3) and mean residual (e.g., R = -32 \\.gm ~3) along the receptor radial but
still result in a fairly low FB (e.g., FB = -0.06). Conversely, modest Rj's related to "small" x's
(e.g., from an elevated release) may yield substantial FB;'s. Such a scenario can include a modest
variance (OR = 0.5 |igm"3) and mean residual (e.g., R = 0.3 \igrn ~3) along the receptor radial but
still result in a sizeable FB (e.g., FB = 0.35). While a useful indicator of correspondence
between two quantities, the FB must be interpreted in the context of other comparison statistics.
At the conclusion of the runs, a performance matrix was created and aggregate statistics
were compiled. For basic residual analysis, the value, run (distinct combination of source type,
wind speed, P-G category, mixing height) and receptor for Ri(min) and Ri(max) were noted.
Likewise, across all runs, the value and run for %Rmin and %RmaK were noted, as were the value
and run for Rmin and ^max. Across all runs, the value and run for oR(min) and oR(max) were also
noted.
-------
Finally, for the FB statistic, the value, run and receptor for FBi(min) and FBi(max) were
noted. Across all runs, the value and run for FBmin and FBmiix were noted. The values and cases
in which FB did not meet the "10% goal" were also noted.
3.1.2 Point Sources (surface and elevated)1
3000m mixing height
To model the four point sources, CALPUFF had to be run 216 times while ISC3 was run 54
times.2 As indicated in Appendix A, CALPUFF was run in the slug mode to emulate ISCS's
Gaussian plume simulation.3 The results indicated good accord (Appendix F). For all cases,
FB < 0.10 (FBmax = 0.02). The maximum residual was 25.0 |igm"3 (0.13% of the concentration
mean at the incident receptor), while the minimum residual was -8.0 |igm"3 (0.03% of the
concentration mean at the incident receptor). Mean residuals for any run were less than one
|igm"3, and total range for OR was 0.0 - 3.2 |igm"3. Overall, perhaps the most practical
performance parameter was %R, which indicates accord well within one percent across all
release heights, meteorological conditions and receptors (the value for %Rm[n was -0.04% and
%JRmax was 0.13%). A qualitative inspection of residuals as they appear along the receptor array
indicated no distinct pattern of bias for any case. Across all runs, a slight negative bias
(CALPUFF relative to ISC3) is apparent for the 2m source, and the greatest variance is
associated with the 2m source, especially for P-G category A
500m mixing height
The array of runs was redone (again, using slugs) with mixing height reduced to 500m to
assess CALPUFF's response to reflection and to evaluate whether reflection is handled
equivalently. The results were quite good. In 43 cases, the plume centerline computed by ISC3
exceeded the mixing height and set ground level concentrations to zero. CALPUFF treated the
same cases equivalently. For the remaining 173 cases, \FB\ < 0.10 (FBmax = 0.02). The
comparison statistics bear a striking resemblance to those for Z; = 3000m. Mean residuals for
Certain runs may be referenced, e.g., D20H100 or Blp5H2. Under this nomenclature, the first signifies a 100m source running
under D stability with 20 ms"1 winds. The second would be a 2m source running under B stability with 1.5 ms"1 winds.
Each source was modeled 54 times for each of two mixing heights. In the current version of CALPUFF, it is impossible to
isolate impacts from more than one source per run. ISC3, however, may be configured to simulate multiple sources during a single
run and isolate impacts individually.
For a description of integrated puff and slug formulations, see Sections 2.1.1 and 2.1.2 of the CALPUFF User's Guide
(Reference #2) and Appendix E of this report.
-------
any run were also less than one |igm"3, and %R indicates accord to well within one percent
across all release heights, meteorological conditions and receptors (the value for %Rmin was
0.0% and %Rmax was 0.08%). A qualitative inspection of residuals as they appear along the
receptor array also indicated no distinct pattern of bias for any case. As with the 3000m Z; case,
a slight negative bias is apparent for the 2m source, and the greatest variance is associated with
the 2m source, especially for P-G category A
3.1.3 Area Source
The area source was modeled with emissions simulated as slugs. While a significant
difference would be expected between the behavior of puffs and slugs, slugs are considered to
treat the area source more closely to the way of ISC3. This is because the "line-source"
integrator, similar to that used in ISC3 to model area sources, is only implemented when
emissions are simulated as slugs. Puffs use the effective oy treatment for area sources. If there
are receptors within or very near an area source, the slug treatment is a better representation. If
receptors are farther away, the puff treatment is reasonable, and less time-consuming. Mixing
height was fixed at 3000m. These runs were done both for oz(init) = 0 and for oz(init) = 2.5m
(specification of non-zero oz(init) is optional in both models). The best accord was seen for the set
in which oz(init) = 0 (Appendix G). For about one fifth of the cases, \FB\ > 0.10 (FBmin = -0.16).
The maximum residual was 561 |igm"3 (2.2% of the concentration mean at the incident receptor),
while the minimum residual was -1537 |igm"3 (33% of the concentration mean at the incident
receptor). Mean residuals and mean standard deviations among runs ranged over three orders of
magnitude. Analysis of the residuals and fractional biases indicate a definite trend toward
negative bias (CALPUFF relative to ISC3), and best accord for any P-G category was seen for
the higher wind speeds. Also, within any P-G category, the variance falls off with higher wind
speed. The parameter %R indicates accord within two percent across meteorological conditions
and receptors (the value for %Rmin was -1.5% and %^max was -0.07%) and again, the tendency
toward negative bias is indicated. A qualitative inspection of residuals as they appear along the
receptor array indicated no distinct pattern of bias for any case.
3.1.4 Volume Source
The volume source was modeled with emissions simulated as slugs. Because ISC3 does not
compute concentrations for receptors within 2.15oy of the source (it's actually 2.15oy + 1m), no
residuals were analyzed for receptors closer than 200m.
-------
There is a fundamental feature of the way in which ISCST3 treats virtual sources (such as
the volume source in question) that is at odds with the way in which CALPUFF treats such
sources. The phenomenon is described and illustrated in Appendix H. A modified version of
ISCST3 was created to ensure conformity in the treatment of virtual sources by both models.
Once this modification was made, the accord between CALPUFF and ISC3 was quite good
(Appendix I).
For all cases \FB\ = 0.0. The maximum residual was 0.15 |igm"3 (0.1% of the concentration
mean at the incident receptor), while the minimum residual was -0.92 |igm"3 (0.4% of the
concentration mean at the incident receptor). Mean residuals for any run ranged from -0.2 |igm"3
to 0.0 |igm"3, and total range for OR was 0.0 - 0.22 |igm"3. The parameter %R indicates accord
well within a tenth of one percent across all meteorological conditions and receptors (the value
for %Rmin was -0.07% and %Rmax was 0.01%). A slight tendency for negative bias was apparent
for the stable P-G categories. As seen for the area source, for any of the stable P-G categories,
variance falls off with higher wind speed. A qualitative inspection of residuals as they appear
along the receptor array indicated slightly more bias for receptors in the near field of the source.
3.2 Variable Meteorological Conditions
3.2.1 Scenarios for Sensitivity Study
For the sensitivity study comparing CALPUFF and ISC3, meteorological conditions
were allowed to vary hourly. The first test scenario was devised to see what effects variable
meteorology would have on hourly averaged concentrations. One annual period of hourly
averaged meteorological data was selected from each of three climatically different regions of the
United States. The concentrations between CALPUFF (emissions simulated as a continuous
series of puffs) and ISC3 (emission release simulated as a continuous plume) were compared in
time and space. The comparisons were examined to try to find the underlying cause of
significant differences. The second scenario was a rerun of the first case with some
modifications. The averaging times were extended to 3-, 24-hour and annual periods. Maximum
concentrations were compared for individual receptor rings at 15 downwind distances. The suite
of four point sources described in Appendix C was used in these comparisons.
The meteorological data consist of hourly values of wind speed and direction, ambient
temperature, stability class, and mixing heights. The three sites selected were: 1991 Boise,
Idaho; 1990 Medford, Oregon; and 1964 Pittsburgh, Pennsylvania.
-------
The Boise data set was selected because of the very directional nature of it winds (see
1991 Boise wind rose, Appendix J). Over 33% of the winds have a northwesterly component
and over 33% of the winds have a southeasterly component with the majority of those winds
having speeds greater than 2 ms"1. With such persistence in wind direction, the puffs simulated
by CALPUFF would be expected to be transported to the most distant receptors.
The Medford data set was selected because of the high number of calm wind situations
(see 1990 Medford wind rose, Appendix J). In 1990, Medford Oregon recorded a value of 22.5%
of calm winds. This compares to the average of 6.5% for the other two sets of data. Since
CALPUFF processes calm winds and ISC3 "zeros" concentrations during calm wind events,
there is good reason to expect differences to be seen in the simulated patterns of surface
concentration values estimated by the two models.
The 1964 Pittsburgh data set was selected because it has been used as a standard test set
for a number of years and because of its fairly well distributed wind directions and wind speeds
(see 1964 Pittsburgh wind rose, Appendix J). Although there is a bias in the wind direction
toward the southwest, this set was included because many data sets show a similar bias for a
particular wind direction and also have a low number of calm winds.
The receptor placement consisted of 15 rings of 36 receptors each for a total of 540
receptors. The rings were spaced at distances of 0.5, 1, 2, 3, 5, 10, 15, 20, 30, 50, 100, 150, 200,
250, and 300km from the source. On each ring, the receptors were spaced every 10° starting at
360°.
3.2.2 Preliminary Studies
Three preliminary studies were done prior to the sensitivity study. In the first preliminary
study, CALPUFF and ISC3 were run to create a plot of concentration curves under steady state
conditions for centerline and laterally placed receptors. If there were differences in the way
dispersion coefficients were calculated between the two models, that would become apparent in
plots of concentration distributions. In the second preliminary study, the puff and slug models
were run for a two-hour segment where a large wind shift occurred in the second hour. The
purpose was to compare concentration footprints from puff and slug mode results. This study
highlights the different manner in which puffs and slugs are treated in CALPUFF. In the third
preliminary study, a detailed examination was made of the concentration output from CALPUFF
(puff mode) and ISC3 using the Boise meteorological data to help understand the large
differences in concentrations between these two models over a multi-hour period involving calm
winds and a wind shift.
-------
First Study
The first study was done to see whether CALPUFF would calculate oy and oz values
differently than ISC3. The same standard input file was created for ISC3 and CALPUFF with the
idea being that any difference in the way the sigmas were calculated would be evident in the
concentration results. In this comparison, CALPUFF was run in both the puff and slug modes.
The models were run for a 2m point source and the basic switch settings for CALPUFF set per
Appendix A. The meteorological data were kept constant except for P-G stability category. For
each run, the stability category was changed until all six stability categories, A through F, were
used for all three models (i.e., ISC3, CALPUFF puff model, and CALPUFF slug model). A
preliminary group of receptors was created with the centerline along the 360° axis and the
receptors spaced every 1 ° for 44°.
The resulting concentrations were compared on a receptor-by-receptor basis. When the
same input data were used, all three models produced concentrations that were within a few
fractions of a percent of one another (Figure 1). Figure 1 displays six sets of three curves, one
set of curves for each stability category. Each curve in each set overlaps the other curves in that
set. The only common difference in each set is that both CALPUFF curves are truncated. This
can be seen by the extrapolation of the ISC3 thin dashed line after the thick dashed line and thin
solid lines of the two CALPUFF curves. This disparity results from CALPUFF concentration
values set to zero for receptors that are more than 3oy from the centerline (Appendix E), whereas
ISC3 sets concentration values to zero for receptors that are more than 11.75oy from the
centerline. However, lateral plume spread in ISC3 is limited to 50° either side of the centerline
and may be further decreased by vertical mixing conditions.
Second Study
In the second study, CALPUFF's treatment of emissions, puffs versus slugs, was evaluated
(Appendix E), using synthesized meteorological data. There is a general difference in the extent
of the hourly CALPUFF concentration "footprints" using the puff and slug models (Appendix K).
Concentrations produced by the puff model produce a concentration field similar to a
concentration field produced by ISC3 but are restricted to the trajectory algorithms in CALPUFF.
The extent of each CALPUFF downwind concentration field is limited by the average wind
speed occurring over a particular hour. The extent of the downwind concentration field in ISC3
is limited only by the farthest downwind receptor. The extent of the downwind concentration
field when the slug model is used is the same as that for the puff model. However, when the
wind direction changes from one hour and to the next, the directional orientation of the slug is
maintained while the slug is advected downwind (Figure 2). During Hour 1, the wind was from
10
-------
1E+5 -=
1E+4 —
I ' T
0 10 20 30 40
Degree Deviation of Receptor from Centerline Direction
Figure 1. Plot of ISC3 and CALPUFF slug
and puff model concentrations at a distance of
0.5 km from the source for all stability
classes. The curves for CALPUFF are
truncated because CALPUFF and ISC3 use
different y/oy criteria for deciding whether to
compute a concentration for a given receptor.
10-
9-
8-
7-
6-
5-
4-
3-
2-
1-
0-
Hour2 Slug Concentration Footprint
from Hour 1 and 2 Emissions
Source
Hour 1 Slug and Puff Concentration Footprints
-1012
10
Distance (km)
Figure 2. Plot of CALPUFF slug and puff
one-hour concentration footprints during a
2-hour, 70 degree wind shift. Note the broad
Hour 2 slug area which was advected from
the area of the Hour 1 slug footprint. This
area was coupled with the Hour 2 emissions.
20-
10-
-2-10-
-20-
-30-
Centerline at Hour 62
CALPUFF Emissions
from Hours 61-62
CALPUFF Emissions
from Hours 57-60
CALPUFF Emissions
from Hours 52-56 Source
ISC3 Emissions
from Hour 62 Only
Group Centerline Concentrations For Hour 62
10 20 30 40
-40 -30 -20 -10 0
Distance
Figure 3. Plot of concentration footprints at
Hour 62 from three CALPUFF hourly
emission groups and one ISC3 emission
group. The Hour 62 ISC3 plume centerline
orientation is drawn through the source
location. Note the overlap of groups in the 5
to 15km downwind range.
Concentration Contributions
i _ » CALPUFF, Hours 52-56
• — CALPUFF, Hours 57-60
— . CALPUFF, Hours 61-62
CALPUFF Total
ISC3,Hour62 only
Hours 61-62
CALPUFF Total
CALPUFFTo
-40 -20 0 20 40
Centerline Distance Downwind from Source (km)
Figure 4. Plot of Figure 3 CALPUFF group
and total and ISC3 one-hour concentrations at
Hour 62. Note that the CALPUFF total is
approximately 50% greater than ISC3
concentrations at 15km.
11
-------
263 °; during Hour 2, the wind switched to 193 °. In Hour 1, the puff and slug model
concentration footprints are almost exactly the same. However, in the next hour, the east-west
oriented slug is advected north-northeastward. This results in a number of receptors being
impacted but at a much lower concentration of about 280 |igm"3. At the end of Hour 2, an area of
approximately 25 km2 has been impacted by the slug model. The puff impacts are restricted to
relatively narrow corridors.
Note that during Hour 2, the slug andpuff models have been simulating emissions from the
source. The emissions have been transported north-northeastward (Figure 2). The
concentrations produced by both models are similar but the puff concentrations are higher in the
area of the maximum (Appendix K), due to the way in which dispersion is treated in the slug
model. Since the slug is elongated and the mass of effluent is spread evenly throughout its
volume, the newly emitted effluent close to the end of the hour has not had time to be transported
past the receptors farther out. At distances of 0.5, 1, 2, 3 and 5km along the 10 degree radial, the
Hour 2 slug concentrations are 91, 82, 64, 46, and 11 percent of the respective puff
concentrations. While the slug model may have a broader spatial impact, its average
concentrations are generally lower than those of the puff model.
Remember that receptors were placed on rings within the modeled domain and that there
were no rings between 5km and 10km. With this arrangement a truncation appeared in the puff
concentration footprint for Hour 2 beyond 5km from the source, and the actual footprint
(appearing as right side Hour 2 in Figure 2) was not detected. (This truncation was not as evident
for the Hour 2 slug footprint). To address this artifact, a finer Cartesian grid was developed for
the second preliminary study that used a spacing of 400 meters and the right side Hour 2 puff
footprint was then detected and expressed (Fig. 2). Note that the right side Hour 2 puff footprint
originates from the terminus of the Hour 1 puff. The left side Hour 2 puff footprint is the result
of Hour 2 emissions from the source. Also note the 400 meter grid resolution was not fine
enough to properly contour the left side Hour 2 puff concentration isopleths. There was no such
contouring problem evident within the other puff and slug concentration footprints. Note that the
Hour 2 slug footprint is superimposed by the Hour 1 and Hour 2 puff footprints and by the
exposed area between them.
Third Study
In the third study, a detailed examination was done on the concentration output from
CALPUFF (puff mode) and ISC3 using the Boise meteorological data to examine the cause of a
large difference in concentrations between these two models' results. These concentrations
12
-------
occurred 5 to 15km downwind from the source at Hour 62 and after a 4-hour period of calm
winds and then a wind reversal (Figure 3).
During the 10 hours preceding Hour 62, hourly emissions were released into one of three
wind regimes. First, there were 5 hours of east-southeasterly winds, followed by 4 hours of calm
winds, followed by a 180-degree wind shift for 2 hours. Emissions were advected west-
northwesterly, then stagnated but the puffs spread out evenly during this calm wind regime, and
finally all emissions were advected east-southeasterly until Hour 62.
The CALPUFF concentration field at Hour 62 consists of three groups of concentrations
based upon the prevailing wind direction at the time of emission release. One group had releases
during Hours 52 through 56. The next group had releases during the calm wind Hours 57
through 60, and the final group had releases during Hours 61 and 62. The fields were depicted to
show their respective group concentration footprints at Hour 62.
In Figure 3, note that all three groups overlap each other in the 5 to 12km range downwind.
This is also affirmed in Figure 4, which shows the centerline concentrations oriented on the Hour
62 wind direction for each group, the total of the three groups, and the ISC3 centerline
concentrations for the receptors nearest the centerline. The centerline concentrations from the
three groups were added together to produce concentrations a factor of two greater than those
estimated by ISC3 at 15km.
Leading up to Hour 62, there were four hours of calm wind conditions. During calm winds,
CALPUFF assumes that the wind speed is zero. However, unlike ISC3 which treats the calm
hour as missing, CALPUFF increases the sigma values of each puff with respect to time. During
an hour of calm winds, the puffs have grown to the point that ground-level concentrations in this
study were calculated at 0.5 and 1.0 km from the puff centers in all directions for the first hour of
calm. After two hours, the effluent reached as far as 2km. The broadness of the Hour 52-56 and
Hour 57-60 groups is reflective of the puff spreading during the calm wind period.
Details of this type of dispersion phenomena can be seen in Table 1. During Hour 57, the
Hour 57 puff releases penetrated a low mixing height (inversion) and continued to spread
horizontally without any concentrations contacting the ground. During the inversion rise in Hour
58, emissions were then mixed to the ground and Hour 57 emissions impacted receptors 0.5, 1
and 2km distance from the source while Hour 58 emissions were dispersed only to receptors at
0.5 and 1 km distance from the source.
13
-------
3.2.3 Sensitivity Study
One of the major tasks of this study is to understand what types of concentrations will be
produced by CALPUFF with respect to ISC3. The results of ISC3 versus CALPUFF using the
puff and slug models were compared for three different climatological regions of the country.
The results are displayed as a series of figures plotting the percent difference in concentrations at
various downwind distances with only ISC3 results in the denominator. Results consist of
maximum and highest of the second highest percent differences for 1-, 3-, 24-hour and annual
averages.
Table 1
CALPUFF Concentration Estimates under Calm Wind Conditions
Concentrations (|igm"3)
Receptor
X
0.00
0.00
0.00
0.00
0.00
0.09
0.17
0.35
0.52
0.87
Coordinates
Y
0.50
1.00
2.00
3.00
5.00
0.49
0.99
1.97
2.95
4.92
Hour 57
emissions at
Hour 57
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
Hour 57
emissions at
Hour 58
1340.44
822.80
115.98
0.00
0.00
1340.68
822.44
115.86
0.00
0.00
I produced by:
Hour 58
emissions at
Hour 58
2989.86
315.08
0.00
0.00
0.00
2992.90
314.55
0.00
0.00
0.00
As illustrated in Figures 3 and 4, the explanation of why and how one model produces
higher concentrations than another can be complex. The effects of inversions, calm winds, wind
shifts, wind reversals, and plume and puff trajectory differences can all lead to enhanced or
reduced effluent impact. The results of these interactions are shown in Figures 5 and 6 (the
series is continued as Figures L-l through L-7 in Appendix L).
As shown in Figure 5, the Medford plots contain the largest number of positive percentage
differences over the widest range of downwind distances. As was seen in Figures 3 and 4, the
results of calm winds and wind reversals can lead to higher than ISC3 average concentrations at
14
-------
respective downwind distances. With the high percentage of calm winds, note also in the annual
average panel (Figure 5d) how the differences increase dramatically for the higher stacks as the
downwind distance decreases. This is caused by the ISC3 plume not reaching the ground, or not
fully dispersing to the ground, whereas CALPUFF can model effluent dispersion with wind
reversals for receptors near the stack base.
As shown in Figure 6, the overall difference pattern with respect to stack height and
downwind distances among the three sites is remarkably similar. Only the magnitude of the
differences and the downwind distance at which the values initially converge is different. The
Pittsburgh plots tend to slope downward with respect to the others but overall the patterns remain
the same with respect to stack height and downwind distance.
As illustrated in Figures L - 1 through L - 7 (Appendix L), sometimes a pattern or trend can
be seen by comparing subsequent or related figures only to find an exception in another figure.
All of this may be the result of complex interactions that are likely to occur in any of the
climatological regimes. For instance, Medford, Oregon has a high percentage of calm winds. If
these calm wind events are coupled with a wind reversal, the same situation illustrated in Figure
3 for Boise, Idaho can occur. The patterns in the percentage differences may reflect a general
pattern found at that site but the pattern can be overlaid by a situation often found at another site.
15
-------
Q
10 100
Distance Downwind (km)
1000
Figure 5a. Results for 1-hour averages using
1990 Medford data.
Q
Tl I \ \ I I I I I I
10 100
Distance Downwind (km)
1000
Figure 5b. Results for 3-hour averages using
1990 Medford data.
600 -i
-200
10 100
Distance Downwind (km)
1000
T| I |
10 100
Distance Downwind (km)
1000
Figure 5c. Results for 24-hour averages using
1990 Medford data.
Figure 5d. Results for annual averages using
1990 Medford data.
Figure 5. Percent differences (ISC3 vs. CALPUFF slug model) as a function of downwind
distance for the highest 2nd high concentrations; 1-, 3-, 24-, and annual averages.
Data are for Medford, Oregon. Note: % Difference = 100 (—
16
-------
Tl I I I I I I I I I
10 100
Distance Downwind (km)
1000
Figure 6a. Results for annual averages using
1991 Boise data.
600 -i
-200
10 100
Distance Downwind (km)
1000
Figure 6b. Results for annual averages using
1990 Medford data (repeat of Fig. 5d).
600 -i
400 -
TTTTT\ ' I r
10 100
Distance Downwind (km)
1000
Figure 6c. Results for annual averages using
1964 Pittsburgh data.
Figure 6. Percent differences (ISC3 vs. CALPUFF slug model) as a function of downwind
distance for annual averages, all three sites. Note: % Difference = 100 (-
•CALPUFF A/SC3
17
-------
4. Summary and Conclusions
4.1 Steady State Meteorological Conditions
CALPUFF and ISCST3 were run with identical meteorological data sets to compare their
estimates. The meteorological data sets were synthesized to represent a variety of wind speeds in
each of the six P-G stability categories (54 cases in all). ISC3 was run in the "regulatory default"
mode and CALPUFF runs were configured to emulate this mode, which included the simulation
of emission releases as slugs (versus integrated puffs). Receptors were located along a straight
line ("due north") at successively distant intervals. Sources included three elevated point sources
(35m, 100m and 200m). Surface releases included a 2m point source and a rectangular area
source 500m on a side. A typical volume source was also examined. For point sources, model
runs were done for two regimes, one in which the mixing height (Z;) was set to 3000m, and the
other for Z; = 500m. The latter regime was explored to inspect CALPUFF's treatment of
reflection. For each source type, a comparison matrix was created to assess comparison across
all meteorological conditions in terms of a variety of robust statistical indicators.
For all point sources (with Z; = 3000m), the results indicated good accord between the two
dispersion models. For all meteorological conditions, the mean fractional bias across the
receptor radial was well below 10%. Maximum residuals at any receptor were on the order of
0.1% of the concentration mean at the incident receptor. While a qualitative inspection of
residuals as they appear along the receptor array indicated no distinct pattern of bias, a slight
negative bias (CALPUFF relative to ISC3) is apparent for the 2m source, while the reverse is true
for the elevated sources. For the low mixing height regime (Z; = 500m), the comparison results
were strikingly similar, suggesting that both CALPUFF and ISC3 treated reflection identically.
The area source was simulated with mixing height set to 3000m. One set of runs was done
with initial oz set to 0, while the other was set to 2.5m. The best accord was seen for the former
case, but for about one fifth of the cases, the mean fractional bias was greater than 10%. The
maximum absolute residual was 33% of the concentration mean at the incident receptor. There
was an apparent trend toward negative bias (CALPUFF relative to ISC3), but there was
substantial variance as well. Mean residuals and mean standard deviations ranged over three
orders of magnitude. Accord improved (and variance diminished) with higher wind speeds,
which is expected as the slugs are stretched from the point of origin.
With a test version of ISC3 in which the virtual source treatment was "corrected", the
models showed close agreement in their treatment of the volume source. Maximum absolute
residual was well below one percent of respective concentration means at incident receptors. For
all cases, mean fractional bias was zero. A very slight tendency for negative bias was seen for
the stable stability categories, and (as for the area source) variance diminished with higher wind
18
-------
speed. A qualitative inspection of residuals as they appear along the receptor array indicated
slightly more bias for receptors in the near field of the source.
4.2 Variable Meteorological Conditions
To examine differences in model estimates when variable meteorological data are used,
several studies were done. Actual full-year data sets from three climatologically different sites
were used. The sites chosen were Boise, Idaho (1991), Medford, Oregon (1990) and Pittsburgh,
Pennsylvania (1964). Using a synthesized meteorological data set, a preliminary set of studies
was done to examine (1) differences in the way both models treat lateral o's (CALPUFF was run
using both puffs and slugs), and (2) puff versus slug differences within CALPUFF alone.
Another study was done using the Boise data to examine the occurrence and location of
concentration maxima estimated by ISC3 and the CALPUFF puff model. Then for all three sites,
extensive sensitivity studies were done in which estimates by ISC3 were compared to CALPUFF
(puff and slug models). In general, 36 - 45 receptors were placed on each of 15 concentric rings
at successively more distant intervals.
In general, the differences between CALPUFF and ISC3 concentration results are caused by
how emissions are transported and dispersed. CALPUFF limits downwind transport in based on
the wind speed while there is no such limitation in ISC3 (it is a plume model). Under calm wind
conditions, CALPUFF continues to disperse each puff while the ISC3 model is arbitrarily set to
not determine concentrations when the wind speed is less than 1 ms"1. CALPUFF is capable of
tracking the puff emitted before, during and after wind shifts and reversals while ISC3 is only
concerned with the current hour transport of its plume(s). CALPUFF continues to disperse each
puff even when they are above an inversion layer while ISC3 determines its plume is above the
inversion layer and cannot be advected to the ground (e.g,. concentrations = 0.0). When the
inversion rises above the old puffs, they are dispersed to the ground creating impacts for any
nearby receptors.
When all these and other meteorological conditions are recorded on an hourly basis and
form a complete year of meteorological data, the effects on concentrations vary between the
models and from region to region. The meteorologically induced variations in concentrations do
not appear to be so much a regional phenomena, but the variations are related to how the hourly
meteorological conditions occur preceding and during a given averaging period. It is possible to
have 4 or 5 hours of winds in one general direction followed by 4 hours of calm winds, and then
followed by several hours of reversed wind flow. This can occur in any one of the regions.
However, the potential frequency of this occurrence may be higher for one region than another.
Since calm winds have a causal relationship leading to higher concentrations, then a site such as
Medford with a relatively greater incidence of calms (i.e., 22% calm hours versus the other
regions having around 6%) will have higher concentrations associated with CALPUFF.
19
-------
4.3 Conclusion
Even though ISC3 and CALPUFF can be made to produce the same concentrations in a
steady state environment, a variable state environment can produce higher-than-ISC3 ground-
level concentrations with CALPUFF. Climatological characteristics of a region appear to be a
factor, but the accumulation of hour by hour meteorological conditions on the transport of
CALPUFF puffs is the key to understanding the differences that are produced by these two
models. This should come as no surprise as the meteorological assumptions used in formulating
the downwind transport of the ISC3 and CALPUFF effluents and the dispersion from the
respective plumes and puffs are different. This is also compounded by the different treatment of
dispersion during calm wind conditions.
This complex interaction of transport, vertical mixing, and dispersion have an effect on
concentrations with respect to downwind distances in CALPUFF. Occasionally, the
accumulation of mass released over several hours will be transported in such a manner that the
combined effect is to produce sharp localized maxima in simulated concentration values. The
occurrence of such events is not predictable. It seems to occur with greater frequency at
Medford. Calm winds play a part in these events. These maxima seem to occur at most
locations in the receptor network, at all downwind distances. When they occur, they seem to
affect in particular the results for the shorter averaging periods.
Overall trends have been noted in the percentage difference comparisons in simulated
concentration values between CALPUFF and ISC3. For taller point sources, there is a trend
toward higher concentrations being simulated by CALPUFF in comparison to ISC3. For annual
averages, the closer a receptor is to the source and the taller the stack, the greater the chance that
the CALPUFF concentration values will be higher than those simulated by ISC3. At the more
distant downwind receptor rings, the bias changes direction from CALPUFF yielding higher
concentrations, to CALPUFF yielding relatively lower concentrations and sometimes these
concentrations are lower than their respective ISC3 counterpart.
20
-------
5. References
1. Environmental Protection Agency, 1995. User's Guide for the Industrial Source Complex
(ISC3) Dispersion Models. Volumes I - IE. EPA-454/B-95-003a-c.
2. Environmental Protection Agency, 1995. A User's Guide for the CALPUFF Dispersion
Model. EPA-454/B-95-006.
3. Environmental Protection Agency, 1995. SCREENS User's Guide. EPA-454/B-95-004.
21
-------
Appendix A
Switch settings for CALPUFF input file
to emulate ISCS's "Regulatory Default" mode
For these comparisons, CALPUFF was run to emulate ISCS's "regulatory mode" (i.e.,
default). Thus, to ensure equivalence for this emulation, certain of CALPUFF's switches were
set as follows:
METFM = 2 ASCH input file used for input
MSLUG = 1 Puffs emitted as slugs
MDRY = 0 Dry deposition NOT used, unless specified otherwise
MWET = 0 Wet deposition NOT used, unless specified otherwise
MSHEAR = 0 Vertical wind shear NOT modeled
WSCALM = 0.9999 A value of 1 ms"1 for the calm wind speed threshold causes a rounding
problem
AVET = 3 Averaging times for o's is 60 min; oy is adjusted as (AVET/60)0'2
MTRANS = 0 NO transitional plume rise (i.e., final plume rise only)
MDISP = 3 PG dispersion coefficients for RURAL areas, computed using the ISC
multi-segment approximation
MGAUSS = 1 Vertical dispersion used in the near-field is Gaussian
MCHEM = 0 NO chemical treatment used
MROUGH = 0 PG oy and oz NOT adjusted for roughness
MPARTL = 0 No partial plume penetration of elevated inversion
MCTADJ = 1 ISC-type of terrain adjustment
MTIP = 1 Stack tip downwash used
PLXO(6) Default wind speed profile power-law exponents for P-G categories A-F
PTGO(2) Default vertical 0 gradient (Km4) for stable P-G categories E & F
For all applicable sources, CALPUFF employs buoyancy induced dispersion (BID); a feature
enabled in ISC3's regulatory mode. Consistent with ISC3's regulatory default mode, missing data
processing was NOT used.
A- 1
-------
Appendix B
Meteorological conditions for the steady state CALPUFF/ISC3 comparisons1
P-G Wind Speed (ms4)
A 1.0 1.5 2.0 2.5 3.0
B 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0
C 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 8.0 10.0
D 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 8.0 10.0 15.0 20.0
E 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0
F 1.0 1.5 2.0 2.5 3.0 3.5 4.0
'Wind speed is at 10m and values are the same as those used in SCREENS. For each combination of P-G stability category,
comparisons for point sources were made with Zj = 500m and 3000m. For the area and volume source, Zj = 3000m.
B- 1
-------
Appendix C
Characteristics for sources used in the CALPUFF/ISC3 comparisons
Point Sources
Stack height
(m)
2
35
100
200
Area (m2)
250,000
Emission Rate
X,Y location &
base elevation
(m)
0,0,0
0,0,0
0,0,0
0,0,0
Length of side
(m)
500
(gs4):
Release height (m):
Initial oy (m):
Initial oz (m):
Emission
rate
(g^1)
100
100
100
100
Ground-level
Emission
rate
(gs-W2)
0.0004
Volume
1.0
10
50
20
Exit velocity
(ms-1)
10.0
11.7
18.8
26.5
Area Source
Effective
Release
Height (m)
1.0
Source2
Stack
diameter
(m)
0.5
2.4
4.6
5.6
Initial
Oz(m)1
2.5
Temperature
(K)
300
432
416
425
'In one set of comparisons, oz(il]it) was set to zero.
Parameter values taken from Figure 9 of SCREENS User's Guide (Reference 3); buoyancy flux and momentum flux = 0; rural
option.
C-l
-------
Appendix D
Receptor array used in the steady state CALPUFF/ISC3 comparisons
Receptors are aligned along a 360° radial at these distances (m):
1 100 32 4000
2 200 33 4500
3 300 34 5000
4 400 35 5500
5 500 36 6000
6 600 37 6500
7 700 38 7000
8 800 39 7500
9 900 40 8000
10 1000 41 8500
11 1100 42 9000
12 1200 43 9500
13 1300 44 10000
14 1400 45 15000
15 1500 46 20000
16 1600 47 25000
17 1700 48 30000
18 1800 49 35000
19 1900 50 40000
20 2000 51 45000
21 2100 52 50000
22 2200 53 55000
23 2300 54 60000
24 2400 55 65000
25 2500 56 70000
26 2600 57 75000
27 2700 58 80000
28 2800 59 85000
29 2900 60 90000
30 3000 61 95000
31 3500 62 100000
D- 1
-------
Appendix E
Puffs versus Slugs: CALPUFF's Two Simulation Modes
CALPUFF may be operated in one of two modes for simulating emissions: puff or slug. In
the puff mode, a continuous plume is represented as a number of discrete packets of pollutant
material. Most puff models evaluate the contributions of a puff to the concentration at a receptor
by a "snapshot" approach, where each puff is "frozen" at particular time intervals, the
concentration due to the frozen puff at that time is computed, and the puff is then allowed to
move, evolving in size, strength, etc. until the next sampling step. The total concentration at a
receptor is the sum of the contributions of all nearby puffs averaged for all sampling steps within
the basic time step. A traditional drawback of the puff approach has been the need for the release
of many puffs to adequately represent a continuous plume close to the source. Another potential
problem arises if the puffs do not overlap sufficiently, causing concentrations at receptors located
in the gap between puffs at the time of the "snapshot" to be underestimated, while those at the
puff centers are overestimated. One alternative to the problems posed by the "snapshot"
approach is the use of the integrated sampling function (originally implemented in MESOPUFF
II). This technique is available in CALPUFF as the integrated puff approach, and is fully
described in Section 2.1.1 of the CALPUFF User's Guide (Reference 2).
Another approach available in CALPUFF uses a non-circular puff (slug) elongated in the
direction of the wind to eliminate the need for frequent releases of puffs. Thus in the slug model,
the "puffs" consist of Gaussian packets of pollutant material stretched in the along wind
direction. A slug can be visualized as a group of overlapping circular puffs having very small
puff separation distances. Actually, the slug represents the continuous emission of puffs, each
containing an infinitesimal mass. The concentrations near the endpoints of the slug (both inside
and outside of the body of the slug) fall off in such a way that if adjacent slugs are present, the
plume predictions will be reproduced when the contributions of those slugs are included (and this
is with steady state conditions). As with circular puffs, each slug is free to evolve independently
in response to local effects of dispersion, chemical transformation, removal, etc. However,
unlike puffs, the endpoints of adjacent slugs are constrained to remain connected (like country
sausages). This ensures continuity of a simulated plume without the gaps associated with the
puff approach. It should be noted that all receptors lying outside of the slug's ±3oy envelope
during the entire averaging time interval are eliminated from consideration. And for those
receptors remaining, integration time limits are computed such that sampling is not performed
when the receptor is outside of the ±3oy envelope. This technique is available in CALPUFF as
the slug approach, and is fully described in Section 2.1.2 of the CALPUFF User's Guide
(Reference 2).
When initial CALPUFF runs were made for point sources, a disparity was seen between
concentration estimates produced by CALPUFF run in the slug mode versus those produced by
ISC3. This discrepancy was unexpected and the matter was brought to the attention of Earth
Tech (CALPUFF's developer). Earth Tech determined that the reason the slug model in Version
960612 did not reproduce the plume model (ISC3) was due to the computation algorithm for
sigmas. In the 960612 version, the receptor-specific sigma was computed by determining the
E- 1
-------
sigma that the puff would have at the receptor, even if the puff hasn't reached the receptor yet (as
does a plume model). This gave nearly exact reproduction of plume results under steady-state
conditions.
However, under non-steady conditions and very high sigma growth rates (e.g., under P-G
category^), this extrapolation can produce puff impacts prematurely (and hence the causality
effect is compromised somewhat). Therefore, the sigmas were "clipped" at the value at the end
of the slug when the receptor is beyond the end of the slug. This arrangement did reasonably
well for causality effects, but caused some deviation from the plume results under steady state
conditions.
As a result of Earth Tech's investigation of this disparity, an experimental version of
CALPUFF was made available to EPA for the purposes of this comparison, and all analyses were
done with this version. This version compromised between the two solutions described above.
The version only allows concentrations to be computed for receptors that are within 4 ox (where
ox is the horizontal puff dispersion parameter) of a puff centroid. This technique was seen to
perform much better with respect to both treating causality and reproducing plume results, and
will be incorporated in the next model to be released (Joe Scire,pers. comm., December 1997).
E-2
-------
Appendix F
Summary statistics from performance matrix - point sources (Zj = 3000m)
Emissions simulated as:
%^mm (%)
%^max (%)
SLUGS3
-0.04 (C1H200)
0.13(D1H100)
Rjfmilrt (ngm'3) -8.0 (A3H2 @ 100m)b
Rnrmax, (|igm-3) 25.0 (F2p5H2 @ 500m)c
^min (^§m~3) '0-1 (see footnote e)
.Rmax (|igm'3) 0.4 (F2p5H2)
°Rfmin^ (l^gm 3) 0-0 (see footnote e)
oRrma^ (|igm'3) 3.2 (F2p5H2)
# cases FB "out of range" :d
NONE
FBmin 0.0 (see footnote e)
max
" ^-*irmin^
0.02 (D20H100)
-0.18(A1H2@ 100km)
FBirmax, 0.53 (D20H100 @ 800m)
aSee text for explanation of ( C1H200 ), etc.
"This value for R, is 0.03% of x (= XCALPUFF + Xisc3) at this receptor. XCALPUFF = 24836 ngnv3; XiSC3 = 24844 (igm'3
This value for Rj is 0.13% of X at this receptor. XCALPUFF = 19040 (igm'3; Xiscs =19015 (
dThere were 216 distinct cases. The "goal" for this range is: -0.10 < FB <0.10
There is no unique run associated with this value.
F- 1
-------
Appendix G
Summary statistics from performance matrix - area source (emissions simulated as slugs)
Initial oz (m):
%*min (%)
%*max (%)
Rirm^ (Hgm"3)
Rirn^ (Hgm'3)
*min (^m")
^max (^g™3)
0™™™ (Hgm'3)
o*rm^ (Hgm'3)
# cases FB "out of range" :d
^ndn
max
T7R
r J-'ifminl
^ ^ifmaxl
0
-1.5 (F1AREA)
-0.07 (see footnote a)
-1537 (FlAREA@3500m)b
561 (E1AREA@ 100m)c
-548 (F1AREA)
-1.1 (D20AREA)
2.4 (D20AREA)
510 (F1AREA)
10
-0.16 (see footnote a)
-0.02 (see footnote a)
-0.39 (El AREA @ 4000m)
0.05 (Al AREA @ 85km)
2.5
-3.2 (F1AREA)
-0.66 (see footnote a)
-4212 (F1AREA @ 300m)
0.08 (Al AREA @ 85km)
-969 (F1AREA)
-10.5 (D20AREA)
28.3 (D20AREA)
895 (F1AREA)
19
-0.19(F1AREA)
-0.04 (see footnote a)
-0.40 (El AREA @ 4000m)
0.05 (Al AREA @ 85km)
There is no unique run associated with this value.
"This value for Rj is 33% of x (= ICALPUFF + Iisc3) at this receptor. XcALpura- = 3869 ngrn3; XiSC3 = 5406
This value for Rj is 2.2% of X at this receptor. XCALPUFF = 25719 (igm'3; Xiscs = 25158 (igm'3
There are 54 distinct cases. The "goal" for this range is: -0.10 < FB <0.10
G-l
-------
Appendix H
ISCSTS's Treatment of Virtual Sources
For volume sources and point sources subject to building wake dispersion, ISC3 makes use of
a virtual source to simulate an initial plume size. That is, if a source has a finite size at the point of
release, its initial oy and oz are "matched" to a point on the corresponding dispersion curve. Because
these curves prescribe the dispersion parameters as a function of distance (starting with a value of
zero at a downwind distance equal to zero), matching the curve to a source with a non-zero initial
sigma entails shifting the apparent position of the source upwind. This shift is known as the virtual
position of the source. If xv denotes the distance of the virtual location of the source upwind of its
actual location, then the value of the dispersion parameter at some distance (x) downwind of the
source should be evaluated at the modified distance (x + xv).
ISC3 adopts this general method, but modifies its implementation in the following way.
Because the P-G curves for oz are expressed as the function axb, where the parameters a and b
themselves depend on the distance, "the ISC model programs check to ensure that the xv used to
calculate oz at (x + xv) is calculated using coefficients a and b that correspond to the distance
category specified by the quantity (x + xv)." (Vol. n of the ISC3 User's Guide (Section 1.1.5.2,
p. 1-20) with the notation for the virtual distance changed from xz to xv. The term xv is calculated
using Equation 1-36.)
The result of this implementation is that the virtual distance becomes a function of receptor
distance downwind of the source, and in fact xv is reevaluated at each receptor the plume encounters
as it moves downwind. Thus, the computed curve of oz as a function of distance is no longer the
continuous P-G curve. This error is illustrated in the following figures. ISC3 was applied to a
volume source with an initial oz of 5m and 20m, respectively, and concentrations were obtained at
receptors within 1000m, for both P-G stability classes A and F (Figs. H-l to H-4). Using
strategically placed write statements in CALC1.FOR (one of ISCSTS's files), the computed virtual
distances and the corresponding oz values were written to a diagnostic file. These values were then
plotted in the figures below as open squares (virtual distances) and as solid circles (oz) in the figures
below. Figures H-l and H-2 are for the P-G A stability category, while Figures H-3 and H-4 are for
P-GF.
In Figure H-l, the virtual distance begins at 33.76m, and grows in steps corresponding to the
"distance ranges" (Table 1-3 of Vol. n) for this P-G curve to almost 120m. The corresponding oz
values "jump" each time a new virtual distance is used. The same phenomenon can be seen in the
other figures, and the departure (ISC3 oz versus P-G oz) increases with downwind distance. Had
more receptors been placed near each of the transition points, a clear "break" in the oz curve would
have been resolved.
For the purpose of the CALPUFF/ISC3 comparison, ISCST3 was re-configured so that a single
value of the virtual distance is computed as a joint function of P-G category and initial oz, with due
regard for the distance ranges imposed on selecting a & b. This single value is then added to all
receptor distances, and the corresponding value for oz computed. Figure H-l indicates the resulting
contour (depicted with open triangles) and suggests the continuous P-G curve for stability class A,
for a virtual location 33.76m upwind of x = Om.
A version of ISC3 with a corrected virtual source algorithm (dated 97363), as was used in this
comparison, was released in January 1998 and uploaded to EPA's SCRAM web site for public use.
H- 1
-------
Stability A; Initial CTz = 5m
100000
10000 -
CO
01
o.
•-C
I
1000 -
100 -
10
- Current Sz
- Corrected Sz
-Xz
- 100.00
- 80.00
120.00
- 40.00
20.00
10
100 1000
Distance Downwind (m)
10000
Figure H-l. Profile of az with distance (m); P-G A and az(init) = 5m
H-2
-------
Stability A; Initial CTz = 20m
100000
10000 -
CO
a
o
01
o.
1000 -
100 -
10
- Current Sz
- Corrected Sz
-Xz
- 220.00
- 200.00
240.00
- 180.00 >
- 160.00
140.00
10
100 1000
Distance Downwind (m)
10000
Figure H-2. Profile of az with distance (m); P-G A and oz(init) = 20m
H-
-------
Stability F; Initial Oz = 5m
40.00
30.00 -
CO
« 20.00 -
01
10.00 -
- 250.00
- 200.00
- Current Sz
- Corrected Sz
-Xz
300.00
a
a
a
3
• 150.00 >
- 100.00
50.00
10
100 1000
Distance Downwind (m)
10000
Figure H-3. Profile of az with distance (m); P-G F and oz(init) = 5m
H-4
-------
40.00
36.00 -
8 32.00 -
CO
8.
24.00 -
20.00
10
Stability F; Initial Oz = 20m
- Current Sz
- Corrected Sz
-Xz
- 1750.00
- 1700.00
1800.00
a
8
Q
"a
- 1650.00
- 1600.00
1550.00
100 1000
Distance Downwind (m)
10000
Figure H-4. Profile of az with distance (m); P-G F and oz(init) = 20m
H-5
-------
Appendix I
Summary statistics from performance matrix - volume source
Emissions simulated as:
%Rmm (%)
%^max (%)
SLUGS
-0.07 (see footnote a)
0.01 (see footnote a)
RH^ (|igm-3) -0.92 (Fl VOL @200m)b
R^rmax^ (jigm'3) 0.15 (C 1 VOL @ 200m)c
,Rmin (|igm-3) -0.2 (F1VOL)
^max (l^g111 3) 0-0 (see footnote a)
°Rfmi^ (l^gm"3) 0-0 (see footnote a)
vKlm^ (l^gm'3) 0.22(F1VOL)
# cases FB "out of range" :d
NONE
FBmjn 0.0 (see footnote a)
max
0.0 (see footnote a)
FBirmin, -0.18 (A1VOL @ 100km)
FBirmax, 0.06 (Al VOL @ 85km)
There is no unique run associated with this value.
"This value for Rj is 0.4% of x (= ICALPUFF + IISCB) at this receptor. XCALPUFF = 238.6 (igm'3; XiSC3 = 239.5 (igm'3
This value for Rj is 0.1% of % at this receptor. XCALPUFF = 135.1 |igm"3; Xiscs = 135.0 |igm"3
There were 54 distinct cases. The "goal" for this range is: -0.10 < FB <0.10
I- 1
-------
Appendix J
Wind Rose Patterns
Bo i se, ID 1991
WIND SPEED (KNOTS)
NOTE; Frequenc i e.
* [ nd Is blow i rig.
OR 1 990
>• 1 -December 3 i ; Mi dn i
-------
Appendix J, continued
P i -t-tsb>ur- \ -gH-t - 1 1 PM
WIND SPEED (KNOTS)
CALM WINDS 6. 7AK
NOTE: Fi-e.qu«r-,c I as
Indicate d I t-act [ t
Fr-orn wh 1 cH tHa
w I nd f s b. I ow I ng.
J-2
-------
Appendix K
CALPUFF Concentrations Estimated by Integrated Puff and Slug Model
Coordinates
X
0
0
0
0
0
0.09
0.17
0.35
0.52
0.87
0.17
0.34
0.68
3.19
1.71
0.25
0.5
1
1.5
2.5
0.32
0.64
1.29
1.93
3.21
0.38
0.77
1.53
2.3
3.83
0.43
0.87
1.73
2.6
4.33
0.47
0.94
1.88
2.82
4.7
0.49
0.99
1.97
2.95
4.92
0.5
Y
0.5
1
2
3
5
0.49
0.99
1.97
2.95
4.92
0.47
0.94
1.88
2.82
4.7
0.43
0.87
1.73
2.6
4.33
0.38
0.77
1.53
2.3
3.83
0.32
0.64
1.29
1.93
3.21
0.25
0.5
1
1.5
2.5
0.17
0.34
0.68
1.03
1.71
0.09
0.17
0.35
0.52
0.87
0
Hour 1
Puff
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
4445.83
4593.7
2440.53
1413.91
678.28
0
Hour 2
Puff
0
0
0
0
0
4400.88
4606.96
2366.87
1408.23
670.78
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
Slug
0
0
0
0
0
4011.7
3785.43
1525.9
658.51
74.94
75.18
211.99
284.43
271.96
233.71
93.83
234.87
284.04
266.32
229.02
108.62
248.26
282.78
262.57
226.52
118.39
254.07
261.61
261.61
226.13
122.47
255.92
281.55
261.75
227.49
120.31
254.23
282.08
263.61
230.65
103.69
234.05
269.69
254.14
226.08
0
K- 1
-------
Puffvs. ISC3
Slug vs. ISC3
10 100
Distance Downwind (km)
Figure L-l(a). Boise meteorological data.
-200
T| I I |
10 100 1000
Distance Downwind (km)
Figure L-l(b). Boise meteorological data.
T| I |
10 100
Distance Downwind (km)
-200
10 100
Distance Downwind (km)
Figure L-l(c). Medford meteorological data.
Figure L-l(d). Medford meteorological Data.
T| I I |
10 100 1000
Distance Downwind (km)
10 100
Distance Downwind (km)
Figure L-l(e). Pittsburgh meteorological data.
Figure L-l(f). Pittsburgh meteorological data.
Figure L-l. Maximum 1-hour average concentrations by distance. Figures a, c, & e show
CALPUFF puffs, whereas figures b, d, & f show slugs.
Note: % Difference = 100 ( "
L- 1
-------
Puffvs. ISC3
Slug vs. ISC3
10 100
Distance Downwind (km)
Figure L-2(a). Boise meteorological data.
-200
T| I I |
10 100 1000
Distance Downwind (km)
Figure L-2(b). Boise meteorological data.
T| I I |
10 100
Distance Downwind (km)
10 100
Distance Downwind (km)
Figure L-2(c). Medford Meteorological Data.
Figure L-2(d). Medford meteorological Data.
T| I |
10 100
Distance Downwind (km)
10 100
Distance Downwind (km)
Figure L-2(e). Pittsburgh meteorological data.
Figure L-2(f). Pittsburgh meteorological data.
Figure L-2. Maximum 3-hour average concentrations by distance. Figures a, c, & e show
CALPUFF puffs, whereas figures b, d, & f show slugs.
Note: % Difference = 100 "
L-2
-------
Puff vs. ISC3
Slug vs. ISC3
I
Q 20°
10 100
Distance Downwind (km)
Figure L-3(a). Boise meteorological data.
400 -
-200
T| I I I
10 100
Distance Downwind (km)
Figure L-3(b). Boise meteorological data.
10 100
Distance Downwind (km)
T| I I |
10 100
Distance Downwind (km)
Figure L-3(c). Medford meteorological data.
Figure L-3(d). Medford meteorological data.
-200
10 100
Distance Downwind (km)
T| I I |
10 100 1000
Distance Downwind (km)
Figure L-3(e). Pittsburgh meteorological data.
Figure L-3(f). Pittsburgh meteorological data.
Figure L-3. Maximum 24-hour average concentrations by distance. Figures a, c, & e show
CALPUFF puffs, whereas figures b, d, & f show slugs.
Note: % Difference = 100 (
L-3
-------
Puff vs. ISC3
Slug vs. ISC3
10 100
Distance Downwind (km)
Figure L-4(a). Boise meteorological data.
-200
T| I I |
10 100 1000
Distance Downwind (km)
Figure L-4(b). Boise meteorological data.
T| I I |
10 100
Distance Downwind (km)
-200
10 100
Distance Downwind (km)
Figure L-4(c). Medford meteorological data.
Figure L-4(d). Medford meteorological data.
400 —
S3
JS
3
a
-200
I '' I r
10 100
Distance Downwind (km)
10 100
Distance Downwind (km)
Figure L-4(e). Pittsburgh meteorological data.
Figure L-4(f). Pittsburgh meteorological data.
Figure L-4. Maximum annual average concentrations by distance. Figures a, c, & e show
CALPUFF puffs, whereas figures b, d, & f show slugs.
. n/ T^--r-r mn ," ~
. % Difference = 100 (
L-4
-------
Puff vs. ISC3
Slug vs. ISC3
10 100
Distance Downwind (km)
Figure L-5(a). Boise meteorological data.
-200
T| I I |
10 100 1000
Distance Downwind (km)
Figure L-5(b). Boise meteorological data.
T| I |
10 100
Distance Downwind (km)
-200
10 100
Distance Downwind (km)
Figure L-5(c). Medford meteorological data.
Figure L-5(d). Medford meteorological data.
T| I |
10 100
Distance Downwind (km)
10 100
Distance Downwind (km)
Figure L-5(e). Pittsburgh meteorological data.
Figure L-5(f). Pittsburgh meteorological data.
Figure L-5. Highest of the second highest 1-hour average concentrations by distance. Figures
a, c, & e show CALPUFF puffs, whereas figures b, d, & f show slugs.
Note: % Difference = 100 ( CALPUFF Xscs^
L-5
-------
Puff vs. ISC3
SlugvsISCl
10 100
Distance Downwind (km)
Figure L-6(a). Boise meteorological data.
400 —
-200
T| I I I
10 100 1000
Distance Downwind (km)
Figure L-6(b). Boise meteorological data.
10 100
Distance Downwind (km)
-200
10 100
Distance Downwind (km)
Figure L-6(c). Medford meteorological data.
Figure L-6(d). Medford meteorological data.
10 100
Distance Downwind (km)
T| I I |
10 100
Distance Downwind (km)
Figure L-6(e). Pittsburgh meteorological data.
Figure L-6(f). Pittsburgh meteorological data.
Figure L-6. Highest of the second highest 3-hour average concentrations by distance. Figures
a, c, & e show CALPUFF puffs, whereas figures b, d, & f show slugs.
Note: % Difference = 100 (Xc^^ " XBC^
L-6
-------
Puff vs. ISC3
Slug vs. ISC3
10 100
Distance Downwind (km)
Figure L-7(a). Boise meteorological data.
-200
T| I I |
10 100 1000
Distance Downwind (km)
Figure L-7(b). Boise meteorological data.
-200
10 100
Distance Downwind (km)
T| I I |
10 100
Distance Downwind (km)
Figure L-7(c). Medford meteorological data.
Figure L-7(d). Medford meteorological data.
400 —
-200
10 100
Distance Downwind (km)
10 100 1000
Distance Downwind (km)
Figure L-7(e). Pittsburgh meteorological data.
Figure L-7(f). Pittsburgh meteorological data.
Figure L-7. Highest of the second highest 24-hour average concentrations by distance.
Figures a, c, & e show CALPUFF puffs, whereas figures b, d, & f show slugs.
Note: % Difference = 100 I
L-7
------- |