United States
Environmental Protection
Agency
Office of Air Quality
Planning and Standards
Research Triangle Park, NC 27711
EPA-454/R-98-019
December 1998
AIR
& EPA
INTERAGENCY WORKGROUP
ON AIR QUALITY MODELING (IWAQM)
PHASE 2 SUMMARY REPORT AND
RECOMMENDATIONS FOR MODELING LONG
RANGE TRANSPORT IMPACTS
-------
EPA-454/R-98-019
INTERAGENCY WORKGROUP ON AIR
QUALITY MODELING (IWAQM)
PHASE 2 SUMMARY REPORT AND
RECOMMENDATIONS FOR MODELING
LONG-RANGE TRANSPORT IMPACTS
U.S. Environmental Protection Agency
Air Quality Modeling Group (MD-14)
Research Triangle Park, North Carolina 27711
National Park Service
Air Resources Division
Denver, Colorado 80225
USDA Forest Service
Air Program
Fort Collins, Colorado 80526
U.S. Fish and Wildlife Service
Air Quality Branch Denver, Colorado 80225
December, 1998
-------
NOTICE
The information in this document has been reviewed in its entirety by the U.S.
Environmental Protection Agency (EPA), and approved for publication as an EPA
document. Mention of trade names, products, or services does not convey, and should
not be interpreted as conveying official EPA approval, endorsement, or
recommendation.
-------
PREFACE
The Interagency Workgroup on Air Quality Modeling (IWAQM) was formed to
provide a focus for development of technically sound recommendations regarding
assessment of air pollutant source impacts on Federal Class I and Wilderness areas.
Meetings were held with personnel from interested Federal agencies, viz. the
Environmental Protection Agency, the U.S. Forest Service, the National Park Service,
and the U.S. Fish and Wildlife Service. The purpose of these meetings was to review
respective modeling programs, to develop an organizational framework, and to
formulate reasonable objectives and plans that could be presented to management for
support and commitment. The members prepared a memorandum of understanding
(MOU) that incorporated the goals and objectives of the workgroup and obtained
signatures of management officials in each participating agency. Even though no
States are signatories, they did participate in IWAQM functions.
This document is being released as a publication of the Environmental
Protection Agency (EPA) in response to a request from the members of IWAQM.
Members of the workgroup include representatives from the Environmental Protection
Agency, the U.S. Forest Service, the National Park Service, and the U.S. Fish and
Wildlife Service. The document includes IWAQM's recommendations for modeling
methods that might be used to estimate Prevention of Significant Deterioration air
quality impacts and National Ambient Air Quality Standards (NAAQS) air quality impacts
associated with long-range transport of pollutant emissions to Class I and Wilderness
areas . The IWAQM recommends that the CALPUFF Lagrangian puff dispersion
modeling system be used for characterization of the transport and dispersion.
The recommendations of IWAQM contained in this document is considered
technical guidance tailored for use in assessing air quality impacts associated with
prevention of significant deterioration.
-------
ACKNOWLEDGMENTS
The members of IWAQM acknowledge the special efforts by Mark Scruggs, John
Notar and John Vimont of the National Park Service; Alan Cimorelli of the U.S.
Environmental Protection Agency (EPA); John Irwin of the National Oceanic
Atmospheric Administration (NOAA) Air Resources Laboratory; Richard Fisher and Bob
Bachman of the U.S. Forest Service; Elwyn Rolofson of the U.S. Fish and Wildlife
Service; Patrick Hanrahan of the State of Oregon, Department of Environmental
Quality; and Kenneth McBee of the Commonwealth of Virginia, Department of Air
Pollution Control for their input and suggestions on assembling this document and their
subsequent review. In compiling the reviews of investigations completed that involved
in some manner the CALMET/CALPUFF modeling system, various authors graciously
provided access to their results and data, and have permitted IWAQM to cite directly
from their reports and papers. The IWAQM gives special recognition and thanks to:
Andrew Gray of SAI, Inc, John Sherwell of the Maryland Department of Natural
Resources, 0. Russ Bullock, Jr., Frank Binkowski, and Robin Dennis of the NOAA Air
Resources Laboratory, David Strimaitis and Joesph Scire of Earth Tech Inc., Jim
Paumier of Pacific Environmental Services Inc., Peter Eckhoff and C. Thomas Coulter
of the EPA Office of Air Quality Planning and Standards.
-------
TABLE OF CONTENTS
1.0 INTRODUCTION 1
2.0 MODELING RECOMMENDATIONS 6
2.1 Screening Analysis 6
2.2 Refined Analysis 9
2.2.1 Meteorology 9
2.2.2 Chemistry 14
2.2.3 Dispersion 15
2.3 Practicalities 16
2.3.1 Screening procedure uncertaintites 17
2.3.2 Refined modeling uncertainties 17
2.3.3 Secondary pollutant uncertainties 18
2.2.4 Technical oversight and review 19
3.0 TRANSFORMATIONS, VISIBILITY, AND DEPOSITION 20
3.1 Chemical Transformations 20
3.2 Visibility Analysis 22
3.3 Deposition Calculations 30
3.4 Assessing Air Quality Related Values - Background 31
4.0 STUDIES AND FINDINGS 33
4.1 MESOPUFF II Implementation Assessment 33
4.2 Revisions to CALMET and CALPUFF 44
4.3 Trajectory Comparisons 48
4.4 Constructing FDDA-MM Data Sets Assessment 54
4.5 Regional Approach 56
4.5.1 An Approach to Implementing a Class I Area Assessments ... 57
4.6 Comparisons of CALPUFF with Tracer Field Data 58
4.6.1 1975 Savannah River Laboratory Tracer Study 58
4.6.2 1977 Idaho Falls Tracer Study 62
4.6.3 1980 Great Plains Tracer Study 67
4.6.4 1992 Project MOHAVE Tracer Study 73
4.7 Comparisons of CALPUFF With ISC3 76
4.8 CALPUFF SCREEN 83
4.8.1 Screening methodology 83
4.8.2 Year to year variability 87
4.8.3 S02 concentrations 88
4.8.4 S04= concentrations 90
4.8.5 S02 deposition 91
4.8.6 S04= deposition 93
4.8.7 Old screen versus new screen estimates 94
4.8.8 Findings and conclusions 96
iv
-------
4.9 CALMET/CALPUFF Enhancements 97
4.9.1 Use of FDDA-MM data with CALMET 97
4.9.2 Use of CALMET to Develop Wind Fields 99
4.9.3 Kincaid SF6 and Lovett S02 Comparisons 104
5.0 CONCLUSIONS 111
6.0 REFERENCES 113
APPENDIX A. CALMET RECOMMENDATIONS A-1
APPENDIX B. CALPUFF RECOMMENDATIONS B-1
APPENDIX C. COMPACT DISK DATA RESOURCES C-1
APPENDIX D. SIXTH MODELING CONFERENCE D-1
-------
LIST OF FIGURES
Figure 1. Hygroscopic relative humidity adjustment factor f(RH) 27
Figure 2. Display of MESOPUFF II computational domain 35
Figure 3. Display of ring source analysis computational domain 40
Figure 4. Highest monthly concentrations 42
Figure 5. 54-km and 18-km MM4 domains 46
Figure 6. Trajectories at 10 m on the 54 km grid 49
Figure 7. Summary of CAPTEX trajectory results 53
Figure 8. Savannah River Laboratory field experiment site 59
Figure 9. Savannah River Laboratory December 10, 1976 results 61
Figure 10. Idaho Falls field experiment site 63
Figure 11. Summary of Idaho Falls April 19, 1977 results 66
Figure 12. Great Plains field experiment site 68
Figure 13. Summary of Great Plains 100 and 600 km July 8, 1980 results 71
Figure 14. Summary of Great Plains 100 km July 11,1980 results 72
Figure 15. Project MOHAVE experimental site 74
Figure 16. Project MOHAVE results for Las Vegas Wash site 75
Figure 17. Summary for Hour 62 of a CALPUFF simulation 79
Figure 18. CALPUFF and ISC comparisons for Medford, Oregon 81
Figure 19. Annual average CALPUFF and ISC comparisons 82
Figure 20. Comparison of screening estimates of ambient sulfate concentrations .. 95
Figure 21. Comparison of screening estimates of total sulfur deposition 96
Figure 22. CALMET computational domain 98
Figure 23. Terrain contours for the Wenachee, WA domain 101
Figure 24. Down slope flow components for 1:00 AM July 1, 1994 102
Figure 25. Upslope flow components for 3:00 PM July 1, 1994 103
Figure 26. Q-Q plots for Kincaid comparing CALPUFF with ISC 107
Figure 27. Scatter plot for Kincaid comparing CALPUFF with ISC 108
Figure 28. Q-Q plots for Lovett comparing CALPUFF with CTDMPLUS 109
VI
-------
LIST OF TABLES
Table 1. Outline of recommendations for screening analysis 8
Table 2. Outline of recommendatinos for refined analysis 10
Table 3. Sources modeled for demonstration analysis 36
Table 4. PSD and AQRV demonstration analysis results 37
Table 5. Source characteristics for ring source analysis 39
Table 6. Source ring characteristics 41
Table 7. Summary of statistical comparisons of wind fields 47
Table 8. Characteristics for point sources used in CALPUFF/ISC comparisons .... 78
Table 9. Year to year variability of S02 concentrations 88
Table 10. Comparison of CALPUFF(ISC/CALMET) S02 concentrations 89
Table 11. Comparison of CALPUFFJISC/CALMET) sulfate concentrations 91
Table 12. Comparison of CALPUFF(ISC/CALMET) S02 deposition 92
Table 13. Comparison of CALPUFFJISC/CALMET) sulfate deposition 93
VII
-------
1.0 INTRODUCTION
Special protection from adverse air quality impacts is afforded certain national
parks and wilderness areas, through the prevention of significant deterioration (PSD)
program (U.S. EPA, 1980). These areas have been designated as Class I areas, and
as such, increases of pollutant levels in these areas are strictly limited. Furthermore,
the Federal Land Manager (FLM) of the Class I area is given an affirmative
responsibility to ensure that Air Quality Related Values (AQRVs) are not adversely
impacted. [The FLMs of the Class I areas are the U.S. Forest Service (USFS), the
National Park Service (NPS), and the U.S. Fish and Wildlife Service (FWS).] Air quality
models are one of the primary tools used to assess the impacts from sources of air
pollution on both the established Prevention of Significant Deterioration (PSD)
increments and the AQRVs. Steady-state models have been generally used for PSD
analyses. As the PSD program has developed, the need for more sophisticated models
to assess air quality impacts in Class I areas, from sources at relatively greater
distances from the Class I areas, has arisen. In some areas, the FLMs have asserted
that Class I areas have been adversely affected by air pollution and that new sources of
pollution over a broad area are further harming the resource. The absence of any
recommended long-range modeling techniques has left permitting authorities without
the means to assess the assertions of the FLMs. The Environmental Protection
Agency (EPA) and the FLMs have undertaken various model development efforts to
address the air quality impacts of pollution transported over relatively long distances.
The Interagency Workgroup for Air Quality Modeling (IWAQM) was formed to
coordinate the independent modeling efforts of the EPA and the FLMs so that a
consistent, technically credible approach can be recommended and used.
Models used to evaluate the impact of sources of air pollution on the PSD
increments and National Ambient Air Quality Standards (NAAQS), are required to follow
Appendix W to 40 CFR Part 51 (Guideline on Air Quality Models), U.S. EPA (1997).
(Hereafter, referred to as the Guideline.) For many situations, preferred models,
considered generally applicable under a variety of circumstances, are defined.
Currently within the Guideline (Section 7.2.6), there is no preferred model listed for
assessing impacts involving transport beyond 50 km. For those situations for which
there are no preferred models, criteria are established in the Guideline to use
appropriate methods on a case-by-case basis. These criteria are:
i. The model can be demonstrated to be applicable to the problem on
a theoretical basis, and
ii. the data bases which are necessary to perform the analysis are
available and adequate, and
-------
iii. performance evaluations of the model in similar circumstances
have shown that the model is not biased toward underestimates, or
iv. after consultation with the EPA Regional Office, a second model is
selected as a baseline or reference point for performance and the
interim procedures are then used to demonstrate that the proposed
model performs better than the reference (preferred) model.
The processes which become important in the transport of pollution over long
distances include the spatial and temporal variability of the winds which transport and
disperse air pollutants in the presence of various terrain and water features, the
chemical transformation of the pollutants as they travel, and the deposition of the
pollutants along the way. Cummulative impact assessments likely will need to address
multiple, geographically disperse sources. There are existing long-range transport
models available which meet some, but not, all of these needs and some which meet
these needs but either have not been sufficiently tested or are perceived to need further
development for routine operational use in regulatory assessment analyses. One of the
primary goals of IWAQM is to evaluate existing modeling codes and either recommend
one as an accepted approach or combine the better elements of several of the existing
codes, creating a new modeling construct.
The IWAQM work plan (U.S. EPA, 1992a) describes a phased approach to
satisfy the modeling needs described above. Phase 1 consists of reviewing EPA
guidance and recommending an interim modeling approach to meet the immediate
need for a long-range transport model for ongoing permitting activity. In developing a
Phase 2 recommendation, the workgroup was to review other available operational
models and make a recommendation of the most appropriate modeling techniques.
The Phase 2 recommendation was to be a compromise between the current modeling
state-of-science and best available operational computer capabilities. If resources
could be found to pursue a Phase 3 recommendation, the workgroup would add more
advanced modeling techniques to its consideration, probably representing a greater
level of scientific and computer hardware sophistication.
Given the practical limitations of resources and hardware, the Phase 1 interim
recommendations (U.S. EPA, 1993) were designed to provide the best approach from
existing "off-the-shelf" techniques. Two models were assessed, the MESOPUFF II
model (U.S. EPA, 1994) and the Acid Rain Mountain Mesoscale Model (ARMS), Morris
et al., (1988). Upon careful examination of both models, coding errors were discovered
in the ARMS, which potentially invalidated its previous evaluations. With this in mind
and other considerations as discussed in U.S. EPA (1993), the Phase 1
recommendation was to use on a case-by-case basis the Lagrangian puff model,
MESOPUFF II, to evaluate the impacts of pollutants from sources located more than 50
kilometers from Class I areas, up to several hundred kilometers from Class I areas.
The impacts are: 1) the allowable Class I increases (increments), 2) the National
Ambient Air Quality Standards (NAAQS), and 3) Air Quality Related Values (AQRVs)
-------
associated with emissions of sulfur, nitrogen and particulate matter. The PSD NAAQS
pollutants were sulfur dioxide (S02) and particulate of size less than or equal to ten
microns (PMio). AQRV impacts include such effects as visibility degradation and acidic
deposition. In order to focus this effort, increase the chance for success in at least
certain scenarios, and to meet an articulated immediate need of permitting authorities
and the EPA, IWAQM chose to address only sulfur and nitrogen derived pollutants. It
was acknowledged that there are other pollutants such as photochemical oxidants that
may injure components of the natural ecosystem, but assessment by IWAQM of the
modeling needs and development of oxidant effects modules was postponed until a
later date. The Phase I interim recommendations also provided a screening
methodology, which (as is discussed later) proved to be too concervative to be of much
use.
For the Phase 1 recommendation, MESOPUFF II was deemed suitable for
conducting single source impact analyses, and in some circumstances cumulative
impact analyses. As the dispersion characterizations in MESOPUFF II were not
designed to handle local-scale dispersion effects, it was recognized that the
MESOPUFF II results would frequently need to be combined with the results from other
modeling techniques used to estimate concentrations from sources closer than 50
kilometers to a receptor area. The Phase 1 recommendation was structured to satisfy
case-by-case Guideline criteria i, ii, and iii above. The iv criterion has meaning only
when there is a preferred model, and such is not the case for modeling impacts
involving transport beyond 50 kilometers.
By restricting the models considered for Phase 1 to "off-the-shelf" techniques,
IWAQM recognized certain limitations. These include a lack of consideration of the
effects of terrain on the long-range transport and dispersion, an underestimation of the
conversion of sulfur dioxide, S02 to sulfate, SO^, when polluted air interacts with
clouds, and a possible overestimation of particulate nitrate when a limited number of
sources are considered. Nonetheless, IWAQM considered the techniques, suggested
to be a significant improvement to those previously used, in that previous techniques
ignored many of the processes important to the assessment of air quality impacts in
Class I areas. Thus, while under some circumstances the impacts on regional visibility
may be underestimated, because the concentrations of sulfates in the atmosphere may
be underestimated due to the inability of the model to treat in-cloud processes, IWAQM,
including the representatives of the land management agencies, considered the
suggested techniques to be technically superior to simply assuming that there are no
impacts on regional visibility.
The IWAQM recognized that there were certain risks involved with
recommending an interim modeling approach. From a regulatory perspective, it is
generally desirable to use an interim model which will yield somewhat higher impact
calculations than a more refined, preferred approach. In the case of steady-state air
quality models, this can be relatively easily ensured because of the independence of
the concentration calculations from one hour to the next. In the case of the Lagrangian
-------
long-range transport models under consideration here, the concentration calculations
for a given hour will be explicitly dependent on the spatially and temporally varying wind
field from that hour and previous hours. Therefore, the exact behavior of a given
modeling system relative to a similar but different modeling system can not be predicted
with certainty.
Not long after the release of the Phase 1 recommendation, the EPA sponsored
the Sixth Modeling Conference which was held August 9-10, 1995 in Washington, D.C.
One of the main topics at this two-day event was a review of IWAQM Phase 1
recommendation, a summary of work in progress, with review comments provided by
several groups. At the conference, IWAQM presented long-range trajectory
comparison, that suggested that use of mesoscale meteorological analyses of wind
fields provided a significant improvement in the accord of modeled and observed
trajectories. The IWAQM endorsed specifically the use of mesoscale meteorological
analyses that employ data assimilation. The IWAQM recommended that the Phase 1
recommendation to use the MESOPUFF II modeling system be replaced with a
recommendation to use the CALMET/CALPUFF modeling system. This was a
relatively new Lagrangian puff modeling system, which had additional algorithms to
provide simulation of local-scale short-range dispersion using methods already
endorsed by the EPA. Thus, use of this newer modeling system allowed one model to
be used for all sources in an analysis, regardless of the transport distance involved.
The IWAQM also endorsed the formation of public/private committees to manage the
myriad of site specific technical decisions that are inherent in assessing mesoscale
PSD and AQRV impacts on Class I areas. Appendix D provides a review of the
information provided at the Sixth Modeling Conference, a summary of comments
received, and IWAQM's response to these comments.
Section 2 presents the Phase 2 modeling recommendation, which represents a
compromise between the current modeling state-of-science and best available
operational computer capabilities. The IWAQM Phase 2 recommendations are for
methods and procedures that might be used to estimate air quality Prevention of
Significant Deterioration impacts to increments and AQRVs and National Ambient Air
Quality Standards (NAAQS) impacts from pollutant emissions, that due to the transport
distances and location are best treated using a long-range transport modeling system.
The Phase 2 recommendations are not limited to recommendations on which modeling
systems to use, as part of IWAQM's efforts are to assist in fostering best assessment of
impacts. Briefly summarized, the Phase 2 recommendation consist of:
• a suggested screening technique for modeling worst-case estimates of long-
range transport impacts,
• a recommendation that the CALPUFF modeling system be listed as the
preferred model in Section 7.2.6 in the Guideline,
-------
• recommendations for assessing PSD increment and NAAQS impacts of a new
source's emissions (or change in an existent source's emissions) involving long-
range transport of emissions,
• recommendations for assessing the AQRV impacts of a new source's emissions
(or change in an existent source's emissions) of sulfur, nitrogen and particulate
matter to regional visibility degradation and deposition, which requires a
knowledge of the current state of the Class I area, and
• a recommendation that consideration be given to forming committees ('Regional
Approach') to assist in the resolution of the myriad of decisions associated with
mesoscale modeling of PSD and AQRV impacts for each Class I area (or groups
of areas).
Section 3 discusses the chemistry limitations inherent in CALPUFF, and how
results from CALPUFF can be used to estimate regional visibility impacts and
deposition of sulfur and nitrogen.
From the comments received at the Sixth Modeling Conference, IWAQM
concluded that in order to provide a firmer basis for a Phase 2 recommendation, more
information was needed comparing mesoscale modeling dispersion results with tracer
field data, and more comparisons were needed to stress test code modifications made
to incorporate local-scale dispersion characterizations. Everyone was in agreement
that the screening analysis, suggested by IWAQM in the Phase 1 recommendation,
was inadequate. It was perceived as providing such large overestimates of S02 and
sulfate impacts that it was of little use. Section 4 provides a review of the activities that
IWAQM has sponsored in developing the Phase 2 recommendations, as well as,
summaries of investigations that IWAQM is aware of that provide information regarding
lessons learned in using and applying the CALPUFF modeling system.
-------
2.0 MODELING RECOMMENDATIONS
For most of the modeling situations discussed in the Guideline where a refined
modeling technique is recommended, a screening analysis is also provided. The
screening analysis is meant to be easy to conduct and to provide a worst-case
maximum impact estimate. If the results of the screening analysis show compliance
with existing regulatory requirements, then no further modeling for compliance with
standards and increments is required.
Basically, IWAQM's recommendations for a screening analysis is an approach of
using a simplified set of meteorology with CALPUFF. To encourage the results to be
higher than would be estimated using a fully developed CALMET and CALPUFF
analysis, rings of receptors are used. The maximum concentration value found
anywhere on the receptor rings are used (rather than restricting the analysis to
receptors only located within the Class I area(s) of interest). More discussion of the
steps to be taken and processing requirements for a screening analysis is provided in
Section 2.1.
IWAQM's recommendations for a refined analysis involve the following
differences from the screening analysis:
• use of a fully developed time and space varying characterization of the
meteorology using CALMET, and
• the receptors are placed within the Class I area(s) of concern, and
• the background concentrations of ozone and ammonia are allowed to vary in
time and space, and
• the concentration and AQRV impacts are computed to more directly correspond
to the standards, increments, and thresholds of concern.
More discussion of the steps to be taken and processing requirements for a refined
analysis is provided in Section 2.2.
Sections 2.1 and 2.2 focus on how to apply models, specific options and data
sets to be employed, and the processing of the input and output data. Section 2.3
provides more general recommendations on practical issues and limitations of long-
range transport modeling assessments.
2.1 Screening Analysis
Section 4.7 presents comparisons of puff and plume model simulation results to
demonstrate what differences might arise in simulated concentration values when the
plume and puff model employ essentially identical meteorology and dispersion
characterizations. The results shown in Section 4.7 show striking evidence that treating
the sequence of meteorological events of all hours (including calms), can result in puff
simulated maxima that are considerably higher than plume simulated maxima for
-------
almost any distance downwind or averaging time. This was most evident for the shorter
averaging times that were 24-hours or less. The IWAQM concludes from these results
that use of a plume model as a screen for a puff model's impacts is unlikely to be
successful. In Section 4.8, a summary is presented of comparisons of results obtained
by using a puff model with single station meteorology (a screening analysis) versus fully
generated wind fields (a refined analysis) for each hour. These results suggest that the
maximum concentration values simulated using the proposed screening approach for a
receptor ring may occasionally underestimate results obtained from a refined model
simulation. To address this tendency, IWAQM recommends use of the maximum
concentration found anywhere on the receptor rings, rather than limiting the analysis to
only receptors within the Class I area(s) of concern (as would be the case in a refined
analysis). These conclusions are for maxima on receptor rings at fixed distances from
isolated point sources where the terrain was relatively flat.
With these thoughts in mind, the following CALPUFF screening procedure is
suggested by IWAQM (as outlined in Table 1):
1) generate five years of ISCST3 input meteorology using PCRAMMET,
2) generate an ISCST3 control file (use standard ISC defaults and create
receptor rings as appropriate for the application); use the ISC2PUF
conversion program to create the CALPUFF control file,
3) edit the CALPUFF control file to select MESOPUFF II chemistry, and
specify domain-wide background concentration values for ozone and
ammonia (see Section 2.2.2),
4) run CALPUFF with the ISCMET.DAT data option, and pick the maximum
concentration for each pollutant, for each receptor ring and averaging time
modeled. Perform increment and AQRV comparisons as required. For
haze impact assessment, use the FLM provided "clean" background
extinction coefficient and assume a RH value of 90%.
In Step 1 above, it is anticipated that an update will be made to PCRAMMET and the
Meteorological Processor for Regulatory Modeling (MPRM) so that they will list the
relative humidity for each hour, along with the other parameters needed to compute
deposition. With this update, the MESOPUFF II chemistry can be activated within
CALPUFF. In Step 2 above, it is envisioned that receptor rings would be created that
would pass through the Class I area(s) of interest. The IWAQM recommends that the
placement of the receptor rings, and assumptions to be employed in the CALPUFF
model runs be discussed with the applicable reviewing authorities, prior to actually
-------
Table 1. Outline of recommendations for screening analysis. Given judgement
required in receptor ring placement and background values to be assumed in
CALPUFF and CALPOST analyses, applicants should gain agreement on how
coordination with Federal Land Managers (FLM) and other reviewing authorities will
be addressed, prior to conducting runs and analyses.
Meteorology
Use five years of PCRAMMET (extended output for deposition).
Receptors
Receptors at least every two degrees on rings that encircle source
and pass through the Class I area(s) of interest.
Dispersion
1. Use ISC2PUF to convert an ISC3 control input file for use by
CALPUFF.
2. Edit control file to use MESOPUFF II chemistry; use wet and dry
deposition (use default setups for these).
3. Use domain average background values for ozone and ammonia
for area.
4. Run CALPUFF using ISC meteorology option (note, define 6 to 10
layers in vertical; top layer should extend above maximum mixing
depth expected); horizontal domain extending 50 to 80 km beyond
outer receptor ring.
Processing
1. For PSD increments: use maximum 3-hr, 24-hr and annual S02
and PMio and maximum annual N02 for comparison with allowable
limits.
2. For haze: use maximum 24-hr S04=, N03", and HN03 values;
assume 90% RH forf(RH) for day, calculate extenction coefficients for
each pollutant (see Section 3.2); and compute the precent change in
extinction using the FLM supplied background extinction, as described
in Section 3.
3. For total S or N deposition: convert deposition flux to
kg/(hectar«year) using maximum values of annual S02, S04=, N03",
HN03, and NOx as described in Section 3.
conducting any screening analyses. These up front discussions are essential for
defining the AQRV's of interest, which dictate the averaging times and pollutants of
interest for the AQRV assessments. The maximum 3-hr, 24-hr and annual S02 and
PMio, and the annual nitrogen dioxide concentrations would be compared with the
current standards and PSD increments as required by the applicable reviewing
authorities. If a haze impact assessment is required, then use the 24-hour maximum
sulfate, nitrate and primary particulate concentrations in equation (8) of Section 3.2. If
total nitrogen or total sulfur deposition impacts are required, then use the maximum
annual concentration values and follow the procedures as described in Section 3.3.
-------
2.2 Refined Analysis
Table 2 outlines the basic IWAQM recommendations for a refined analysis. It is
understood that conducting a CALMET and CALPUFF analysis for any scale of
application will necessarily involve case-specific judgements. It is IWAQM's conclusion
that it is not possible to prescribe all the decisions needing to be made. We have
attempted to provide suggestions where possible. We have recommended that all
users of CALMET and CALPUFF start with a common set of default conditions and
input data. Then, as the decisions are made to change default settings and discard or
augment the input data, these can be discussed and reviewed with the applicable
authorities, prior to actually committing expensive and time-consuming resources.
Within this context use of technical review committees, as suggested in the Regional
Approach (Section 4.5), would greatly assist applicants in these endeavors.
2.2.1 Meteorology
Expertise Needed
Currently, developing CALMET meteorological fields is considered a difficult task
just managing the sheer volume of input and output data of CALMET, and excellent
computer skills are needed to manage the operation of the various processors to
CALMET. The software was not written to accept a variety of input data formats. The
software was developed with the assumption that the user is capable of screening the
data for anomalous values. It was assumed that if the data are not in the required
format, the user has the programming skills to write special programs to translate the
data format to the format required.
Renovating these programs was beyond the resources available to IWAQM.
The IWAQM focused what resources it had to the issues of testing the technical
aspects of the modeling system (comparisons with tracer experiments, enhancement of
processing controls, etc.) to see if the CALMET/CALPUFF modeling system was
technically sufficient for routine use. It has been assumed that support to renovate
these processors can be found, if and when, the CALMET/CALPUFF modeling system
becomes a recommended modeling approach.
Currently, adjusting and tailoring the CALMET options to see if the generated
wind fields are reasonable requires strong computer graphics skills to visualize the
results generated by CALMET. This component of the analysis can be simplified by
using existing visualization software.
The control of the CALMET options requires expert understanding of mesoscale
and microscale meteorological effects (such as terrain slope flows) on meteorological
-------
Table 2. Outline of recommendations for refined analysis. Given judgement required
in CALMET, CALPUFF and CALPOST processing, applicants must gain agreement
on how coordination with Federak Land Managers (FLM) and other reviewing
authorities will be addressed, prior to conducting runs and analyses.
Meteorology
1. Use five years of representative hourly NWS surface and
precipitation observations, and twice-daily upper air observations. On
a case-by-case basis, one may use less than five years of FDDA-MM
data with representative NWS data; or use less than five years of on-
site meteorological data with available NWS data.
2. CALMET (minimum of 6 to 10 layers in vertical; top layer must
extend above maximum mixing depth expected); horizontal domain
extending 50 to 80 km beyond outer receptors and sources being
modeled; terrain elevation and land-use data resolved for situation
(tailor land-use parameters to situation).
Receptors
Within Class I area(s) of concern, provide coverage tailored to
situation (insure FLMs concur).
Dispersion
1. CALPUFF with default dispersion settings.
2. Use MESOPUFF II chemistry; use wet and dry deposition (use
default setups for these).
3. Define background values for ozone and ammonia for area (tailor
spatial variability to situation needs and data availability).
Processing
1. For PSD increments: use highest second-highest 3-hr, 24-hr
concentration values, and use the maximum annual concentration
values of S02 and PMio and maximum annual for N02 for
comparisons with allowable limits.
2. For haze: process the 24-hr S04=, N03", and HN03 values adjusted
using hourly RH as discussed in Section 3; calculate extinction
coefficients for each pollutant (Section 3.2), compute the percent
change in extinction using the FLM supplied background extinction, as
described in Section 3; compare with thresholds as directed by
applicable FLMs and reviewing authorities.
3. For total S or N deposition: convert deposition flux to
kg/(hectar«year) using maximum values of annual S02, S04=, N03",
HN03, and NOx as described in Section 3; compare with threshold as
directed by applicable FLMs and reviewing authorities.
conditions, and finesse to adjust the available processing controls within CALMET to
develop the desired effects (e.g., Section 4.9.2). The IWAQM does not anticipate a
lessening in this required expertise in the future. Developing three-dimensional time-
varying fields of meteorological conditions is a demanding task, which can not be left to
unskilled or inexperienced staff. The IWAQM does not foresee a time when
10
-------
development of time-varying mesoscale meteorological wind fields (by whatever
means) will become as simplified as the running of the ISC3 modeling software has
become in recent years.
Appendix A provides a listing of the default settings recommended by IWAQM at
this time. Some of these settings require testing, and we have attempted note these.
The information provided in Appendix A should not be interpreted as a cookbook
approach to be applied, regardless of results obtained.
Length of Modeling Assessment
Significant year-to-year variations were seen in the Demonstration Assessment
(Section 4.1) and also in the developmental work towards attempting to build a new
screening methodology (Section 4.8). These results suggest that several years of
refined analysis are needed to address the variation in pollution impacts likely to occur.
For consistency with other Guideline requirements, a five-year period of analysis using
representative NWS meteorological data is recommended. CALMET employs
diagnostic algorithms to tailor the available meteorological data for slope flow effects,
land-sea circulations, etc. If special meteorological observations are available within
the modeling domain that would assist CALMET's diagnostic analysis, or if FDDA-MM
data are to be employed in the CALMET analysis, then less than a five-year period of
analysis can be accepted, on a case-by-case basis by the applicable reviewing
authorities. This accommodates use of refined meteorological data for those locations
where refined meteorological data have been suitably processed, or where mesoscale
meteorological campaigns have been accomplished.
Geophysical Data
Terrain heights and land-use are needed for input to CALMET. The IWAQM has
provided one set of these values for the contiguous United States. The grid resolution
is approximately 900 m for the terrain data and 1/12 degree latitude (9.25 km) by 1/8
degree longitude (9.8 km at 45° latitude) for the land-use data (CALMET, CALPUFF,
And CALPOST Modeling System (Version 1) CD-ROM; see Appendix C). Default
values characterizing the surface parameters associated with each land-use class are
also provided. The resolution of this data set is considered adequate for assessments
involving transport distances of 50 km and greater, although the land-use categorization
should be examined carefully. The resolution of this data set may also be adequate for
many local-scale assessments.
The IWAQM recommends use of this data set, as a first choice. The IWAQM
recognizes that there are instances when these terrain heights and land-use data must
be considered inappropriate, such as the situation discussed in Section 4.9.2. It is
anticipated that rejection of this data set will most often be for treatment of sources
located in very rugged terrain, which is anticipated to significantly affect the transport
trajectory of the plume. Using this data set, as a first choice with assignment of the
11
-------
default surface parameters as listed on the distribution CD, should provide consistency
in future analyses and helps to standardize the input requirements for software
development.
Another source of terrain heights and land-use data is from the United States
Geological Survey (USGS). Some of these data can be accessed through the World
Wide Web. The general USGS site is at http://www.usgs.gov. At the time of this
writing, the data was available at http://edcwww.cr.usgs.gov/doc/edchome/
ndcdb/ndcdb.html.
Precipitation Data
Precipitation is notoriously spotty, with many localized maxima and minima. It is
recommended by IWAQM that all precipitation reports reasonably available be used.
Precipitation data are available for the United States from the National Climatic
Data Center (NCDC) in a format called TD-3240. No data filling is necessary for these
data sets for use by CALMET. But as discussed in EPA (1995a), the formatting of
these data for a large domain is not trivial. The software available is not robust and
could stand to be improved.
Another source of precipitation data is from private firms that purchase the data
from NCDC, repackage it, and provide software for extracting the repackaged data.
This may be an attractive alternative to some users.
National Weather Service Data
The number of surface and upper air sites will be determined by the size of the
modeling domain and the availability of meteorological data. It is recommeded that for
refined analyses one would use all sites that are within and near the modeling domain,
so as to provide as good a characterization as possible of the spatial variation of the
meteorological conditions.
The CD-ROM data sets (Appendix C) are made available by the National
Climatic Data Center. These hourly surface weather observations (Solar and
Meteorological Surface Observation Network (SAMSON) and Hourly United State
Weather Observations (HUSWO) 1990-1995), and twice-daily upper air observations
(Radiosonde Data for North America) are recommended by IWAQM as first choice for
acquiring access to these data. Using these data sets, as a first choice, should provide
consistency in future analyses and helps to standardize the input requirements for
software development.
The current CALMET software requires the user to closely inspect the upper-air
data prior to use. Since the program does not fill in missing levels or time periods
(which sometimes occur in the observations), the user should inspect the data to insure
12
-------
that complete data is available at least for all levels below and just above the
anticipated maximum mixing depth for the modeling domain. Furthermore, CALMET
can not process if the time difference between successive upper-air soundings is
greater than 12 hours. If this occurs, the user must use judgement to fill in the requisite
data. When multiple NWS sites are used for characterizing the hourly surface weather
conditions, missing observations from one station can be handled simply by filling in a
missing data indictor for that site for that hour. But the user should review the hourly
surface observations to insure that the data coverage in space and time is adequate.
This requires judgement and finesse and given the complexities of long-range transport
analysis, is not amenable to cookbook cures.
FDDA-MM Data
As mentioned in Section 4.4 there are various groups capable of developing
FDDA-MM data suitable for use as input to the CALMET/CALPUFF modeling system.
Alternatively, consideration could be given to using the 1990 MM4 data set (NCDC,
1995) or other data sets which might be completed for other years. The problem with
these data sets is that the they are costly to construct and advances are currently being
made at such a rapid pace, they are becoming dated almost as soon as they are
constructed. Furthermore, there are situations defined in the Guideline when on-site
meteorological data are to be collected and used in the modeling assessments (e.g.,
when using CTDMPLUS for a complex terrain impact assessment, where the transport
distances are less than 50 km). Hence, there appears to be a never-ending need to
construct additional years of data. The IWAQM recommends finding a link to some
operational source of FDDA-MM data, as the optimal solution.
Towards this end, IWAQM recommends that links to the operational model
outputs that are produced by the NOAA Mesoscale Modeling Branch (MMB) (see
discussion by Schulze and Turner, 1998). This group is developing data suitable for
use in the CALMET/CALPUFF modeling system. They have an active plan to refine
and improve the modeling science and to reduce the grid size in future years. These
model outputs are being developed to directly support operational needs of the National
Weather Service, and should prove to be a stable source of such data in the future.
The major obstacle is access to these data. An operational means for gaining
easy access to comprehensive mesoscale meteorological data sets has yet to be
developed. Therefore, IWAQM recommends that a solution be developed that will
provide the public routine, inexpensive access to such data. The data can be
depended on being available in future years. Users who either are required to collect
and use on-site data, or need such data for best characterization of their situations, can
use CALMET to manage the blending of their on-site data, with routine NWS
observations and available FDDA-MM data.
13
-------
Once an operational link is established to an operational source of FDDA-MM
data and a sufficiently long record of the data is available, IWAQM recommends the
use of these data.
2.2.2 Chemistry
The MESOPUFF II chemistry option, currently available in CALPUFF, is
adequate for representing gas phase oxidation of sulfur dioxide to sulfate and for the
nitrate chemistry. The algorithms currently do not adequately account for the aqueous
phase oxidation of sulfur dioxide to sulfate. The aqueous phase chemistry can
dominate the formation of sulfate. Therefore, in many applications sulfate is likely to be
underestimated.
The IWAQM recommends use of the MESOPUFF II chemistry option, although it
is recognized that this module may grossly underestimate the conversion of sulfur
dioxide to sulfate when the pollutants interact with clouds or fog.
Ozone Background
CALPUFF provides two options for providing the ozone background data: (1) a
single, typical background value appropriate for the modeling region, or (2) hourly
ozone data from one or more ozone monitoring stations. The second and preferred
option requires the creation of the OZONE.DAT file containing the necessary data. For
the Demonstration Assessment, the domain was large (700 km by 1000 km) such that
the second option was necessary. The IWAQM does not anticipate such large domains
as being the typical application. Rather, it is anticipated that the more typical
application will involve domains of order 400 km by 400 km or smaller. But even for
smaller domains, the ability to provide at least monthly background values of ozone is
deemed desirable. The problem in developing time (and perhaps spatial) varying
background ozone values is having access to representative background ozone data.
Ozone data are available from EPA's Aerometric Information Retrieval System
(AIRS); however, AIRS data must be used with caution. Many ozone sites are located
in urban and suburban centers and are not representative of oxidant levels experienced
by plumes undergoing long range transport.
Ammonia Background
A further complication is that the formation of particulate nitrate is dependent on
the ambient concentration of ammonia, which preferentially reacts with sulfate. The
ambient ammonia concentration is an input to the model. Accurate specification of this
parameter is critical to the accurate estimation of particulate nitrate concentrations.
Based on a review of available data, Langford et al. (1992) suggest that typical (within a
factor of 2) background values of ammonia are: 10 ppb for grasslands, 0.5 ppb for
forest, and 1 ppb for arid lands at 20°C. Langford et al. (1992) provide strong evidence
14
-------
that background levels of ammonia show strong dependence with ambient temperature
(variations of a factor of 3 or 4) and a strong dependence on the soil pH. However,
given all the uncertainties in ammonia data, IWAQM recommends use of the
background levels provided above, unless specific data are available for the modeling
domain that would discredit the values cited. It should be noted, however, that in areas
where there are high ambient levels of sulfate, values such as 10 ppb might
overestimate the formation of particulate nitrate from a given source, for these polluted
conditions. Furthermore, areas in the vicinity of strong point sources of ammonia, such
as feed lots or other agricultural areas, may experience locally high levels of
background ammonia.
2.2.3 Dispersion
Expertise Needed
The control of the CALPUFF options requires expert understanding of terrain
affects on meteorological conditions and some finesse to adjust the available
processing controls. Appendix B provides a listing of the default settings for CALPUFF
recommended by IWAQM at this time. Some of these settings require testing, and
IWAQM has attempted note these. The information provided in Appendix B should not
be interpreted as a cookbook approach to be applied, regardless of results obtained.
Emissions
Developing an inventory with agreed upon emission rates is not trivial. The
inventory could differ depending on whether the analysis is addressing NAAQS
assessments (which typically address maximum allowable emission rates from only
PSD sources) versus AQRV assessments (which typically address actual current
emission rates from all existing sources). The IWAQM recommends that the manner in
which the sources and emissions are to be characterized be agreed upon in the initial
up-front discussions with the reviewing authorities.
Receptors
For some Class I areas the FLMs have developed receptor networks for use in
the modeling analyses, which facilitates assessment of cumulative impacts from
successive applicants. In such cases, the applicants should use these predefined
receptor networks. In other situations, IWAQM suggests use of judgement. The more
rugged the terrain and the presence of local sources being included in the analysis will
necessitate use of more closely spaced receptor networks in order to adequately
characterize the concentration patterns to be simulated for the Class I area(s). This is
another area where judgement and review by others will be needed.
15
-------
Local-scale
The IWAQM is recommending use of the CALPUFF modeling system for the
characterization of all sources being explicitly modeled. This eliminates the need to
simulate the long-range impacts (those involving transport greater than 50-km)
separately, and then combine these results with those obtained using some other
model for the local-scale impacts (those involving transport of less than 50-km). We
have CALPUFF and ISC comparison results using both steady-state and non-steady-
state hourly meteorology. It was the conclusion of IWAQM that CALPUFF could
reproduce the steady-state results of the ISC plume dispersion model. A benefit of
using one model for all sources is that CALPUFF has the MESOPUFF II chemistry,
which provides characterization of pollutant species that are not treated by currently
available local-scale models (such as ISC and CTDMPLUS).
Comprehensive tests results are not available comparing CALPUFF with
CTDMPLUS for steady-state impacts on isolated hills and ridges. The IWAQM
assumes that once these have been accomplished and the results have been shown to
be similar, that long-range impact assessment with CALPUFF (using fully developed
CALMET meteorology) can also include more explicit hill impaction assessments, as
necessary, for all sources regardless of the transport distances involved.
2.3 Practicalities
The IWAQM recommendations for conducting a long-range transport screening
analysis are presented in Section 2.1, and the IWAQM recommendations for
conducting a long-range transport refined analysis are presented in Section 2.2. The
focus of these discussions was how to run the simulations and the management of the
input and output data. There are other practical concerns that applicants and reviewing
authorities should be mindful of, namely: the uncertainties associated with the
screening analysis, the uncertainties associated with the refined analysis, differences
between plume and puff simulation results, and commensurate difficulties in providing
technical oversight.
2.3.1 Screening procedures uncertainties
In Section 4.8 comparisons are presented of CALPUFF simulation results
generated either through the use of ISC or CALMET meteorology. Anticipating that
most analyses will involve a moderate to tall stack, of order 35-m to 200-m in height, it
is seen that the screening estimates of sulfur-dioxide and sulfate concentration maxima
obtained using ISC meteorology, typically range within ±70% of that simulated using
CALMET meteorology. The sulfur dioxide and sulfate deposition fluxes obtained using
ISC meteorology, typically range within ±60% of that simulated using CALMET
meteorology. This suggest that the screening analysis as proposed is not providing a
biased (overestimate) of these impacts. It was for this reason that IWAQM
recommended that all receptors on the ring be included in the screening assessment. It
16
-------
was hoped that this would provide a measure of conservatism to the screening
analysis. Adding a measure of conservatism is deemed reasonable, as the proposed
screening analysis completely disregards the terrain and land-use induced wind effects,
that would arise if fully-developed three-dimensional wind fields were developed using
available surface and upper-air observations. The IWAQM concludes that the impacts
estimated by the screening procedure proposed are conservative and yet less onerous
than results as would be obtained by the Phase 1 Level 1 screening, and addresses
concerns raised at the Sixth Modeling Conference
2.3.2 Refined modeling uncertainties
In Section 4.6 comparison results of CALPUFF simulations with tracer field data
are provided. The studies presented are considered representative of a few of the
better tracer field studies that involve comprehensive sampling along arcs at downwind
distances of order 50- to 100-km. As summarized in Section 4.6, in general the
CALPUFF simulated concentration values were within a factor of two of that observed.
Little differences in overall performance was seen, whether Pasquill or similarity
dispersion curves were employed. This is understandable since as transport times and
distances increase, the dispersing material is becoming well-mixed in the vertical, and
the horizontal extent of the dispersing material is more related to the wind field
variations than is the rate of relative dispersion about simulated puff centroids. This is
dramatically made apparent by the results shown in Section 4.7, where direct
comparisons were made of the ISC plume model and CALPUFF. In these comparisons
it was shown that for steady-state assumptions, CALPUFF could suitably mimic results
as would be obtained by ISC. And yet with identical specification of the dispersion and
meteorology, once the puff model was allowed to simulate the 'causality' of hour-by-
hour variations, the puff model's results no longer were similar to that obtained by ISC.
In fact, as a result of explicitly treating calms and wind reversal effects by the puff
model, the simulated maximum concentrations by CALPUFF were generally greater
than that simulated by ISC for all transport distances and averaging times.
The quality and uncertainties associated with long-range transport simulations is
driven more by the characterization of the mixing depth and by the characterization of
the transport winds. Better characterization of the mixing depth may be possible in the
future, but it should be understood that the mixing in the vertical reflects both local and
mesoscale effects. The influence of local variations in land-use (forests, lakes,
farmland, cities) can be significant. Hence uncertainties of order ±40% in mixing depth
are likely (Irwin and Paumier, 1990) and will be difficult to avoid. Improvement in the
characterization of the transport winds is possible through the use of FDDA-MM
meteorological data, as discussed in Section 4.3. Both the comparisons of simulated
trajectories and the comparisons of trajectories derived using the CAPTEX tracer field
show improvement through the use of sophisticated mesoscale meteorological data
employing FDDA.
17
-------
Based on the tracer comparison results presented in Section 4.6, it appears that
CALPUFF provides reasonable correspondence with observations for transport
distances of order 100 km. Most of these comparisons involved concentration values
averaged over 5 to 12 hours. The CAPTEX comparisons, which involved comparisons
at receptors that were 300 km to 1000 km from the release, suggest that CALPUFF
tends to overestimate surface concentrations by a factor of 3 to 4. Use of the puff
splitting option in CALPUFF might have improved these comparisons, but there are
serious conceptual concerns with the use of puff dispersion for very long-range
transport (300 km and beyond). As the puffs enlarge due to dispersion, it becomes
problematic to characterize the transport by a single wind vector, as significant wind
direction shear may well exist over the puff dimensions.
With the above thoughts in mind, IWAQM recommends use of CALPUFF for
transport distances of order 200 km and less. Use of CALPUFF for characterizing
transport beyond 200 to 300 km should be done cautiously with an awareness of the
likely problems involved. Since the long-range transport results appear to be relatively
insensitive to the exact selection made for characterizing the puff dispersion
parameters, IWAQM recommends use of dispersion parameters that provide results
most similar to the local-scale model of choice (which currently is ISC). It would appear
that CALMET is capable of treatment of highly complex wind fields that are strongly
influenced by terrain slope flows (Section 4.9.2). However, as mentioned at the Sixth
Modeling Conference, there are situations that one can imagine involving highly rugged
terrain, that any model simulation's results must be viewed as uncertain.
2.3.3 Secondary pollutant uncertainties
The CALPUFF simulation for gas phase oxidation of sulfur dioxide to sulfate and
for the nitrate chemistry is considered adequate. The algorithms currently do not
adequately account for the aqueous phase oxidation of sulfur-dioxide to sulfate, which
can become dominant in the presence of fog or clouds. Even if planned updates to
CALPUFF include consideration of the aqueous phase chemistry, the input data may
preclude routine use of this enhancement. Finally, it must be mentioned that evaluation
studies have not been conducted regarding CALPUFF's simulations of secondary
formed sulfate and nitrate. The IWAQM is not aware of comprehensive tracer-field-
data studies that would lend themselves to such evaluation studies. Given that
aqueous phase chemistry is not treated, it is likely that CALPUFF simulations of sulfate
would be less than observed. Mention is made in Section 4.9.1 that the total nitrogen
deposition estimates by CALPUFF were about a factor of 3 less than estimates
provided by a more physically complete model (RADM). Notwithstanding these obvious
deficiencies, IWAQM recommends use of CALPUFF's estimates of sulfate and nitrate
for purposes of addressing the need to assess AQRV impacts in Class I areas (in lieu
of assuming that there are no impacts at all).
18
-------
2.3.4 Technical oversight and review
It would be convenient if objective criteria and cookbook procedures could be
constructed that would preclude inappropriate application of air dispersion models.
This has proved to be troublesome for local-scale modeling, and likely is impossible for
mesoscale and long-range transport modeling. As in any air quality simulation, the
usefulness of the results obtained depends mostly on the expertise brought to the
analysis in characterizing the situation, and on the experience applied in interpreting the
results obtained. In response to these considerations, IWAQM has attempted to warn
the modeling community that conducting a long-range transport assessment requires
esperts. We have also tried to warn the modeling community that application of the
CALPUFF modeling system to any situation will require expert judgment, it will likely
involve site-specific decisions, and it will require strong interaction and coordination with
the applicable reviewing authorities.
In this regard, the use of technical review committees, as suggested in the
Regional Approach (Section 4.5) would likely prove useful to both the applicants and to
the reviewing authorities. These technical committees could assist in sorting through
the site-specific decisions, and they could provide a forum for reaching consensus.
Having a standing technical committee would provide applicants with some assurance
of being treated equitably, and could provide data sets for use to facilitate comparability
between individual analyses. Whether such a committee would prove useful is
dependent on availability of experts and on the ability to obtain long-term commitments
of service. Furthermore, not all Class I areas will require use of such committees.
Hence, while IWAQM endorses the use of such committees, IWAQM does not
recommend Federal agencies mandate or require their use.
19
-------
3.0 TRANSFORMATIONS, VISIBILITY, AND DEPOSITION
In principle, estimating the impact of emissions from an individual source is best
accomplished using a Lagrangian dispersion model, as a Lagrangian model 'follows'
the emissions as they are transported downwind. This provides within the model the
ability to directly assess how much each source's emissions are impacting each
receptor in the analysis. However, there are some physical processes (like nonlinear
chemical transformations) that can be more explicitly and completely characterized
using Eulerian grid modeling techniques. To treat nonlinear chemical transformations
within a Lagrangian modeling framework is presently so computationally demanding as
to be impractical for routine use.
The IWAQM recognized these limitations and tradeoffs between Lagrangian and
Eulerian grid modeling techniques. The purpose of the Phase 2 recommendations was
to recommend techniques useful for permitting individual PSD sources. For individual
source impacts involving complicated terrain and transport distances of order 50 to 250
km, a Lagrangian puff model like CALPUFF is an optimal choice. In this section we
summarize the approximations made in the CALPUFF chemical transformations and
associated inherent limitations. We then summarize the methodology recommended
for computing haze and deposition impacts in Class I areas. Finally, we discuss
conceptually how assessment of adverse impact of Air Quality Related Values is
somewhat different than traditional NAAQS assessments.
While drafting this report, the National Park Service, the U.S. Fish and Wildlife
Service and the U.S. Forest Service have been holding intensive meetings to promote a
greater consistency in the procedures Federal Land Managers use in identifying and
evaluating AQRV impacts. We have discussed in this report the assessment of
regional visibility impacts using the deciview, which at the time of the drafting of this
report was the preferred metric. As time progresses, it is looking more like the change
of extinction may become the preferred metric. Hence, although the information
provided here is useful, the details and implementation may be somewhat different as a
consequence of the ongoing discussions. For the latest information on procedures and
metrics, we suggest visiting the web site: http://www.nature.nps.gov/ard/flagfree/index.html.
3.1 Chemical Transformations
The MESOPUFF II chemistry option, currently available in CALPUFF(4.0), is
adequate for representing gas phase oxidation of S02 to S04= and for the nitrate
chemistry. The algorithms currently do not adequately account for the aqueous phase
oxidation of S02 to S04=. The aqueous phase chemistry can dominate the formation of
sulfate. Therefore, in many applications sulfate is likely to be underestimated. A further
complication is that the formation of particulate nitrate (N03") is dependent on the
ambient concentration of ammonia, which preferentially reactswith S04=. The ambient
ammonia concentration is an input to the model. Appropriate specification of this
20
-------
parameter is critical to the estimation of realistic values of participate nitrate
concentrations.
For instance, if we were simulating effects of emissions coming from an as yet to
be built new single source, then the ammonia available to react with these new
emissions could be estimated from monitored background levels of ammonia. If there
are large major sources near the proposed new source included in the analysis, then
the background ammonia should in principle be adjusted to a value somewhat above
that monitored to reflect ammonia available that these existing sources typically
scavenge from the atmosphere. By way of this example, we see that CALPUFF is most
easily applied for isolated new emissions, and becomes more problematic as the
number of sources increases. This supports limiting application of CALPUFF to a
relatively few sources of emissions, so that the 'background' levels of ozone and
ammonia can be derived using appropriate monitoring data.
The spatial and temporal scales, where CALPUFF might be used, are long
enough so that the chemical conversion of S02 to S04= and NOX to HN03 are of
interest. The oxidation of SOX and NOX may occur by gas and aqueous phase
reactions. The gas phase reactions for both SOX and NOX involve free radical
photochemistry and, therefore, are coupled to the oxidation of reactive organic gases
(ROG). Homogeneous gas phase reaction is the dominant S02 oxidation pathway
during clear, dry conditions. Ozone and hydrogen peroxide are believed to be the
principal oxidants for aqueous phase oxidation of S02. Homogeneous gas phase
reactions may convert S02 at most a few percent per hour, whereas aqueous phase
reactions can convert S02 up to 100% per hour.
The oxidation of NOX is dependent on gas phase ROG/NOX/03 photochemistry. It
is generally more rapid than S02 gas phase oxidation. NOX can be oxidized to nitric
acid (HN03) and organic nitrates (RN03) such as peroxyacetylnitrate (PAN). HN03
combines with ammonia gas to form solid or aqueous ammonium nitrate (NH4N03).
Unlike sulfate formation, the nitrate process is reversible. Equilibrium is established
between nitric acid, ammonia, and ammonium nitrate:
NH4NO3 ^ HNO3 + NH3
The equilibrium constant for this reaction is a nonlinear function of temperature
and relative humidity. The equilibrium constant can vary several orders of magnitude
over a typical diurnal cycle. Given fixed amounts of total nitrate, ammonia, and water
vapor, higher NH4N03 concentrations are expected at night due to lower nighttime
temperatures and higher relative humidities. Thus, the nitrate aerosol can not be
considered a stable product like sulfate.
The transformation rate expressions, used in the MESOPUFF II algorithm, were
developed statistically analyzing hourly transformation rates produced by a
21
-------
photochemical model. Plume SOX/NOX dispersing into background air containing ozone
and reactive hydrocarbons was simulated over a wide range of conditions representing
different solar radiation intensities, temperatures, dispersion conditions, background
ozone concentration and surface-level relative humidity, RH, plume NOX concentrations
and emissions times. The following transformation rate expressions, representing
curve fits to the daytime hourly conversion rates predicted by the photochemical model,
were determined:
lr -'JApO.SSr/o iO.71o-l.29 , ir
k\ ~36R \.°T\ *> + k\(aq)
fc, = i26i[<93]L4lrL34[M}j-°-12
Where, k., is the S02 to S04= transformation rate (percent/hour),
k2 is the NOX to HN03 + RN03 transformation rate (percent/hour),
k3 is the NOX to HN03 (only) transformation rate (percent/hour),
R is the total solar radiation intensity (kw/m2),
S is a stability index from 2 to 6 (PG stability A&B = 2, C=3, etc.,
RH is the surface-level relative humidity (percent),
[03] is the background ozone concentration (ppm),
[NOX] is the plume NOX concentration (ppm), and,
k1(aq) is the aqueous phase S02 oxidation term.
The term, k1(aq) peaks at a value of 3 percent/hour at 100% relative humidity. This
is much less than might be expected if the plume interacted with clouds or fog. These
transformation rate expressions only apply during the day. At night oxidation rates of
0.2% and 2.0% for S02 and NOX, respectively, are suggested as default values in the
model.
3.2 Visibility Analysis
In the context of the Phase 2 recommendation, the focus of the visibility analysis
is on haze. These techniques are applicable in the range of thirty to fifty kilometers and
beyond from a source. At source-receptor distances less than thirty to fifty kilometers,
the techniques for analyzing visual plumes (sometimes referred to as 'plume blight')
should be applied.
There are two approaches to determining visibility effects (NAPAP Report to
Congress). One is a technically rigorous, complex, and situation-specific method, while
the other is a more generalized approach. The more rigorous approach requires the
calculation of aerosol growth dynamics and the application of Mie theory to determine
the optical characteristics of the aerosol distribution. Sophisticated radiative transfer
models are then applied, using the aerosol optical characteristics, the lighting and
scene characteristics and the spatial distribution of the pollutants to calculate the path
22
-------
and wavelength of image-forming and non-image-forming light that reaches a specific
observer at a specific time and date from all points in the scene being viewed. While
this detailed analysis may be useful for assessing specific cases, it is impractical for
addressing haze issues where visibility is experienced in a nearly infinite variety of
situations and where detailed characteristics of the pollution, lighting, and scene
conditions are rarely known.
The generalized approach uses aerosol species extinction efficiencies, with water
growth functions, to determine the light extinction coefficient of the aerosol, from its
composition and from the relative humidity of the atmosphere. The extinction
efficiencies and relative humidity dependence of the aerosol are based upon typical
results from the more rigorous analyses mentioned above.
A generalized approach is recommended by IWAQM for Class I area analyses.
Under this approach, the concentrations of pollutants (in this case calculated by an air
quality model) are used to calculate the extinction coefficient due to these pollutants.
This is then compared against the light extinction coefficient of the background air. A
constant fractional change in the extinction coefficient produces a similar perceptual
change for a scene regardless of baseline conditions. Therefore, under cleaner
visibility conditions, relatively less pollution will cause a perceptible change than under
more polluted background conditions.
Visibility is an instantaneous phenomenon. When an observer looks at a scene,
the view is what is seen at that moment. Many of our measurements and modeling
techniques deal with averaged values from one hour and longer. Therefore, some
consideration is needed to accommodate this dichotomy.
Visibility Parameters
Visibility is usually characterized by either visual range (MR) (the greatest distance
that a large dark object can be seen) or by the light-extinction coefficient (bext) (the
attenuation of light per unit distance due to scattering and absorption by gases and
particles in the atmosphere) (Sisler,1996). Under certain assumed conditions, these
parameters are inversely related to each other by Equation 1.
VR(hn) = 3'912 (1)
The dimensions of MR are length and the dimensions of bext are 1/length. Visual range
is usually expressed in kilometers. The extinction coefficient is sometimes expressed
as "inverse kilometers" (km"1) or as "inverse megameters" (Mm"1) (the reciprocal of 1
million meters). If bext is expressed in Mm"1, the coefficient 3.912 in Equation 1
becomes 3912. The value 3.912 = -ln(0.02), which assumes a two percent contrast
23
-------
threshold for the viewer. Other researcher have assumed a five percent threshold,
which would change the value to 2.995.
A constant fractional change in the extinction coefficient produces a similar
perceptual change for a scene regardless of baseline conditions. Using the relationship
of a constant fractional change in the extinction coefficient resulting in a similar
perceived visual change, an alternate visibility index, the deciview (dv) has been
defined (Equation 2).
b (km-1)
dv = 10 In -^ (2)
-
This index was specifically designed so that anywhere along its scale, haziness
changes that are equally perceptible correspond to the same deciview difference. For
example, a 3dv difference caused by a change in air quality should result in about the
same perceived change in haziness, whether under clean or highly polluted conditions.
This characteristic of the deciview scale requires that the scene being viewed has
sufficient sensitive scenic features to detect changes in visibility from the baseline haze
level. No one scene is likely to have such scenic features for all conceivable haze
levels. However, the nearly infinite variety of scenes available, where hazes are
concerned, ensures that many will have the desired characteristic for any haze level
(NAPAP report to congress).
Calculating the Extinction Coefficient
Visibility is degraded by light scattered into and out of the line of sight and by light
absorbed along the line of sight. Light extinction is the sum of light scattering and
absorption, and is usually quantified using the light extinction coefficient (bext). Using
the generalized approach to estimating visibility effects, outlined above, one can
calculate the extinction coefficient as the sum of its parts, i.e., bext = bscat + babs, where
bscat and babs are the light scattering and absorption coefficients. The light scattering
and absorption coefficients can be further broken down by their respective components.
The scattering coefficient is affected by Rayleigh scattering (bRay) from air molecules
and from particle scattering (bsp); the particles can be natural aerosol or result from air
pollutants. The absorption coefficient is affected by gaseous absorption (bag) and
particulate absorption (bap). Nitrogen dioxide is the only major light-absorbing gas in the
lower atmosphere; it generally does not affect hazes, although it can be an important
element in a coherent plume assessment. Therefore, only particle absorption is
considered in the suggested haze analyses.
Particle scattering, bsp, can be broken down by the contributions of different
particulate species. It has been convenient to consider the scattering coefficients of
24
-------
fine particles (PlVb.s) (particles with mass mean diameters less than or equal to 2.5 urn)
and coarse particles (mass mean diameters greater than 2.Sum but less than or equal
to 10um). The fine particle scattering coefficient can be further defined by the sum of
the scattering coefficient due to sulfates (bs04), nitrates(bw03), organic aerosols (boc),
and soil (£>So,/); the coarse scattering coefficient (bcoarse] is typically not refined any
further. Thus the particle scattering coefficient (bsp) can be expressed as in Equation 3.
b = bc,^. + b^i + br,^ + be, ., + bn (V\
sp SO4 NO3 OC Soil Coarse \J )
Each of the particle scattering coefficients can be related to the mass of the
components using the relationships in Equation 4.
bSQ4 = 3 [(NH4}2SO,} flRH)
bN03 = 3 [NH4N03] flRH)
boc = 4 [OC] (4)
bSoil = 1 [Soil Mass}
b Coarse = °'6 \CoOTSe Mass]
The quantities in brackets are the masses expressed in ug/m3. (It is assumed that the
forms of the S04= and N03" are ammonium sulfate [(NH4)2S04] and ammonium nitrate
[NH4N03].) The numeric coefficients are the dry scattering efficiencies (m2/g). The term
f(RH) is the relative humidity adjustment factor. The extinction coefficients are in Mm"1.
If the dry scattering efficiencies are divided by 1000 (i.e., 0.003 instead of 3) the
resultant extinction coefficients will be in km"1.
Particle absorption (bap) is primarily due to elemental carbon (soot). For purposes
of analyzing the effects of soot on visibility in a modeling analysis, the relationship in
Equation 5 should be used. Again, the quantity in brackets is the mass of elemental
= 10
(5)
carbon in ug/m3 and 1 0 is the extinction efficiency.
The total atmospheric extinction can be expressed as in Equation 6. To the
extent that a source contributes to the formation of some of these constituents, those
b f = bv^. + b*m + br,n + bv ., + bn + b + bD (f,\
ext SO4 NO3 OC Soil Coarse ap Ray \y)
25
-------
contributions can be summed to yield the source's contribution to extinction. This will
be discussed in more detail below.
Examination of Equation 4 reveals that the sulfate and nitrate components of the
extinction coefficient are dependent upon relative humidity. These aerosols are
hygroscopic and the presence of water enhances their scattering efficiency. It is
sometimes convenient to consider the sulfate and nitrate components of extinction
separately from the remaining components of Equation 6 and to keep the relative
humidity adjustment factor (f(RHJ) separate. Equation 6 can then be rewritten as in
Equation 7,
+ b, (7)
where
bSN = 3 [(NH4)2S04 + NH4N03]
j = bnr + b~ ., + br + b + bD
dry UL boil Coarse ap Kay
where bSN is the combined extinction coefficient of sulfate and nitrate, excluding the
relative humidity adjustment factor, and bdry is the sum of boc, bsojh bcoarse, bap, and bRay.
The relative humidity adjustment factor requires some further explanation. The
variation of the effect of relative humidity on the extinction efficiency of sulfates and
nitrates is shown in Figure 1. As can be seen, the effect of relative humidity on the
extinction efficiency of these aerosols is non-linear, and is several times greater at
higher relative humidities. These factors are applicable on a short-term basis. If the
particulate concentrations are only available over a longer averaging time (i.e., a 24-
hour sample or a seasonal average) then the average relative humidity adjustment
factor for that time period must be applied, no? the average relative humidity.
(Alternately, short-term extinction coefficients (i.e., 1-hour) may be averaged to yield a
longer-term average.)
26
-------
0 20 40 60 8
Relative Humidity, RH, (percent)
Figure 1. Variation of f(RH) (see Equation 7).
Background visibility conditions
As noted previously, visibility analyses are compared against a background
condition. The estimates of background visibility conditions at Class I areas are derived
from the IMPROVE (Interagency Monitoring of PRotected Visual Environments)
network. There are several methods of obtaining estimates of the background visibility.
These include reconstructed extinction from speciated measurements of particulate
matter, direct measurement of extinction with a transmissometer, and estimates of
extinction from photographs (Malm et al., 1996). Reconstructed extinction is usually
used to estimate background conditions, since this can be directly related to pollutant
loadings. It should be noted that reconstructed extinction values from the IMPROVE
network are based on 24-hour average particulate concentrations. The temporal
average at a point is used to represent a short-term spatial average.
The background conditions provided for a Class I visibility analysis will be
representative of clean conditions. Changes in visibility are most sensitive under clean
conditions. By using clean conditions for all comparisons in a Class I analysis, it
ensures that already clean conditions will not be impaired. Additionally, the Clean Air
Act states as a national goal that the visibility in Class I areas is to be unimpaired by
man-made air pollutants and that any such impairment is to be remedied. To represent
clean conditions, the average of the cleanest 20% of the data from IMPROVE, at that
site, is generally used. Even the data from the cleanest days usually exhibit some
made-made influence. This average of 24-hour values for the 20% cleanest conditions
is used as representative of a clean background condition.
27
-------
Background conditions may be provided using any of the visibility parameters,
defined above, although the preferred method is to use the extinction coefficient, with
the hygroscopic components and other components dissociated, as in Equation 7.
Using the form of Equation 7 allows the effects of relative humidity to be applied
consistently to both the background aerosol and the aerosol attributable to a new
source. If only one of the other visibility parameters is available, without the
dissociation of the hygroscopic components, then that value would be used. However,
the relative humidity conditions that represent the average background will likely be
different than the condition being analyzed. The usual effect of this is that the relative
effect of sulfate and nitrate emissions is overstated under high relative humidity
conditions.
Calculating a change in extinction
The modeling techniques outlined in this recommendation will provide ground
level concentrations of visibility impairing pollutants. These concentrations are then
used to calculate the extinction coefficient due to these pollutants, using the
relationships outlined in Equations 4 and 5. The results of this are then compared
against the background extinction from Equation 7. The metric used for this
comparison is usually the change in deciview (Adv) from a "clean" background
condition. Thus, for a given background extinction, bback, and a source or sources
contribution to extinction of bsource, Adv is given by Equation 8.
= 10 ln
"back
These methods are embodied in the CALPOST program which post-processes
the concentrations from the CALPUFF air quality model. However, if another model is
used it is necessary to be able to perform these calculations separately. Even if the
CALPUFF system is used, it is sometimes more convenient to calculate changes in
visibility outside of the post-processor.
Example Problem
This example assumes that a dispersion model has been run and yielded
concentrations of S04= and soot (elemental carbon). From these concentrations the
analyst wishes to calculate a change in visibility.
First, we will consider the background visibility condition. If the background 24-
hour average visibility at the Class I area of interest has a combined sulfate and nitrate
extinction coefficient (bSN) of 1 .8 Mm"1 (neglecting the effects of relative humidity) and
an extinction coefficient from the other components (bdry) of 19.6 Mm"1, then the
background extinction (bback), expressed in the form of Equation 7 would be:
28
-------
bhack= 1.8 J(RH) + 19.6
In a typical analysis, the model will be run for an extended period, such as a
month, a season, annually, or for multiple years. This will produce a corresponding
number of 24-hour averaging periods, which will each need to be analyzed against the
background condition. In our example we are only considering one 24-hour averaging
period. For this example we will assume that the sources in the analysis contributed
0.218 ug/m3 of sulfate (S04=) and 0.05 ug/m3 of soot (elemental carbon). The first step
is to convert the mass of S04= to ammonium sulfate ((NH4)2S04), which is accomplished
by multiplying by the ratio of the molecular weights of (NH4)2S04 to S04=, which is
1.375. This yields a concentration of (NH4)2S04 of 0.30 ug/m3. This result is then
multiplied by the dry scattering efficiency of (NH4)2S04 (which is 3, from Equation 4),
yielding an extinction coefficient for the sulfate of 0.9 Mm"1; the relative humidity
adjustment has not yet been applied. In this example, our modeling does not require
any conversion of the mass of soot, so we need only to multiply the soot concentration
(0.05 ug/m3) by the extinction efficiency of elemental carbon (which is 20, from Equation
5). This yields an extinction coefficient of 1.0 Mm"1. Therefore, following the form of
Equation 7, the source contribution would be:
b = 0.9 -KRH) + 1.0
source. J ^ '
The relative humidity adjustment factor for this averaging period has not yet been
applied. Our example is based on a 24-hour average. The representative hourly RH
values for this day would need to be obtained. For each hour, the corresponding f(RH)
must be obtained from Figure 28 (or a corresponding table). These values are then
averaged together. Let us assume that for this day the average f(RH) is 3.4. With the
day-average relative humidity adjustment factor (f(RHJ) of 3.4, bbackwould be 25.72
Mm"1 (corresponding to a visual range of 152 km from Equation 1) and bsource would be
4.06 Mm"1. Using these values in Equation 8 would yield a Adv of 1.46.
These calculations would be repeated for each 24-hour average concentration in
the analysis, using the corresponding day-average f(RH). Background visibility
conditions may be given for seasons or months. The corresponding background values
should be used. A spread sheet program is suggested if the results are not processed
with the CALPOST processor.
Visibility Summary
The following list summarizes the steps necessary for conducting a visibility
analysis under this recommendation.
29
-------
• Consult with the appropriate regulatory agency and with the appropriate FLM
• Run an air quality model which yields ambient concentrations of visibility impairing
pollutants
• Obtain data on the background visibility conditions for use in the form of Equation
7,bext = bSNf(RH) + bdry
• Calculate the extinction coefficient for the source or sources being analyzed using
Equations 4 and 5.
• Apply appropriate and consistent relative humidity adjustment factors to both the
source contribution to light-extinction and the background light-extinction.
• Calculate a change in deciview (Adv) using equation 8.
3.3 Deposition Calculations
Estimates of atmospheric deposition are obtained by selecting the options in
CALPUFF to calculate and output the wet and dry fluxes of the pollutants modeled.
The units of the fluxes are in g/m2/s of the pollutant modeled (i.e., g/m2/s of HN03).
Generally AQRV analyses require values of total deposition (background plus modeled
impact) to be given in units of kg/ha/yr of an element, such as nitrogen (N) or sulfur (S).
Therefore, the modeled deposition flux of each of the oxides of sulfur or nitrogen from
CALPUFF must be adjusted for the difference of the molecular weight of their oxides
and the element and the various forms must be summed to yield a total deposition of
sulfur or nitrogen. This can be accomplished using a multiplier in CALPOST to do all of
the conversions. The CALPOST program will produce an average flux (i.e., annual
average), therefore, the average value must be multiplied by the number of seconds in
an hour and the total number of hours used in the averaging period for the total
deposition.
The wet and dry fluxes of S02, S04=, NOX, HN03, and N03" need to be calculated
and saved in a CALPUFF run. It is necessary to make a separate CALPOST run of the
wet and dry fluxes for each species modeled, normalizing each species by the
molecular weight of a common compound or element (usually S or N), converting the
units, and adjusting for the total number of averaging periods used in the CALPOST run
(i.e., 8760 for 1 year). Then the results of the sulfur CALPOST runs are summed and
the results of the nitrogen CALPOST runs are summed to yield a total deposition value
for sulfur and nitrogen, respectively. The following table indicates the multipliers to use
to correct for molecular weight differences and unit changes, as well as the correction to
go from a short-term flux to annual deposition.
30
-------
Deposition of:
S from SO,
S from SO,,
N from NOX
N from HN03
N from N03
Ratio Mol wt. of
Oxidant to S or N
0.50000
0.33333
0.30435
0.22222
0.22581
gtokg
ID'3
10'3
10'3
ID'3
ID'3
m2 to ha
104
104
104
104
104
sec to hours
3600
3600
3600
3600
3600
Number
of Hours
in
N
N
N
N
N
Multiplier (a in
CALPOST) x
N
1.800000E+04
1 .200000E+04
1 .095652E+04
8.000000E+03
8.129032E+03
For example, if CALPUFF was run for S02 and S04 , for one year, it would be
necessary to run four different CALPOST runs for wet and dry deposition of S02 and
S04=. For the S02 runs, the multiplier in CALPOST would be set to 1.8x104x8760 =
1.576x108, assuming a non-leap year with 8760 hours and to 1.2x104x8760 =
1.0512x108 for the S04= runs. If one was interested in the deposition over the month of
January, assuming a run length of 744 hours, one would use 744 in place of the 8760
to calculate the multiplier in CALPOST.
3.4 Assessing Air Quality Related Values - Background
It usually is not possible to assess Air Quality Related Values (AQRVs) in
isolation of the existing background stress. To illustrate this point, we consider two
examples. The first using a leaf injury model and the second discusses discerning
differences in visibility.
Larsen et al., (1983) presented a model of leaf injury, in which leaf injury was
seen to be a function of the hourly sulfur-dioxide concentration raised to the 1.845
power times the ozone concentration raised to the 1.271 power. The impact for a
series of hours was to be computed as a summation over each hour's computed effect.
What should be noted is the fact that the injury was not linear, but was proportional to a
power, in this case, greater than 1. We can simplify this to an equation of the form:
D = a ip
X = b + b1
where D = damage, a is some constant, p is a power (like those cited above), x is the
total concentration (or deposition, etc.), which is composed of a background
concentration, b, and an additional concentration, b', associated with one or more
sources. If p is not equal to 1, then it is not possible to assess the effects of the
background concentration, b, separately from the effects associated with one or more
sources, b'.
Example: Let p = 2, b = 10 and b' = 1, then D = 121 a. If one separately computes the
damage as a summation, as D = D(b) + D(b'), the result will underestimate the total
damage, as D = 100a + a = 101 a.
31
-------
The deciview, dv change associated with adding some new source's effects onto
an existing background haze effect can be computed as:
dv- 10 ln[<* + b\
where b is the extinction coefficient associated with the background and b' is the
extinction coefficient associated with the additional source or sources. It is worth noting
in the above equation that the human perception of a change in visibility is
fundamentally a function of the existing background condition. A one dv change is
computed whenever the ratio (b+b')/b = 1.11. Or stated in other terms, when b' is 11 %
of b, a one dV change in the visibility will be computed.
These two examples were provided to illustrate that assessing some Air Quality
Related Vales (e.g., as might be related to crop injury, or visibility effects) is
fundamentally tied to knowing the current stress (background) being exerted on the
system. Assessing the response of a resource (plant health, visibility) is related to the
cumulative effects of all the current existing stresses. Whether a model is used to
estimate the existing condition, or whether existing monitoring measurements can be
used to define the current stress (background), is decided for each case based on
available information.
32
-------
4.0 STUDIES AND FINDINGS
Sections 4.1 through 4.5 provide extended summaries of the investigations
conducted prior to the Sixth Modeling Conference. Following the Sixth Modeling
Conference, a series of investigations were conducted in response to the comments
received, and these results are summarized in Sections 4.6 through 4.9. In Section 4.6,
summaries are presented of several investigations in which surface concentration
values estimated by the CALPUFF puff dispersion model were compared with tracer
field data. These are presented to provide further information, as requested at the Sixth
Modeling Conference, for deriving conclusions on the performance of the
CALMET/CALPUFF modeling system in characterizing long-range transport. In Section
4.7, a summary is presented in which direct model-to-model comparisons are made
between CALPUFF and ISC simulations of point source release dispersion. These are
presented to provide evidence on the ability of CALPUFF to replicate the dispersion
results of ISC for steady-state meteorological conditions. Some coding deficincies were
detected in compariing the steady-state CALPUFF results with ISC, but the primary
lessons learned were what differences could be expected due to the fact that puff
models treat the sequence of meteorological events ('causality'). This information is
valuable since CALPUFF was to be used to simulate all sources (even though for
some, the transport may be less than 50 km) in a long-range transport assessment.
Section 4.8 summarizes results available towards developing a new screening
technique that might be used to see if it is worthwhile or needful to develop a full wind-
field puff-dispersioin analysis using the CALMET/ CALPUFF modeling system. It was
evident at the Sixth Modeling Conference that use of the ISC plume dispersion model to
develop screening estimates of long-range transport impacts was not providing much
help. Section 4.9 summarizes studies that have recently been completed of further
enhancements and refinements, some of which are implemented in Version 5.0 of the
CALMET/CALPUFF modeling system. Not all of these enhancements have been fully
tested, but it was felt desirable to summarize that which was known.
4.1 MESOPUFF II Implementation Assessment
A case study was conducted to apply the MESOPUFF II air quality modeling
system following IWAQM Phase 1 interim recommendations (U.S. EPA, 1993). This
study would identify and summarize the decisions made, would record and summarize
the resolution process for these decisions, and would provide a written record of the
resources used to complete the effort. The objective was to learn by experience where
the difficulties are in the process of conducting such an analysis, and when possible, to
provide a means for resolving these difficulties. It was not an objective to provide a
meaningful assessment of PSD, NAAQS or AQRV impacts for the Class I areas
considered in the study. A complete description of the study results is presented in
U.S. EPA (1995a). As part of this study the following tasks were carried out:
The MESOPUFF II model and associated processors were tested using the
example problem intended for Support Center for Regulatory Air Models
33
-------
bulletin board (SCRAM BBS) distribution. The SCRAM BBS example
problem computer files were evaluated and some suggested improvements
were implemented.
A five-year meteorological data set suitable for input to the MESOPUFF II
model was developed for a multi-state area surrounding Shenandoah
National Park (SNP), including the James River Face (JRF) wilderness area.
Demonstration model simulations were performed using three years of the
developed five-year meteorological data set for the assessment of visibility,
acidic deposition, and PSD increments for a set of real sources in the states
surrounding Shenandoah National Park.
Model simulations were performed to test the sensitivity of concentrations to
the distance between sources and receptors using a set of "pseudo" sources
placed in successive rings around Shenandoah National Park.
As discussed in the summary report for this project (U.S. EPA, 1995a), on
several occasions significant departures were made in conducting this study from that
which would be expected if a realistic assessment were to be developed. For instance,
the source inventory considered only some of the states surrounding the Shenandoah
National Park and the James River Face wilderness area, and thus is incomplete. To
conserve resources, the sources were consolidated into ten surrogate sources for the
purposes of this study. These departures allowed the emphasis of the project to be
focused on a critique of the process and resource needs, which were the primary study
objectives.
A realistic assessment, would require that all important sources be modeled
(without consolidation). If the modeling objective is to determine PSD impacts, then all
relevant sources that consume PSD increment must be considered. If one desires to
determine the impact of a single new (or modified) source, then the PSD increment
from the new source must be added to all pre-existing PSD sources. It would be
possible to model the impacts from a single source and then add those impacts to prior
MESOPUFF II results, assuming the prior results were available. If not, it would be
necessary to model all relevant PSD sources to assess the total PSD increment
consumed.
34
-------
420
620
40 .
4360
MESOPUFF II Computational Domain
Figure 2. Display of the MESOPUFF II computational domain, showing positions of
all PSD sources compiled for the demonstration assessment. Crosses represent all
27 original PSD sources, while circles denote the ten final aggregated sources used
for the demonstration analyses. The source characteristics for the ten aggregate
sources (numbered) are given in Table 2. One unit along left and bottom axes
equals 20 km. Top and right axes are the UTM coordinates in km.
Demonstration Results
In the MESOPUFF II modeling for this project, only one run was performed (for
three years) using sources beyond 50 km of Shenandoah National Park. The
MESOPUFF II results (for one month) were then added to ISCST2 results to
demonstrate the integration process. For the demonstration study, sources were to be
located as far as 200 km from SNP. All sources and receptors to be modeled must be
contained within the computational domain. Puffs are not tracked after they leave this
grid. To avoid underestimating concentrations by immediately losing puffs from near-
boundary sources or missing short-term recirculation events at near-boundary
receptors, all sources and receptors should be at least 20-50 km from the boundaries of
35
-------
the computational grid. Provision for a 80 km buffer zone on all sides of this source
region yielded a computational domain of 30 by 30 grid points (Figure 2), with a grid
spacing of 20 km.
For the demonstration study, existing sources that began operation since the
beginning of the PSD program would be modeled. For practical considerations and
ease of communications between affected states, the sources were limited to those
within EPA Region III, comprising the states of Delaware, Maryland, Pennsylvania,
Virginia and West Virginia, and the District of Columbia. Small PSD sources, with
emissions of both S02 and NOX less than 5 g/s, were excluded. This resulted in a set of
27 sources (indicated by crosses in Figure 2). To reduce run times the set of 27
sources were consolidated into 10 sources. Table 3 shows the final set of 10
consolidated sources, the facilities that are included in each, and the location and
emission parameters. Figure 2 also displays the locations of these ten sources
(indicated by circles) in relationship to all 27 original PSD sources.
Table 3. Final condensed set of sources modeled with MESOPUFF II for the
demonstration assessment. HS, DS, VS, and TS are the source's stack height, stack
diameter, effluent exit velocity, and effluent exit temperature, respectively.
1
2
3
4
5
6
7
8
9
10
Sources
PEPCO II, Patawmack
Warrior Run
Ogden-Martin, SEO Birch
N. Branch, Mettike
LG&E AltaVista, Multitrade
Mecklenburg, Old Dominion
Doswell, Cogentrix-Richmond, Cogentrix-
Dinwiddie, LG&E Hopewell
Brandon
Cambria Cogen, Colver PP, Ebensburg
P.M. Glalter, Harrisburg, Lancaster, Solar
Turbine, York Co.
UTMX
(km)
803.6
693.6
823.6
639.6
653.6
711.6
819.6
885.2
689.6
861.6
UTMY
(km)
4341 .0
4385.0
4265.0
4349.0
4109.0
4067.0
349.6
4345.0
683.6
4439.0
HS
(m)
48
82
106
76
60
109
70
187
86
67
DS
(m)
8.77
3.75
3.24
4.69
4.44
7.11
5.35
6.71
5.08
6.60
VSI
(mis)
36.6
23.6
10.0
21.3
21.4
18.6
14.0
27.3
21.3
15.4
TS
(K)
557
398
372
389
407
331
352
413.
418
438
S02
(g/s)
348.2
54.8
48.0
112.1
24.7
194.6
252.1
1893.6
287.8
14.1
S04=
(9/s)
15.7
2.5
2.2
5.0
1.1
8.8
11.3
85.2
13.0
0.6
NOX
(g/s)
230.4
26.0
151.9
83.6
42.6
425.1
426.4
630.8
155.8
212.3
For this study, PMio increments were calculated as described in the Phase 1
recommendations. Modeled concentrations of sulfate and aerosol nitrate were
converted to ammonium sulfate and ammonium nitrate, and summed to estimate PMio
increments. The PSD increments were calculated for each receptor, and the highest
values within SNP and JRF were identified. Since there are no standards or mandated
increments for AQRVs, it was necessary to define AQRV criteria before postprocessing
could begin. For visibility, incremental extinction coefficients were computed from 3-
hour average modeled sulfate and nitrate concentrations using relative humidity data
36
-------
and the equation provided in Appendix B of the Phase 1 recommendations. The
maximum 3-hour extinction resulting from modeled emissions over all receptors was
compared to measured total extinction from the SNP IMPROVE monitoring site. The
number (and percentage) of 3-hour periods for which the maximum extinction
represents 10 percent or more of the measured extinction for the 90th percentile
cleanest day was reported. Deposition impacts were calculated for each receptor for
total sulfur (S02 plus sulfate, expressed as S) and total nitrogen (NOX plus HN03 plus
nitrate, expressed as N). Deposition impacts were expressed as the cumulative annual
sum of wet and dry deposition, in units of kg/hectare.
Table 4. PSD and AQRV parameters calculated from MESOPUFF II demonstration
assessment.
Parameter
S02
NOX
PMio
Extinction
Total S
Total N
Averaging Period
Annual (ug/m3)
24-hour (ug/m3)
3-hour (ug/m3)
Annual (ug/m3)
Annual (ug/m3)
24-hour (ug/m3)
3-hour (% of year)
Annual (kg./Ha)
Annual (km/Ha)
Shenandoah NP
1988
1989
1990
James River Face
1988
1989
1990
PSD Increments
0.23
3.92
16.56
0.12
0.08
1.24
0.32
5.03
19.62
0.15
0.12
3.32
0.21
3.15
9.55
0.13
0.07
1.31
0.08
1.39
5.38
0.05
0.05
0.85
0.10
2.93
7.35
0.07
0.07
1.25
0.06
1.61
3.65
0.04
0.04
1.31
Visibility
19.9
27.9
22.4
7.3
8.2
6.8
Deposition
0.38
0.09
0.36
0.12
0.28
0.07
0.11
0.04
0.15
0.16
0.08
0.03
Allowable Class I
Increment
2
5
25
2.5
4
8
Extinction = percent of 3-hour periods for which incremental extinction is greater than 10 percent of clear-day extinction levels.
Table 4 displays a summary of calculations from three years of MESOPUFF II
output (1988-1990). Results are given for receptor groups in both SNP and JRF.
There is considerable year-to-year variations in all of the simulated impacts, typically of
order 20 to 25%. As for concentrations of criteria pollutants, annual averages of S02,
NOX, and modeled PMio in both Class I areas were predicted to be small fractions of
the total allowable Class I increments for all years modeled. Short-term concentration
increments were predicted to approach the allowable Class I increments, particularly at
SNP. The second highest 3-hour and 24-hour average S02 concentrations approach or
exceed the allowable Class I increments, while the second highest 24-hour average
modeled PMio is between an eighth and one-half the allowable limit. Also note, that
37
-------
whereas the summer 1988 was characterized by widespread high pollution levels
throughout the eastern U.S. (particularly for ozone), all modeled concentration,
extinction, and deposition predictions were distinctly higher during 1989 than during the
other two modeled years at both Class I areas.
The extinction measure reported in Table 4 corresponds to the percent of each
year in which calculated sulfate plus nitrate concentrations lead to extinction levels
more than 10 percent above the defined "clean" background levels for SNP. According
to these calculations, it is predicted that secondary particulate matter from the PSD
sources modeled in this analysis lead to such conditions between 20 and 28 percent of
the year for SNP, and about 7 to 8 percent of the year at JRF. At SNP, maximum 3-
hour extinction for the years 1988, 1989, and 1990, were 0.0507, 0.3348, and 0.1345
km"1, respectively. Using a simple standard estimation procedure for clear-sky visual
range (3.0 divided by extinction in km"1), the extinction levels computed for SNP
translate to minimum visual ranges of 59, 9, and 22 km, respectively, for each year
modeled. The maximum extinction levels are quite high when one considers that these
estimates were calculated solely as a result of particulate matter generated by these
emission sources alone, and that other S02 and NOX sources, and natural and
anthropogenic sources of organics and dust, were not taken into account.
Source-Receptor Separation Distance Sensitivity Study
A set of 24 MESOPUFF II simulations were performed with sources at varying
distances from Shenandoah National Park (SNP) to provide some insight into the
relationship between distance from SNP and potential PSD and AQRV impacts (U.S.
EPA, 1995a). Four months of one year (representing each season of a year) were
modeled under six different source scenarios. Each source scenario included a
number of identical hypothetical point sources placed on a ring at a constant distance
from the "spine" of the SNP. The source rings were established at 50, 100, 125, 150,
175, and 200 km from the spine (Figure 3).
The distribution of hypothetical point sources around each ring was determined
such that the linear density of sources was held as constant as possible. The emission
rates and source emission characteristics were set to be similar to those listed in Table
3. The emission rate and stack parameters used for the hypothetical sources are
summarized in Table 5. The stack parameters are average values from the PSD
source data in Table 3. Emission rates of primary sulfate were specified at 3 percent of
S02 rates, with an additional 1.5 factor to account for the larger sulfate molecular
weight. Source characteristics of each ring are summarized in Table 6, and the
locations of the sources are shown in Figure 3.
38
-------
Table 5. Emission rates and stack parameters for the idealized sources.
Stack Parameters
S02 Emission Rate (g/s)
S04 Emission Rate (g/s)
NOX Emission Rate (g/s)
Stack Height Im)
Stack Diameter (m)
Stack Gas Exit Velocity (mis)
Stack Gas Exit Temperature (K)
Value for Idealized Sources
181.0
8.1
135.0
81.0
3.97
20.7
425.0
A total of 24 MESOPUFF II runs were made, corresponding to six distances for
each of the four months. MESOPUFF II output for the ring source analysis consisted of
3-hour average concentrations, and wet and dry fluxes of all species. 24-hour averages
and monthly averages were produced for NOX, nitric acid, sulfate (S04=), nitrate (N03")
and modeled PMio, S04= and N03" concentrations output by MESOPUFF II were
multiplied by factors of 1.38 and 1.29 to convert to ammonium sulfate, (NH4)2S04, and
ammonium nitrate, NH4N03, respectively (these conversions are discussed further in
Section 3). Following the Phase I recommendations, the sum of ammonium sulfate and
ammonium nitrate was reported as modeled PMio. Total sulfur deposition was
calculated by summing wet and dry deposition fluxes for S02 and S04= over the month.
Conversion factors were applied to convert S02 and S04= to a sulfur basis, and to
convert 3-hour average fluxes to 3-hour cumulative deposition. Total nitrogen
deposition was calculated by summing wet and dry deposition fluxes for NOX, HN03,
and N03" over each month. Conversion factors were applied to convert NOX, HN03,
and N03" to a nitrogen basis, and to convert 3-hour average fluxes to 3-hour cumulative
deposition. (For further discussion of the computations of total S and N deposition, see
Section 3.3). An important AQRV associated with PMio is visibility. MESOFILE was
not capable of producing visibility estimates as outlined in the Phase I
recommendations, where extinction is a function of relative humidity. Thus, visibility
impacts were not assessed for the ring source analysis.
39
-------
420
620
820
^ 4360
- 4160
11 21
MESOPUFF II Computational Domain
3960
Figure 3. MESOPUFF II computation domain for ring source analyses, showing the
location of the ring sources. One unit along left and bottom axes equals 20 km. Top
and right axes are the UTM coordinates in km.
Figure 4a shows the results obtained for the monthly average S02 concentration
for each of the four months as a function of source distance from SNP. The nearly
linear decrease as a function of distance from SNP seen for these results is typical of
that seen in the highest 24-hour average S02 results. As might be expected for a
primary pollutant, the highest impacts occur for the 50-km source ring. The highest
monthly average concentration is 0.45 ug/m3, and decreases by a little over a factor of
2 for sources located on the 200 km ring. The maximum 3-hour and 24-hour S02
concentrations were also for the sources on the 50 km ring and were 20 ug/m3 and 2.8
ug/m3, respectively. The computed maximum 3-hour S02 concentration for sources on
the 50 km ring is close to the allowable 3-hour increment of 25 ug/m3. The computed
maximum 24-hour S02 concentration for sources on the 50 km ring is slightly more than
half the allowable Class I PSD increment of 5 ug/m3.
40
-------
Table 6. Source ring characteristics.
Distance from
SNP (km)
50
100
125
150
175
200
Number of
Sources
4
6
7
8
9
10
Distance Between
Sources (km)
138
144
146
148
149
150
Total Emissions (g/s)
S02
724
1086
1267
1448
1629
1810
NOX
540
810
945
1080
1215
1350
Figure 4b shows the highest monthly average concentrations for NOX. (Plots of
shorter-term NOX concentrations were not prepared because there are no short-term
PSD standards for NOX.) Comparison of Figure 4a to Figure 4b for monthly average
S02 illustrates the faster rate of chemical decay for NOX. For sources on the 50 km
ring, NOX concentrations are generally on the order of 60 percent of the S02
concentration for January, April, and October, reflecting the ratio of the emission
strengths of S02 and NOX in the ring source input files. NOX concentrations at 50 km
are 40 percent of S02 concentrations for July because NOX reacts much more rapidly
than S02. For sources on the 200 km ring, NOX concentrations are on the order of 50
percent of the S02 concentration for January, April, and October, and 30 percent of S02
concentrations for July. The highest monthly average NOX concentration was slightly
less than 0.3 ug/m3, or about 11 percent of the allowable increment of 2.5 ug/m3 for NOX
annual average in Class I areas. The annual average will be lower than the highest
monthly average. Thus, much greater NOX source strengths than those used in this
analysis would be needed for NOX concentrations to approach the allowable increment.
Figure 4c presents the highest monthly average modeled PMio concentrations as
a function of source distance from SNP. These impacts show a completely different
pattern than S02 or NOX, in that there is no clear decrease with source ring distance.
This pattern is expected as, unlike S02 or NOX, sulfate and nitrate are actively being
formed during transport downwind. The seasonal effect is very pronounced for the
monthly average modeled PMio, with July concentrations more than twice those in
January, and April and October falling somewhere between the two. Modeled sulfate
and nitrate
41
-------
(A)
0.5
1
jj 0.4-
0.3-
0.2-
25 50 75 100 125 150 175 200
Distance from SNP (km)
• Jan A Apr » Jul a Qct
(B)
•
•
D 1
D a
* i
* •
A A
25 50 75 100 125 150 175 200
Distance from SNP (km)
Jan A Apr * Jul o Oct
(C)
£ 0.05
1
O)
I
0
25 50 75 100 125 150 175 200
Distance from SNP (km)
• Jan A Apr * Jul D Oct
Figure 4. Highest monthly concentrations of (A) S02, (B) NOX, and (C) PMio (ug/m3)
from ring sources for January, April, July and October, 1988.
42
-------
concentrations were nearly equal for all months but July. This is contrary to measured
PMio data for SNP and other locations in the eastern U.S., where sulfate
concentrations are typically three times as large as nitrate concentrations. Two factors
are likely to be responsible for this apparent inconsistency. The high background
ammonia concentration of 10 ppb used as a default value in the MESOPUFF II analysis
will result in an overestimate of aerosol nitrate concentrations. However, the lack of a
parameterization of rapid in-cloud sulfate formation in MESOPUFF II may lead to an
underestimate of sulfate concentrations.
The highest monthly total S deposition of nearly 0.05 kg/hectare occurred for
July for the 50 km ring. Modeled S deposition decreased with source distance from
SNP, although the decreasing trend was very weak for October. For all sources, the
highest S deposition occurred in July, and the lowest in January. Dry S02 deposition
accounted for most of the modeled S deposition in all months. Deposition of S04= was
small for all months except July. The main reason for this is that most of the total sulfur
remained in the form S02 at all distances modeled. Wet deposition shows greater
random variability than dry deposition. This makes sense, as wet deposition requires
both a puff and precipitation to be present at the same time at a given receptor. Thus,
trends in wet deposition with source distance are somewhat obscured by random
variability.
The highest monthly total N deposition of 0.013 kg/hectare occurred for the 50
km ring for July. Modeled N deposition decreased slowly with source distance from
SNP. For all sources, the highest N deposition occurred in July, and the lowest in
January. Total N deposition was less than half the modeled total S deposition for all
months and distances. MESOPUFF II assumes zero wet deposition for NOX. The
majority of modeled N deposition in July is due to dry deposition of HN03. In the other
months, NOX dry deposition is important, and N03" deposition can be important as well.
Considering the high solubility of HN03, the modeled wet deposition of HN03 appears
surprisingly low.
The equilibrium between HN03 and N03" affects the results presented here, both
for the modeled PMio concentrations and the nitrogen deposition. The ring source
analysis utilized the default value for background ammonia of 10 ppb. This value is
likely to be too high, especially for winter. When ammonia concentrations are high, the
nitrate equilibrium favors the formation of aerosol nitrate. As a result, modeled aerosol
nitrate values may be too high. For N deposition, it is less clear what the effect of high
background ammonia would be. Dry deposition is faster for HN03 than it is for N03",
but wet deposition is faster for N03". Snow is assumed in MESOPUFF II to scavenge
particles, but not gases. If background ammonia is high, N03" deposition will be
overestimated and HN03 deposition will be underestimated. The net effect on N
deposition may be small.
43
-------
4.2 Revisions to CALMET and CALPUFF
In the course of completing the Phase 1 recommendations, IWAQM had become
aware of the CALMET/CALPUFF modeling system (Scire et al., 1990ab), which was
actively under development. Building from lessons learned from the
MESOPAC/MESOPUFF II modeling system, the CALMET/CALPUFF modeling system
was a Lagrangian puff model designed to include: 1) the capability to treat time-varying
point and area sources, 2) suitability for modeling domains from tens of meters to
hundreds of kilometers from a source, 3) predictions for averaging times ranging from
one-hour to one year, 4) applicability to inert pollutants and those subject to linear
removal and chemical conversion mechanisms, and 5) applicability for rough or
complex terrain situations. The CALMET meteorological processor was designed to be
compatible with both CALPUFF and a photochemical grid model, called CALGRID
(Scire et al., 1989). Even though the current focus of IWAQM was on puff model
simulations, the Phase 3 work was anticipated to include grid modeling. Hence the
compatibility of CALMET to a grid model was considered advantageous.
There were two areas where IWAQM felt further enhancements were needed.
With a view towards allowing one model to be used for all sources (which might include
sources-receptor distances of less than 50 km), the first area for enhancement was to
include within CALPUFF dispersion additional algorithms, so that CALPUFF simulation
results would be consistent with ISC and CTDMPLUS (Perry et al., 1989) modeling
results for steady-state meteorological conditions. The ISC plume dispersion model is
recommended in the Guideline for use in gently-rolling terrain, and the CTDMPLUS
plume dispersion model is recommend in the Guideline for use in complex terrain
where plume impaction on elevated isolated terrain features is likely. Both ISC and
CTDMPLUS are limited to source-receptor distances of less than 50 km. The IWAQM
has concluded that one of the most challenging issues for long-range transport is the
characterization of the time-varying three-dimensional wind field. Therefore, the second
area for enhancement to the CALMET/CALPUFF modeling system was to include
provisions within CALMET to allow use of mesoscale meteorological modeling results
created using data assimilation techniques, for example Stauffer and Seaman (1989)
and Stauffer etal., (1990).
At the time of the Sixth Modeling Conference, there were only preliminary
sensitivity testing results to show that the modifications to CALPUFF would be
successful in replicating ISC. There were no comparison results available showing
consistency between CALPUFF and CTDMPLUS. Comparisons of CALPUFF with ISC
are presented in Section 4.7. In following discussion, we summarize the modifications
that were made to CALMET to expand its use to long-range transport applications.
The wind field module in CALMET is based on the Diagnostic Wind Model
(DWM). In anticipation of using CALMET and CALPUFF for long-range transport
distances, a series of modifications were made (U.S. EPA, 1995b). These included
options to use a spatially variable initial guess field based on observations or results
44
-------
from coarse-grid mesoscale meteorological modeling analyses, optional use of Lambert
conformal coordinates for sources and receptors versus Universal Transverse Mercator
(UTM) Cartesian coordinates, and (as mentioned before) provisions to allow processing
of mesoscale meteorological modeling results created using data assimilation
techniques.
The DWM in CALMET uses a two step procedure in developing the final wind
fields. An initial guess field is developed based on a domain-average wind profile, and
this domain-average profile of winds is adjusted for terrain effects and divergence
minimization to produce a "Step 1" wind field. The second step in the processing of the
wind field is the introduction of the observational data into the terrain-adjusted Step 1
wind field. As originally configured, the initial guess domain-average profile was
constant over the domain. For long-range transport analyses, it is easily conceivable
that terrain features (such as a ridge, or a deeply cut river gorge that turns sharply
within the computational domain) would invalidate the appropriateness of using an initial
guess wind profile that is everywhere the same over the domain (i.e., homogeneous).
To address this, options were added to CALMET that allow spatially variable winds as
an initial guess field. The spatially variable winds are computed using an inverse-
distance interpolation from the available wind profile observations. The provision to
allow use of a Lambert conformal projection versus a flat Cartesian mapping of
receptors and sources was perfunctory, but necessary, in anticipation of analyses that
might include transport distances where the curvature of the earth might become a
significant factor in the analyses (say computational domains larger than 200 km by 200
km).
The adaptations needed to allow use of meteorological wind fields as analyzed
by sophisticated mesoscale meteorological models (hereafter referred to as FDDA-MM
data) involved more than simply providing a new data input option. Four Dimensional
Data Assimilation (FDDA) and the development of FDDA-MM data are discussed
further in Section 4.4. The representativeness of the fine-scale observations (which
can be viewed as point-value observations) as compared with winds derived from
FDDA-MM analyses (which can be viewed as grid-average observations) was expected
to depend on such factors as the height above the surface, subgrid-scale terrain
variations, and the ratio of the input FDDA-MM data grid size to the output grid size of
the CALMET analyses. For example, the FDDA-MM results having a grid spacing of
80-km will not reflect potentially important local features of the surface flow field
induced by terrain variations (e.g., in vicinity of the Shenandoah National Park or the
Columbia river gorge) which can not be resolved by a grid resolution of 80 km. On the
other hand, the point-value observations in such areas do not necessarily represent
larger-scale flow fields as well as the FDDA-MM data fields. Therefore, IWAQM
investigated development of a weighting factor based on the subgrid-scale terrain
variations, that could be employed to blend the FDDA-MM data fields into the network
of available surface and upper air observations. The developmental work to define the
blending weights is described in the summary project report U.S. EPA (1995b). The
45
-------
detailed description of how to compute the weights is provided in the current CALMET
user's guide.
The decision as to where to introduce the FDDA-MMwinds in the CALMET
processing involves judgment. Basically, the finer the grid resolution used in developing
the FDDA-MM winds the more reasonable it is to bring these data directly in as
observations in developing the Step 2 winds. The coarser the grid resolution used in
developing the FDDA-MM winds, the more reasonable it is to bring in these data to
initialize the Step 1 analyses, and allow the diagnostic wind model of CALMET to
develop the local terrain effects.
it i i j i iiiiLH i\ ill i i i i n in i\n mil mill
Figure 5. The MM4 domains for the 54-km grid (outer box) and the 18-km grid (inner
box).
46
-------
Table 7. Summary of statistical comparisons of wind fields.
80-km MM4 Interpolated
CALMET-L Observations only
CALMET-S Observations only
CALMET-L MM4 as initial guess field
CALMET-S MM4 as initial guess field
CALMET-L MM4 as Step 1 field
CALMET-S MM4 as Step 1 field
CALMET-L MM4 as observations
CALMET-S MM4 as observations
54-km Results
R2
0.961
0.789
0.542
0.926
0.932
0.962
0.962
0.960
0.959
RMSE
0.992
2.211
3.220
1.378
1.316
0.992
0.993
1.020
1.021
18-km Results
R2
0.958
0.858
0.814
0.920
0.922
0.959
0.959
0.952
0.952
RMSE
0.949
1.774
2.039
1.328
1.300
0.932
0.938
1.011
1.011
R2 = correlation coefficient, and RMSE = root mean squared error.
L = large radius of influence settings, and S = small radius of influence settings.
R2 and RMSE are computed on a concatenation of the time series of the east-west (u) and north-south (v)
components of the wind (each of length N) into one time series (of length 2N).
The radii used were defined as: Small Large
R1 50 km 500 km
R2 100 km 1000 km
Rmaxl 100 km 500 km
Rmax2 200 km 1000km
RmaxS 1000 km 1000 km
where R1, R2 = the distances at which the Step 1 wind and observations have equal weight in the surface layer
(R1 ) and the upper layers (R2), Rmaxl , Rmax2 = the maximum radius of influence of observations in the surface
(Rmaxl) and upper layers (Rmax2), and RmaxS = the over-water radius of influence.
To investigate the effectiveness of the terrain weights developed, the CALMET
diagnostic wind field model was used to analyze two episodes, one summer episode
(August 1 -6, 1988) and one winter episode (December 3-10, 1988). The summer
episode was characterized by light wind, stagnating conditions. The winter episode was
characterized as an active period that included the passage of a front and low-pressure
system through the domain. Penn State Mesoscale Meteorological (MM4) results were
available employing four dimensional data assimilation (Stauffer and Seaman, 1989) for
both episodes, for three different grid resolutions, 18-, 54- and 80-km. Figure 5 depicts
the domains over which comparisons were made of CALMET simulated wind fields,
developed for comparison with the MM4 54-km winds (outer box in Figure 5) and with
the MM4 18-km winds (inner box in Figure 5). The CALMET winds were developed
using the MM4 80-km winds as input (at various stages in the CALMET processing),
47
-------
and CALMET winds were developed using only the hourly surface weather
observations and twice-daily upper air observations from the National Weather Service.
Table 7 summarizes the comparison results obtained by using the available 80-
km MM4 winds in various ways with CALMET to develop either 54- or 18-km gridded
wind fields. As a further comparison, the hourly 80-km MM4 winds were linearly
interpolated directly to the 54- or 18-km grid resolution. In general, the introduction of
the 80-km MM4 winds improves the ability of CALMET to reproduce the reference 54-
and 18-km MM4 wind fields. Slightly better agreement was achieved when the 80-km
MM4 winds were brought in after the diagnostic terrain adjustment procedures (i.e., as
the Step 1 wind fields or as "observations"). This is conjectured to occur due to the fact
that 1) the 80-km MM4 are already close to the 54- and 18-km MM4 results, and 2) the
CALMET diagnostic adjustments may duplicate terrain effects that were already
accounted for in the development of the 80-km MM4 winds. The similarity of the
interpolated 80-km MM4 winds to the 54- and 18-km MM4 winds (as evidenced by the
close agreement achieved by simple linear interpolation) suggest that there might not
be significant new terrain effects between the 80-, 54-, and 18-km scales for this region
of the United States.
4.3 Trajectory Comparisons
As discussed in the previous section, modifications were made to CALMET in
anticipation of using CALMET and CALPUFF for long-range transport distances (U.S.
EPA, 1995b). It was anticipated that use of FDDA-MM winds would improve
CALPUFF's characterization of trajectories of dispersing pollutants. To investigate this,
two trajectory studies were conducted: 1) a numerical simulation study, and 2) a
comparison with regional-scale observations of trajectories.
Numerical simulation study
Trajectories were computed from four release locations at three levels (10m,
200 m, and 400 m) for each of the wind fields discussed in Section 4.2 for the summer
episode (U.S. EPA, 1995b). Trajectories were generated at each location every 4, 6
and 12 hours from the beginning of the simulation, for up to 24 hours before the end of
the simulation. A statistical analysis was conducted on the trajectories to assess the
effect of the different wind fields.
Figures 6a and 6b show the trajectories at an elevation of 10 m on the 54 km
grid for each of the four locations. The trajectories developed using the MM4 54 km
gridded meteorology directly are shown in both figures with small "x's." The trajectories
developed using the CALMET generated 54 km winds are shown in both figures with
48
-------
(A)
n7,54
378 810 1242
SOUTH
1674
-621
(B)
NORTH
-54 378 810 1242 1674QQQ
999| i i i I i i I | i i i I i ILI-M i i i M i | i i i i/visi | ij999
i i I/TI i i I i i i i i i i I i I 521
-54
378
810 1242
SOUTH
Figure 6. Trajectories at 10 m on the 54 km grid produced by CALMET (A) using the
large radius of influence and observations only, and (B) using the large radius of
influence and introducing the MM4 80 km winds in the Step 1 initial guess. CALMET
results are shown with "o" and trajectories developed directly from MM4 54 km winds
are shown with "x". The starting date was August 1, 1988 at 7 AM. The small
numbers indicate hours after release.
small "o's." Figure 6a illustrates the results obtained by CALMET with observations
only, and Figure 6b illustrates the results obtained by CALMET by introducing the MM4
49
-------
80 km winds in the development of the Step 1 initial guess wind fields. The most
striking differences between the two figures is for the trajectory developed from the
Knoxville release location. The CALMET winds (based on observation only) show a
light southwest flow, whereas the winds resulting from introducing 80 km MM4 winds to
CALMET show the corrent movement towards the north and northeast.
Trajectory statistics were computed from each release time for each site and
three levels. In general, the introduction of the 80 km MM4 winds into CALMET to
develop either 54 km or 18 km gridded wind fields significantly improved the
comparisons with the trajectories developed from the 54 km and 18 km MM4 wind fields
directly, versus using only the routine hourly weather observations and twice-daily
upper air observations as input to CALMET. There was a slight difference to be seen in
using a large or small radius of influence in developing the trajectories with CALMET
when only observations were input to CALMET. When the MM4 data are used as input
to CALMET, the choice of the radius of influence appears statistically to have no effect
on the trajectory comparisons.
CAPTEX comparisons
One of the objectives of the CAPTEX comparisons (Irwin et al., 1996) was to
assess whether use of mesoscale dynamic wind fields developed using Four-
Dimensional Data Assimilation (FDDA), exhibiting improved spatial and temporal
resolution versus typical mesoscale wind fields determined diagnostically from the
available hourly surface and twice-daily upper air observations, would improve the
quality of the characterization of the transport and dispersion. Results were generated
for CAPTEX releases 3, 5 and 7.
The Cross-APpalachian Tracer Experiment (CAPTEX) is a unique series of
tracer releases, which besides testing a particular tracer technology, was conducted for
the purpose of providing data to evaluate and improve computer models of pollutant
dispersion and to provide insight into the mechanisms involved in long-range transport
and dispersion (Ferber et al., 1986). A three-hour ground-level release of
perfluoromonomethylcyclohexan (C7H14, PMCH) was made five times near Dayton,
Ohio and twice from near Sudbury, Ontario when winds were expected to transport the
tracer over the ground-level sampling network. Samplers were operated at 86 sites in
Ohio, Pennsylvania, New Jersey, New York, New England and southern Canada at
distances from 300 to 1100 km from the release site. Air concentrations were collected
for 3- and 6-hour durations for several days following each release.
Meteorological data available for use in developing the CAPTEX wind fields
consisted of 122 National Weather Service (NWS) surface locations reporting hourly
and 13 upper-air locations reporting twice-daily (0000 GMT and 1200 GMT) throughout
the region. Furthermore, mesoscale wind fields developed using FDDA were available
on an 80-km grid. Three wind field models were used to obtain a gridded field of
meteorological data with a horizontal resolution of 18-km: MESOPAC II, CALMET, and
50
-------
CALMET using the mesoscale wind fields as STEP-1 inputs. In the discussion to
follow, the latter modeled wind field is referred to as CALMET/MM4.
Version 8 of the Penn State/NCAR Mesoscale Model - Generation 4 (MM4) was
used to develop profiles of atmospheric wind, temperature, and moisture. The MM4 is
a primitive equation, mesoscale, hydrostatic Eulerian modeling system of the
atmosphere (Anthes et al., 1987). The vertical layers are terrain following. FDDA is
incorporated as Newtonian nudging by adding to the general momentum and
thermodynamic equation a "Force-Restore" term, to effectively nudge the numerical
solution towards the observed data (Stauffer and Seaman, 1989; Stauffer et al., 1990).
Nudging was applied for the east-west and north-south wind components, temperature
and mixing ratio at all levels, except for temperature at the surface and at levels aloft
within the lowest 6 layers of the analyses (top of sixth layer is typically 1500m). The
horizontal grid spacing of the MM4 simulation was 80 km in both dimensions, with a grid
array size of 84 by 55 centered on 90°W longitude and 40°N latitude to cover the
domain of the CAPTEX releases.
MESOPUFF II was driven by each of the modeled wind fields to produce a set of
three simulations of ground-level concentrations for each of the three CAPTEX releases
(9 simulations in all). CALPUFF was driven by two of the modeled wind fields,
CALMET and CALMET/MM4, to produce a set of two simulations of ground-level
concentrations for each of the three CAPTEX releases (6 simulations in all).
In both the MESOPUFF II and CALPUFF simulations, chemical transformations,
dry deposition and wet removal were not modeled because PMCH was assumed to be
inert and non-depositing. No attempts were made to optimize the choice of model
options within CALPUFF, such as the dispersion coefficients, mode of incorporation of
MM4 data, and meteorological vertical layer structure. In effect, CALPUFF was run in a
mode designed to make it most like MESOPUFF II, in order that the effects of different
wind fields and transport characterization could be identified. MESOPUFF II and
CALPUFF differ in the way transport winds are computed for each puff. MESOPUFF
II uses a two-layer wind field, the lower layer for the transport of puffs within the mixed
layer, and an upper layer field for puffs above the mixing height. For a surface release,
as in CAPTEX, MESOPUFF II will always use the lower layer (mixed-layer averaged)
wind field. CALPUFF internally computes for each sampling step, a transport wind
averaged over the depth of the puff from the multi-layer winds provided to it from
CALMET. As the puff grows in the vertical, the depth through which the wind is
averaged is increased.
The distance from the nearest to furthest receptor was approximately 800 km
and there where 86 receptor locations. The average distance between the receptors
was roughly 86 km = [(800)2/86]1/2, which meant that typically there were 4 to 6
receptors with nonzero concentration values for analysis for the shorter travel times
(300 to 600 km transport) and from 8 to 16 receptors with nonzero concentration values
for analysis for the larger travel times (600 to 900 km transport). This suggests that the
51
-------
maximum concentration was likely not well characterized and is uncertain both in the
observations and in the simulations.
For each 6-hour period, the centroid position was computed for the observed and
simulated tracer puff as it was transported downwind over time. The results for each
release are shown in Figure 7. It was concluded that the simulated trajectories are
sensitive to the manner in which the wind field is characterized, but are insensitive to
the model employed. This can be seen by the fact that the simulated trajectories are
nearly identical for MESOPUFF and CALPUFF when the models have the same input
wind fields.
Ratios were formed by dividing the simulated and observed maximum
concentration values for each six-hour period. There was no noticeable trend seen as
a function of travel time (hours after release), but clearly there was a tendency to
overestimate the maximum concentration value by roughly a factor of 3.7. To further
investigate the tendency to overestimate the surface concentration maxima, ratios were
formed by dividing the simulated and observed lateral dispersion (as determined from
the second moment of the concentration values about the centroid positions) for each
six-hour period. There was no apparent trend in the ratio values with travel time
following release, however there was a clear trend to underestimate the
horizontal extent of the tracer puff for each six-hour period. There were only slight
differences seen for the different wind fields employed. The overall geometric average
was 0.54 and the geometric standard deviation was 1.81. If the only difficulty or "bias"
in the simulation was to underestimate the horizontal extent of the puffs, the effect on
the simulated maximum concentration values would be proportional to the inverse
square of the bias in the horizontal dispersion, which would be 3.43 = (1/0.54)2. It is
concluded that the underestimation seen in the simulated horizontal dispersion is able
to explain the overestimates seen in the simulated maximum concentration values. The
large scatter seen in concentration and lateral dispersion ratio values in part may relate
to the stochastic nature of atmospheric dispersion, but also is traceable to poor
definition of the maximum concentration values and lateral dispersion.
52
-------
90.00 —,
80.00 —
9 60.00 —
(B)
Release 5
October 26, 1983
0345-0645 GMT
30.00 40.00
50.00 60.00 70.00
X-GRID (18km)
80.00 90.00
E
9 60.00
50.00 —
40.00 —
30.00 •
(C)
Release 7
October 29, 1983
0600-0900 GMT
o o
30.00 40.00
50.00 60.00 70.00
X-GRID (18km)
80.00 90.00
90.00 —,
80.00 —
Release 3
October 2, 1983
1900-2200 GMT
Summary Over All Releases
n | i | i | i | i | r
30.00 40.00 50.00 60.00 70.00 80.00 90.00
X-GRID (18km)
0.00
1 ' I ' I ' I
400.00 600.00 800.00 1000.00
Travel Distance (km)
Figure 7. Summary of trajectory results for each of the releases are shown in parts (a), (b) and (c). In
part (d) is shown the fractional difference computed between the observed and simulated trajectories,
summarized over all releases for each of the three wind fields. The open smaller circles indicate
sampling locations. The X and Y coordinates have been specified in terms of the meteorological grid,
which had an 18-km size.
53
-------
The purpose of this investigation was to assess whether different methods of
characterizing the wind fields affect the performance of the simulated trajectory. The
comparison results presented were conducted with simplified puff dispersion model
assumptions, hence the model-to-model differences were minimized. The results
shown in Figure 7 suggest a noticeable improvement in the puff simulation results of
the centroid trajectory when mesoscale dynamic wind fields developed using FDDA
were used to characterize the time and space varying wind field, versus using
mesoscale wind fields determined diagnostically from the available hourly surface and
twice-daily upper air observations. The favorable comparison of the CALMET wind field
results with the MM4 results is influenced by Release 3 results, where CALMET
happened to well characterize the average transport direction. The trajectory
simulations for Releases 5 and 7 derived using MM4 winds are as good or better than
the results obtained with CALMET wind fields derived solely from routine National
Weather Service observations. It was concluded that use of MM4 is preferred and that
both CALMET and MM4 wind fields provide better simulations of the trajectories than
MESOPAC wind fields.
The analysis of the concentration maxima and lateral dispersion values suggest
that the simulation assumptions employed in these results consistently underestimate
the horizontal extent of the tracer puff as it is transported downwind. The centroid
maximum surface concentration was found to be correspondingly overestimated and
relatively insensitive to the mesoscale wind characterization. In these simulations, no
provisions were made to address delayed shear enhancement of the dispersion as
described by Moran and Pielke (1994) and Shi et al. (1990). Inclusion of some sort of
puff splitting is obviously warranted, but the computational demands are not trivial if one
is attempting to develop an operational model for routine use. Furthermore, in those
cases where the puff model dynamics have been enhanced, for example Draxler (1987)
and Davis et al. (1986), there was a tendency to underestimate the surface maximum
concentrations.
4.4 Constructing FDDA-MM Data Sets Assessment
To foster interest in the use of mesoscale meteorological (MM) data processed
using Four Dimensional Data Assimilation (FDDA) in routine air pollution modeling
assessments and to learn what problems might be associated with such a project,
IWAQM sponsored the development of a one-year meteorological data set for 1990
that spans the contiguous United States, southern Canada and northern Mexico.
Hourly profiles of wind, temperature and moisture were provided at 23 levels in the
atmosphere on an 80-km gird. The Penn State mesoscale meteorological model
(MM4) with FDDA was used in developing these data. The horizontal grid spacing of
the MM4 simulation was 80 km in both dimensions, with a grid array size of 85 by 56
centered on 90 W longitude and 40 N latitude to cover most of the North American
continent and adjacent oceanic areas (Bullock, 1993). The model simulation variables
from the outer 2 columns and rows of grid points were not included in the published
data set to avoid the boundary effects typical of fixed-grid numerical models. The
54
-------
domain of the model's vertical coordinate system extended from the earth's surface to
the 100 millibar pressure level (approximately 15 km above sea level). The 15-level
structure provides height-resolved information similar to that routinely obtained by the
National Weather Service at 12-hour intervals from approximately 80 rawinsonde
balloon sounding locations across North America. However, the data set obtained from
the MM4 simulation provides synthetic soundings at 1-hour intervals for 4080 model
grid-point locations, or about 600 times more information than is available from routine
observational networks. The 1990 data set contains over 20 billion bytes of
information.
The annual data set was developed from the MM4 output of 80 separate
simulations. Each simulation was 5 days in length with a 12-hour overlap with
chronologically adjacent simulations. This 12-hour overlap was used to allow a model
"spin-up" so each simulation's hour 12 conditions would match the previous simulation's
ending conditions without the detrimental effects of model initialization. The first 6
hours of each simulation were not used in the definition of the final data set. Hour 7
thru 11 results were blended with the previous simulation results with a time-linear
weighting function to produce temporally continuous fields in the final data set.
Beginning at hour 12, the simulation results were copied directly in the production of the
annual data set. Differences between adjacent simulations for these overlap periods
were monitored and no significant discontinuities were detected. At hour 12, the
differences were often zero at the floating-point numerical precision of the CRAY Y/MP
at EPA's National Environmental Supercomputing Center where the 80 MM4
simulations were performed. Each of these simulations required about 2.5 hours of
CPU time to compute on the Cray Y/MP. The MM4 output files were then transferred to
a DEC 3500 AXP workstation where the data were chronologically blended and chained
and various QA checks were performed. The final product was then re-partitioned into
36 sequential files of about 600 Mbytes each. Three files were produced for each
month, the first containing days 1 thru 10, the second containing days 11-20, and the
third containing day 21 through the end of each month. These files were then
processed by the National Climagic Data Center to produce a 12 Compact Diskette
data set (NCDC, 1995).
The science of mesoscale analysis using data assimilation is rapidly developing.
Major advancements have been occurring every several months during the period from
1995 through 1997. There are various research groups who have active development
programs investigating mesoscale meteorological modeling employing data
assimilation, e.g., Pielke et al., (1997), Turner and DeToro, (1998).
A major obstacle is access to these data. The 1990 MM4 data set in a
compressed format, and providing only profiles of wind, temperature and moisture,
requires 12 Compact Diskettes. An operational means for gaining easy access to
comprehensive mesoscale meteorological data sets, as alluded to here, has yet to be
developed.
55
-------
4.5 Regional Approach
Federal and State modelers have struggled with the issues related to Class I
area analyses and have found their resolutions to be elusive. Currently the consensus
opinion within IWAQM is that air quality impact assessements for Class I areas are
fundamentally different from local-scale assessements, typically associated with Class
II assessments and State Implementation Plans (SIP's). Unfortunately, the Federal
Agencies have attempted to implement the assessment of Class I impacts as if the
assessment is similar to that associated with Class II and SIP impacts. This has lead to
a significant mis-match between the analysis process and the inherent needs of a Class
I impact assessment.
Local-scale assessments require that analyses be performed within a domain on
the order of 50 km or less and centered on the source (the domain changes from permit
to permit). Given the small domain, Class II analyses lend themselves to individual
state implementation. Class I analyses, on the other hand, are centered on specific
land areas. Therefore, the modeling domain does not change from permit to permit.
Furthermore, these analyses involve a modeling domain on the order of hundreds of
km, thus requiring multi-state coordination. Additionally, Class I area analyses for
AQRV's may require estimating the deposition of secondary pollutants and their impact
on visibility. Finally, the affirmative responsibilities of the FLM's inherantly adds to the
coordination difficulties.
As an alternative to the current perm it-by-perm it practice, Class I air quality
modeling assessments could be designed for each Class I area (or cluster of Class I
areas). The cornerstone of this approach is an up-front comprehensive increment and
AQRV analysis of the area. We envision an "initialization" study being accomplished
outside the context of a permit application. An up-front study is preferred since many of
the decisions which need to be made (e.g., inventory, AQRV's criteria, etc.) are specific
to the Class I area, not the applicant's source. Further, it avoids having these decisions
colored by the negotiations which occur for a single source. If desired, the
"initialization" can involve technical experts from private and public groups. Finally (and
perhaps most importantly), it provides future applicants with up-front information
needed for planning and assurance of what is expected for the given situation.
In large part, the emissions inventory and meteorological data which are
developed during the initializing process remain fixed in subsequent analyses.
Therefore, once "initialization" is complete, each additional new source need only
determine its additive contribution; as such, increment and AQRV's are directly tracked.
This approach has the benefit of removing the burden, from each applicant, of
developing an emissions inventory specific for their application. Updates may be
necessary to account for changes in actual emissions from other sources. Within this
approach, provisions can be made for updating the initial analysis under certain
circumstances. For instance, a re-initialization analyses could result from any one of
56
-------
the following: 1) periodic State audit as contemplated by the PSD regulations, 2) by
choice of the permit applicant, or 3) as the scientific understanding changes. Once the
system is "re-initialized", new meteorological data and grids may result.
4.5.1 A Regional Approach to Implementing a Class I Area Assessments
The following outline presents a conceptual approach how the air quality
modeling could be accomplished for Class I areas. This approach was designed to
provide a framework in which many of the issues could be resolved.
I. INITIALIZING ANALYSIS: Use the recommendations of the Interagency
Workgroup on Air Quality Modeling (IWAQM) to initialize each Class I area
outside the context of a permit application.
A. Preprocess a fixed 5 year data base. All subsequent analyses will use this
data set. That is, the meteorological grid will be fixed for all future analyses.
B. Define both the computational and receptor grids. These grids are also fixed
for all future analyses (unless a reanalysis warrants changes).
C. Decide on the measures to be used in the Air Quality Related Values
(AQRV's) analyses, and determine the significance (de minimi's) criteria for
the AQRV's (this should be done specific to the Class I area)
D. Develop an Inventory:
1. Source inventory for PSD increment analyses.
2. Source inventory for AQRV analyses.
E. Run IWAQM recommended approach to produce appropriate
concentration/impact fields.
F. Archive these fields for use by future Prevention of Significant Deterioration
(PSD) applicants.
II. INITIALIZATION STUDY PARTICIPANTS: The initialization work should be a
cooperative effort among the FLM for the area and the EPA regions and states
who have or could have sources which affect the Class I area. The initialization
study participants could include technical experts from industry and academia.
III. PERMITTING:
A. The State within which the Class I area resides could be the ultimate
caretaker of all data bases1. The data bases would include the
meteorological data, the computation and receptor grid definitions, the Class
I area specific emissions inventories, the various topographical and other
data and the concentration/impact fields.
1 An issue yet to be resolved is developing an effective means to assist those
States faced with assessing impacts on Class I areas not within their respective
borders. Will coordination problems arise between States on roles and responsibilities?
57
-------
B. Using the data provided, an applicant need only determine its own
incremental impacts. These impacts will then be added to the archived
concentration/impact fields for comparison against appropriate increments
and AQRV's.
IV. DATABASE UPDATES: It is envisioned that re-initialization analyses could
result from the following:
A. Periodic state audit as contemplated by the PSD regulations.
B. By choice of the permit applicant. An applicant will be permitted to use a
more recent 5 year set of meteorological data with the requirement that they
perform a complete re-initialization analysis. Such an analysis would update
the comprehensive data base for future analyses.
C. As scientific understanding changes.
4.6 Comparisons of CALPUFF with Tracer Field Data
There are very few intensive tracer field experiments available for investigating
model simulations of mesoscale transport and dispersion. The IWAQM is aware that
there are several other expriments for which the CALMET/CALPUFF modeling system
has not been exercised (mostly in Europe). The results that are summarized here
represent those that IWAQM was aware of and for which the data could be obtained.
4.6.1 1975 Savannah River Laboratory Tracer Study
In this study (U.S. EPA, 1998a), concentration estimates from the CALPUFF
dispersion model were compared to observed tracer concentrations from a short-term
field experiment conducted at the Savannah River Laboratory (SRL) in South Carolina
on December 10, 1975 (U.S. DOE, 1978). This experiment was designed to examine
long-range transport of inert tracer materials to demonstrate the feasibility of using
other tracers as alternatives to the more commonly used SF6. Several tracers were
released for a short duration (3-4 hours) and the resulting plume concentrations were
recorded at an array of monitors downwind from the source.
For the CALMET/CALPUFF simulations, a meteorological grid extending from
32° N to 34° N latitude and from 80° W to 82° W longitude was used. Figure 8 shows
the region of the SRL field experiment. The SRL facility is near the west edge of the
domain and the sampling monitors are located along Interstate 95. A 24-by-24
horizontal grid with a 10-kilometer resolution was used for the SRL modeling. To
adequately characterize the vertical structure of the atmosphere, six layers were
defined: surface-20, 20-50, 50-100, 100-500, 500-2000, and 2000-3300 meters.
58
-------
} ~
\
H I
— r
BAMiERG
ALLENDALE
0 S 10 II 20
I I I I I
HHLE3
<> 'P ? ? *," TILLMAN
KIL0METEJH
-95
ST. GEORGE
WALTERBORO
'MP35
MP46
*-MP42
• YEMASS1E
SC336J
RID6ELAND
Figure 8. Savannah River Laboratory field experiment site.
The CALMET preprocessor utilizes National Weather Service (NWS)
meteorological data and on-site data to produce temporally and spatially varying three
dimensional wind fields for CALPUFF. Only NWS data were used for this effort and
were extracted from two compact disc (CD) data sets (see Appendix C). The first was
the Solar and Meteorological Surface Observation Network (SAMSON) compact discs,
which were used to obtain the hourly surface observations. The following surface
stations were used:
Georgia:
North Carolina:
South Carolina:
Athens, Atlanta, Augusta, Macon, Savannah
Asheville, Charlotte, Greensboro, Raleigh-Durham,
Wilmington
Charleston, Columbia, Greer-Spartanburg
Twice daily soundings came from the second set of compact discs, the Radiosonde
Data for North America. The following stations were used:
Georgia:
South Carolina:
North Carolina:
Athens, Waycross
Charleston
Greensboro, Cape Hatteras
59
-------
The geophysical parameters were derived from geographical information system
(CIS) land-use categories. Terrain and land-use data were available on the CALMET,
CALPUFF, and CALPOST Modeling System (version 1.0) CD (hereafter referred to as
the CALPUFF CD, see Appendix C). The terrain and CIS land-use data on the
CALPUFF CD were used to define gridded land-use data for each field experiment.
These data are defined with a resolution of 1/6° latitude and 1/4° longitude.
SF6 and two heavy methanes were released. For this analysis, the SF6 tracer
emission rates were used. The source parameters for this analysis were, a release
height of 62 m, with 154 kg of SF6 tracer released over a 4 hour period (10.69 g/s) with
no buoyant plume rise.
The distance to the monitoring arc was approximately 100 kilometers. The
monitors were located along I-95 (Figure 8) from MP76 on I-95 near St. George south
to Hwy 336 west of Tillman, SC and along SC 336. The monitors subtended an arc of
about 70°. Receptors for modeling were placed along an arc every 1/4° degree from
MP76 to MP22 near Ridgeland, resulting in 261 receptor locations. The distance
between receptors was about 450 meters.
Two separate CALPUFF model runs were made: 1) using Pasquill-Gifford (PG)
dispersion parameters, and 2) using dispersion coefficients from internally-calculated ov
and ow from the micrometeorological variables calculated in CALMET (hereafter
referred to as similarity dispersion). The central maximum concentration is estimated
from a Gaussian fit to the modeled and observed data (Cmax) and computed from the
crosswind integrated concentration (CWIC) and the lateral dispersion, oy, as Cmax =
CWIC^/ZTT; oy). The CWIC was computed by trapezoidal integration. The program that
computed these measures utilized only those values that were 1 % or greater of the
maximum.
The observed concentrations are the cumulative concentration from bag samples
located along Interstate 95 from about St. George south to Ridgeland (Figure 9).
Background concentration was estimated to be 0.5 ppt (DOE, 1978). The tracer
release started at 10:25 Local Standard Time (LST) and continued until 14:25 LST.
The bag samplers were started at different times, ranging from about 10:40 to 12:30
LST, and the duration of the sampling ranged from 7.0 to 7.5 hours. Since the release
started at 10:25 LST, it seems likely that sampling at the monitors would have begun
prior to the arrival of the plume. The arrival time of the modeled plume was the hour
ending at 13:00 LST for both PG and similarity dispersion. The simulated plume
required seven hours to pass the arc with the PG dispersion coefficients, but only six
hours with similarity dispersion coefficients. Therefore, seven-hour-average modeled
and observed concentrations were computed for comparison with the measurements.
Since the first monitors were turned on prior to 11:00 LST and only cumulative
concentration is reported for the observed data, the simulated concentrations were
summed over the seven-hour period from 11:00 LST through 1800 LST.
60
-------
(x\ Savannah River
w Dec. 10, 1975
70_1.QD.Jflm.aj:c
P-G Dispersion
Similarity Dispersion
—•— Observations
0.0
100
110
120 130 140
Azimuth (decrees)
150
Figure 9. Simulated and observed seven-hour average concentration values along
sampling arc for the Savannah River Laboratory December 10, 1975 tracer field
experiment.
Figure 9 shows the plots of the concentration estimates at the receptors
(continuous curves) and the observed concentrations at the receptors (labeled points).
The modeled peaks are 10° to 20° further to the south than the observed peak. It
appears that the CALMET meteorology derived using routine NWS was not able to
characterize this initial difference in wind direction sufficiently to transport the plume
more toward the north.
Clearly, there is general agreement in the shape and magnitude of the
distributions. Note that there are two local maxima in the observations near 135° and
145°. The winds were more northerly shortly after the release and may have resulted
in the observed local peaks (DOE, 1978) that were not captured in the modeled
meteorology. The observed lateral dispersion is 50-100% larger than the modeled
dispersion due to these local peaks. If these two secondary peaks are omitted from the
analysis, then the statistical measures of the simulated plumes are in better agreement
with the measures of the observed plume. Without these secondary peaks, the fitted
central maximum to the observations increases by 37% to 3.8 ppt (modeled: PG 7.2 ppt
and Similarity 5.1 ppt), The observed computed lateral dispersion is reduced by 33% to
7.77 kilometers (modeled: PG 6.9 km and Similarity 5.0 km). The observed CWIC is
reduced only slightly to 0.732 ppt-m (modeled: PG 1.29 ppt-m and Similarity 0.8 ppt-m).
61
-------
With only one realization for comparison, general conclusions regarding model
performance are not possible. But the simulation results are in reasonable accord and
do not suggest there are severe problems in the modeling system. It is encouraging
that the correspondence is as close as it is given that only routine NWS observations
were employed in developing the meteorological fields.
4.6.2 1977 Idaho FallsTracer Study
CALPUFF dispersion modeling results (Irwin, 1997) were compared with data
obtained following a single 3-hour late afternoon tracer release, lasting from 1240 to
1540 Mountain Standard Time (MST), conducted on April 19, 1977 near Idaho Falls,
Idaho. The tracer release results (Clements, 1979) were obtained as a consequence of
an investigation into the feasibility of using certain perfluorocarbons and heavy
methanes as alternative tracers in place of sulfur hexafluoride (SF6). Hence, although
the results have found use for testing alternative characterizations of dispersion and
transport, this was not a primary purpose in the original design of the investigation.
Draxler (1979) included this experiment in an assessment of the effects of alternative
methods of processing wind data for characterization of the mesoscale trajectory and
dispersion. He concluded that a network of wind observations having a spacing on the
order of 25 kilometers might be needed to simulate mesoscale transport associated
with variable-flow situations, and that spacing of order 100 kilometers might prove
adequate for stationary and homogeneous flow situations.
The design for meteorological data collection and sampling locations relative to
the release location is shown in Figure 10. Since locations of towers and sites were
extracted from data volume figures, the relative positions are likely accurate but the
absolute positions are no better than 0.5 km. The receptor arcs at 48 and 90 km
downwind from the release are shown in Figure 10. Meteorological data were available
from eleven sites providing hourly-averaged winds; four sites providing hourly-averaged
winds and temperatures, three sites providing hourly pibal observations of winds aloft
(CFA, MTV, DBS). Two of the pibal sites (CFA and DBS) also provided hourly-
averaged winds and temperatures. Hourly rawindsonde observations were taken at
about 600 m northwest of the release location. The meteorological masts ranged in
height above ground with two at 6.1 m, eleven at 15.2 m, three at 22.8 m, and two at 30
m. The pibal observations taken at Billings, Montana (well past the farthest sampling
arc downwind) were not used in this investigation. The skies were clear of clouds and
no precipitation occurred during the experiment. The National Weather Service
observations taken at Pocatello, Idaho (approximately 75 km southeast of the release
location) were included to provide station pressure (required input for CALMET).
62
-------
100.00
80.00 —
60.00 -
40.00 —
op
£ 20.00 —
-20.00 —
-40.00
| Winds and Temperature
t Winds
A PIBAL
O Receptors
0
°°
°S
1 1
-40.00 -20.00 0.00 20.00 40.00 60.00 80.00 100.00
East-West (km)
Figure 10. The Idaho tracer experiment sampling arcs and meteorological data
collection network. The sampling arcs at 48 km and 90 km are shown. The receptor
arc at 3.2 km downwind of the release is omitted for clarity.
To estimate the effects of drainage flow on the near-surface wind field, gridded
values of land-use and terrain heights are needed. The land-use data are used as
surrogates for typical values of surface roughness, albedo, soil heat flux, anthropogenic
heat flux and leaf area index. These surface parameters are used in estimating the
surface energy balance. For this analysis, U.S. Geological Service land-use and terrain
height data were extracted from data bases included in U.S. EPA (1996). The basic
grid size for these data is approximately 900 m. They were processed into a 20 by 20
grid with a grid resolution of 10 km. Default values, as defined in U.S. EPA (1996), for
the surface parameters to be associated with the land-use data were used. The
southwest corner of this grid was approximately 50 km southwest of the release. The
area depicted in Figure 10 is fairly flat, but the terrain sharply increases in height to the
west and north of the area depicted. The dominant land-use was rangeland; and the
surface roughness was estimated based on land-use to be on the order of 10
centimeters.
Hourly-averaged winds and temperature were available from midnight April 18
through midnight April 19. To mitigate the effects of not having surface data beyond
midnight of April 19, the surface meteorological tower data were duplicated to form two
63
-------
24-hour periods, having identical meteorology. The assumption being made is that
conditions were steady-state. The pibal and rawindsonde data, which were available
from 0700 MST to 1900 MST, were treated in a similar manner. CALMET assumes all
upper-air observations are from rawindsondes, and thus expects upper-air observations
to provide winds, dry-bulb temperature and pressure with height. CALMET interpolates
in height for missing data values at intermediate heights in an observation, but CALMET
will not extrapolate upper air data. Thus observations are rejected that fail to reach the
user-prescribed top of the modeling domain (3300 m for this analysis), or have missing
data values at the surface. To make use of the hourly pibal observed winds,
temperature and pressure values were added by linearly interpolating in time and height
between available rawindsonde observations, which were available every 1 to 3 hours.
The pibal wind directions were consistent with those from the one rawindsonde, but the
wind speeds were generally less in magnitude.
A purpose of this investigation was to assess the effects of having different
amounts of meteorological data for use in the development of the time varying field of
meteorological data. For this purpose four separate runs were made: Case 1 using all
available upper-air and surface mast observations, Case 2 using all surface mast
observations but only the one on-site rawindsonde upper-air observation, Case 3 using
only the CFA wind and temperature observations with the one on-site rawindsonde
upper-air observations, and Case 4 using only the CFA wind and temperature
observations with all upper-air observations. In Cases 1 and 2, all the on-site hourly
wind and temperature data are employed but different amounts of upper-air
observations are used. In Cases 3 and 4, hourly winds and temperatures taken close
to the release are used with different amounts of upper-air observations. For all the
CALMET simulations, winds and temperatures were computed for six layers in the
vertical, the midpoints of which were: 10 m, 35 m, 75 m, 300 m, 1250 m, and 2650 m.
The winds at CFA were higher than those generally seen throughout the
network. Hence in Cases 1 and 2 when all the on-site winds were employed the low-
level winds were lower than when only CFA data were used. In Cases 1 and 2, the
afternoon stability was Pasquill category B/C (Monin Obukhov lengths of order -30 m).
As a consequence of higher winds in Cases 3 and 4, the surface friction velocities were
higher, and the Monin Obukhov lengths were larger (in magnitude), thus closer to
neutral stability. The afternoon mixing heights are similar regardless of data used. This
results because the "upper-air" temperatures all have a common source, namely the
rawindsonde observations taken 600 m northwest of the release. The nighttime mixing
heights are mostly a function of the magnitude of the friction velocity. Hence, where
estimated friction velocities were largest and differ most among the various processing
methods, differences were seen in the nighttime mixing height values.
Each of the four analyses of meteorology was used to produce two CALPUFF
simulations of ground-level concentrations for each of the three sampling arcs. In the
first simulation, the dispersion was described using Pasquill-Gifford dispersion
parameters. In the second simulation, the dispersion was described using dispersion
64
-------
parameters suggested by Draxler (1976), which require values of the standard deviation
of the vertical and lateral wind fluctuations (referred to hereafter as "similarity
dispersion"). The wind fluctuation standard deviations estimated within CALMET are
primarily dependent on the surface friction velocity. The surface friction velocity is a
strong function of stability (largest during unstable conditions), roughness length and
wind speed (increases as roughness length or wind speed increase).
For each sampling step, CALPUFF internally computes a transport wind
averaged over the depth of the puff from the multi-layer winds provided to it from
CALMET. As a surface release puff grows in the vertical, the depth through which the
wind is averaged increases. The SF6 tracer emission was reported to be steady at
25.37 g/s over the three hour period, and was simulated within CALPUFF as a 3-hour
point-source release at 10-m starting at 1300 MST. The release height was set at the
midpoint of the lowest CALMET layer, to insure that the internally computed standard
deviations of lateral and vertical velocity fluctuations (for use in the similarity dispersion
parameter characterizations) at the specified release height, were in accord with the
wind speed used by CALPUFF for the lowest layer.
For each 6-hour period, the second moment (lateral dispersion, oy) of SF6
concentration values about its centroid position along the arc was computed. The
crosswind integrated concentration, CWIC, was computed by trapezoidal integration.
By assuming the concentration profile along the arc is Gaussian, the central maximum,
Cmax, was computed as, Cmax = CWIC/(y2u oy ).
A goal of this investigation was to assess the sensitivity of the modeling results to
different treatments of processing the meteorology, as well as to assess the
performance of CALPUFF in characterizing dispersion for transport distances beyond
50 km. Figure 1 1 depicts the observed SF6 concentrations with the simulation results
where all the surface and upper-air observations were used to generate the hourly wind
fields. For the observed values, there were from 14 to 17 receptors along each arc with
valid data for analysis. For analysis of the simulation results, receptors were spaced at
each arc distance at 2 degree intervals, over the 90 degree sector northeast of the
65
-------
(B)
0.00 20.00 40.00 60.00 80.00
Arc Azimuth (degrees)
(C)
(D)
20.00 40.00 60.00
Arc Azimuth (degrees)
I
80.00
HourofDay (LSI)
Figure 11. Six-hour average SF6 concentration values observed and
estimated for April 19, 1977, (A) 3.2-km arc, 1300-1900 MST; (B) 48-km arc,
1400-2000 MST; (C) 90-km arc, 1600-2200 MST. Azimuth is defined as
viewed from the release position with 0 due North and 90 due East (see
Figure 9). Receptor numbers are shown just above each observed
concentration value. (D) Time history of observed PDCH and estimated SF6
concentrations along the 48-km arc for April 19, 1977. Observed PDCH
values were multiplied by 3.16 for comparison with estimated SF6 values
(volume of SF6 divided by volume of PDCH released equals 3.16).
66
-------
release location. The second moment, oy, represents a measure of the puff horizontal
dispersion. For these 6-hour periods, the observed lateral dispersion ranged from
roughly 22% to 15% of the travel distance downwind. The crosswind integrated
concentration (CWIC) value characterizes the amount of pollutant mass seen at the
surface. The observed CWIC values at all of the arcs is close to what one would
expect if the tracer had become well mixed in the vertical. As shown in Figure 11
(which is typical for all of the simulations), the simulated transport was somewhat south
of the observed position along the first two arcs. It is also apparent that the
concentrations simulated for the first arc are at least a factor of 5 higher than observed.
Figure 11d provides a comparison of the time history of the puff, as it passed by
the 48-km arc. Sampling results are shown for the two-trap sampler which provided 5-
minute samples, and a cassette sampler which provided approximately 15-minute
samples. These samplers were quite close to the observed position of the 6-hour SF6.
maximum along this arc. The dispersion results are for the simulated position of the
maximum, which was somewhat displaced from that observed. The Pasquill dispersion
results are in remarkable accord with the tracer results. The similarity results arrive and
depart slightly later than observed. The slower transport for the similarity dispersion
occurs because the vertical dispersion was less than that simulated by Pasquill
dispersion, hence the transport speed was computed over a more shallow layer for the
similarity results. These results and those discussed above suggest that the similarity
dispersion was underestimating the vertical dispersion for this case.
The primary purpose of this investigation was to assess whether the CALPUFF
simulations were in reasonable accord with the observed concentrations. The
comparison results presented reveal as yet unexplained differences for the nearest arc,
3.2 km downwind from the release. The simulated pattern of dispersion was displaced
as much as 40 degrees from that observed, regardless of how the wind fields were
characterized. For all arcs, the lateral dispersion along the sampling arcs was best
characterized by both dispersion characterizations when all the surface tower winds
were used. Except for the first sampling arc, the simulated maxima along the arcs were
typically within a factor of 2 of that observed. The Pasquill simulations were most
sensitive to how the wind fields were characterized, showing the most variability
between the various wind field results. Having but one puff release limits conclusions to
be reached. For this one realization, it would appear that simulations by both
dispersion characterizations were in best accord overall with observations when all the
low-level winds and upper-air observations were used. And for this case, the similarity
dispersion simulations may have underestimated the vertical dispersion.
4.6.3 1980 Great Plains Tracer Study
In this study (U.S. EPA, 1998a), concentration estimates from the CALPUFF
dispersion model were compared to observed tracer concentrations from a short-term
field experiment (the Great Plains experiment) near Norman, Oklahoma (Ferber et al.,
1981) in July 1980. This experiment examined long-range transport of inert tracer
67
-------
materials to demonstrate the feasibility of using other tracers as alternatives to the more
commonly used SF6. Several tracers were released for a short duration (3-4 hours) and
the resulting plume concentrations were recorded at an array of monitors downwind
from the source. For the Great Plains experiment, arcs of monitors were located 100
and 600 kilometers from the source.
98°
43°W
O BATS Samplers
© BATS & LASL Sampler!
'l A RiwiroondB Station!
Figure 12. Great Plains field experiment site.
Previous studies have compared the results from the Great Plains experiment to
dispersion model results. Carhart et al. (1989) intercompared the results from eight
short-term, long-range dispersion models to the Great Plains results and to a longer-
term study at the Savannah River Laboratory (not the study discussed in Section 4.6.1).
The primary method for evaluating model performance was the use of the American
Meteorological Society (AMS) statistical measures (Fox, 1981) and graphical
techniques. They concluded that model results compared in space and in time to
observations were generally poor and that predictions for a specific location and time
for averaging periods less than one day were not reliable. They also noted that
unpairing decreases the scatter. They concluded that "model improvement can be
made by better representing the wind field. The use of multiple layers seems to
improve results substantially."
The transport and diffusion of a tracer gas was simulated by Moran and Pielke
(1995a,b) using the Colorado State University mesoscale atmospheric dispersion
modeling (CSU MAD) system, which consists of a prognostic meteorological model
coupled to a mesoscale Lagrangian particle dispersion model. Results from several
simulations with the model were compared to observations from the Great Plains
experiment. Their baseline simulation generally compared favorably to observations for
68
-------
both arcs although directional errors of up to 20° were apparent. The results also
suggest that the nocturnal low-level jet plays an important role in transport and
deformation of the tracer plume and that some flow regimes require better temporal
resolution of boundary layer winds than is available from the National Weather Service
(NWS) twice-daily rawinsondes.
The Great Plains site is shown in Figure 12. Two arcs of monitors were
deployed during the field experiment - 100 and 600 kilometers. For this analysis, two
separate grids were defined. For the 100-kilometer arc, a grid extending approximately
from 35° N to 36.5°N latitude and from 96° W to 98.5° W longitude was defined. A 42-
by-40 horizontal grid with a 10-kilometer resolution was used for this arc. For the 600-
kilometer arc, the grid extended from approximately 35° N to 42°N latitude and from
89° W to 100° W longitude. A 44-by-40 horizontal grid with a 20-kilometer resolution
was used for this arc.
To adequately characterize the vertical structure of the atmosphere, six layers
were defined: surface-20, 20-50, 50-100, 100-500, 500-2000, and 2000-3300 meters.
The CALMET preprocessor utilizes NWS meteorological data and on-site data to
produce temporally and spatially varying three dimensional wind fields for CALPUFF.
Only NWS data were used for this effort and came from two compact disc (CD) data
sets (see Appendix C). The first was the Solar and Meteorological Surface Observation
Network (SAMSON) compact discs, which were used to obtain the hourly surface
observations. The following surface stations were used:
Arkansas: Fort Smith
Iowa: Des Moines
Illinois: Springfield
Kansas: Dodge City, Topeka, Wichita
Missouri: Columbia, Kansas City, Springfield, St. Louis
Nebraska: Grand Island, Omaha, North Platte
Oklahoma: Oklahoma City, Tulsa
Texas: Amarillo, Dallas-Fort Worth, Lubbock, Wichita Falls
Twice daily soundings came from the second set of compact discs, the
Radiosonde Data for North America. The following stations were used:
Arkansas: Little Rock Nebraska: North Platte, Omaha
Illinois: Peoria Oklahoma: Oklahoma City
Kansas: Dodge City, Topeka Texas: Amarillo
Missouri: Monett
CALMET requires a file of terrain elevations and geophysical parameters in order
to prepare the wind fields and other meteorological parameters. The geophysical
parameters were derived from geographical information system (CIS) land-use
categories. Terrain and land-use data were available on the CALMET, CALPUFF, and
69
-------
CALPOST Modeling System (version 1.0) CD (see Appendix C). These data are
defined with a resolution of 1/6° latitude and 1/4° longitude.
The primary purpose of the Great Plains field study was to demonstrate the
efficacy of perfluorocarbons as tracers in atmospheric dispersion field studies.
Perfluoromonomethylcyclohexane (PMCH), perfluorodimethylcyclohexane (PDCH),
SF6, and two heavy methanes were released during this experiment. For this analysis,
the PDCH emission rates were used since the monitoring data appeared to have a
more complete record of PDCH concentrations. The following source parameters were
used for this analysis:
Source
Oklahoma
(July 8)
Oklahoma
(July 11)
Release
height
(m)
10.0
10.0
Stack
diameter
(m)
1.0*
1.0*
Exit
velocity
(m s-1)
0.001
0.001
Exit
temp.
(K)
ambient
ambient
Total tracer
released
(kg)
186
26
Length of
release
(hr)
3.0
3.0
Emission rate
(g s"1) and
tracer
17.22
PDCH
2.41
PDCH
The stack diameter for each study is the same as was used for the study with the INEL data
For both experiments, the emission rate was assumed to be constant over the
entire period of the release, and the release was assumed to be nonbuoyant.
For the July 8 Great Plains experiment, sampling was conducted using two arcs
of monitors: 100 kilometers and 600 kilometers as shown in Figure 12. Two separate
CALPUFF model runs were made: 1) using Pasquill-Gifford (PG) dispersion
parameters, and 2) using dispersion coefficients from internally-calculated ov and ow
from the micrometeorological variables calculated in CALMET (hereafter referred to as
similarity dispersion).
Beginning at 1300 LST on July 8, the PDCH and PMCH tracer gases were
released at a constant rate for a three-hour period from an open field at the National
Severe Storms Laboratory in Norman, Oklahoma. A background concentration for
PDCH of 26 ppt (Ferber et al., 1981) was removed from the observed concentrations.
The five-hour average modeled and observed concentrations for the 100-kilometer arc
on July 8 of the Great Plains field experiment are shown in Figure 13a along with Moran
and Pielke's baseline simulation (experiment 4b). Two things are immediately
apparent: the monitoring did not capture the entire plume and the observed maximum
concentration is very likely less than the simulated maxima. Given the incomplete
sampling of the observed plume at 100 kilometers for this release, the statistical
measures of the observed plume likely are suspect and are not sufficient to draw
conclusions regarding model performance. In comparing the CALPUFF results to
Moran and Pielke's simulation, the CSU MAD model placed the maximum about 25°
70
-------
west of the actual plume. Moran and Pielke's result for the 100 kilometer arc are very
similar to the CALPUFF simulations.
(A)
(a) Great Plains
JulyS, 1980
100 km arc
3.0
-20 -10 o 10 20 so
Azimuth (degrees)
(B)
(b) Great Plains
JulyS, 1980
600 km arc
0.40
10 20 30
Azimuth (degrees)
Figure 13. Simulated and observed concentrations for the Great Plains tracer field
study for July 8, 1980 for A) 5-hour average concentrations along the 100-km arc,
and B) 12-hour average concentrations along the 600-km arc.
The 12-hour average modeled and observed concentrations for the 600-
kilometer arc on July 8 are shown in Figure 13b. Two things are apparent: 1) the
observed maximum concentration is about three times higher than the simulated
concentrations and 2) the maxima of the simulations are in relatively good agreement
with each other. As noted above, the tracer arrived at the sampling arc earlier than
anticipated and the sampling likely missed some of the tracer material. Ferber et al.
(1981) speculate that the plume probably arrived just before the samplers were
activated and a small amount of plume material was not collected. As described by
Moran and Pielke (1995a), the most likely reason for the earlier-than-expected arrival
was the formation of a low-level nocturnal jet. Hoecker (1963), in detailed studies of the
low-level jet over the Midwestern plains (from Amarillo, TX to Little Rock, AR) using a
series of pibal stations, found that jet speed maxima occur between 300 and 800
meters above local ground. In examining available data for the 1980 Great Plains field
experiment, Moran and Pielke (1994) note an approximate doubling of the average
nocturnal wind speeds from their daytime values. Examination of the upper air wind
profiles for Oklahoma City through the period indicate the presence of a jet between
500 and 1000 meters for the 1200 Greenwich Mean Time (GMT) soundings.
71
-------
Great Plains
July 11, 1980
100 km arc
o
Q
0 20
Azimuth (degrees)
Figure 14. Simulated and observed 6-hour average concentration for the Great
Plains tracer field study on July 11,1980 along the 100-km arc.
Beginning at 1300 LSI on July 11, tracer gases were released for a three-hour
period using the same system as on July 8. The PDCH was released as an aerosol
spray at an average rate of 2.41 g/s, about 1/7 the release rate on July 8. A
background concentration for PDCH of 26 ppt was removed from the observed
concentrations. The transit time for the observed plume was six hours. The transit time
of the simulated plume in CALPUFF using both P-G and similarity dispersion
coefficients also was six hours. Therefore, six-hour average concentrations were used
in this part of the analysis.
The six-hour average modeled and observed concentrations for the 100-
kilometer arc are shown in Figure 14. As with the July 8 study, the monitoring did not
capture the entire plume at 100 kilometers, although the peak appears to be a little
better defined, with an observed maximum at receptor 18. There were no aircraft flights
to assist in determining the western extent of the plume. The simulated plumes using
PG and similarity dispersion agree with each other very well, but, as with the July 8
results for the 100-kilometer arc, the peaks are more than twice the magnitude of the
observed plume and the simulated lateral dispersion is less than the observed plume. .
As with the 100 kilometer arc for the July 8 study, the question remains - why do
the simulated plumes have higher central maxima and narrower dispersion. With a
more sophisticated modeling system, Moran and Pielke (1995a,b) encountered similar
differences in their examination of the July 8 simulation at 100 kilometers, and they
could not explain to their satisfaction why their dispersion model was not able to more
closely represent the observed dispersion patterns at the receptor arcs.
72
-------
4.6.4 1992 Project MOHAVE Tracer Study
In this investigation (Vimont, 1998), the CALMET/CALPUFF modeling system's
simulation of tracer gas dispersion was compared with field measurements collected
during the 1992 summer intensive monitoring period of Project MOHAVE. The Project
MOHAVE tracer monitoring sites used in this analysis were at Dolan Springs (DOSP),
Las Vegas Wash (LVWA), Meadview (MEAD), and Overton Beach (OVBE), (Figure
15). The tracer was released from the stack of the Mohave Power Plant (MOPP) in
direct proportion to the time-varying sulfur dioxide emissions. The tracer was collected
in 24-hour samples starting at 7 AM Mountain Standard Time (MST) at each of the
sites, except at MEAD which had two 12-hour samples with start times of 7 AM and 7
PM MST. The monitoring network is not well suited for rigorous model evaluation as it
is too sparse. Ideally, a network designed for model evaluation would have arcs of
monitors placed at distances of interest (as in the studies mentioned in previous
sections of this report). Also, it would be of interest to have sufficiently short sampling
intervals so that passage of tracer past and along an arc could be analyzed.
The meteorology is strongly influenced by the east-west river canyon that runs
from Las Vegas Wash and past Meadview. The Colorado River sharply turns south at
Las Vegas Wash along a north-south transect that connects Las Vegas Wash with the
location of the Mohave Power station. Upper air wind measurements were available
from three radar profiles located at the Mohave Power Plant (MOPP), Meadview
(MEAD), and Truxton (TRUX); RASS temperature profiles were available at MOPP,
(Figure 14). Twice-daily balloon soundings (winds and temperature) were available
from Dolan Springs (DOSP). Since CALMET requires both winds and temperatures at
each upper air station, it was necessary to generate temperature profiles for the radar
wind profiles collected at MEAD and TRUX. These were constructed by linear
interpolation in time using the temperature profiles at DOSP (DOSP is closer in
elevation to MEAD and TRUX, whereas MOPP is 600 to 800 m lower in elevation). The
radar winds were reported in both a high resolution mode, and a low resolution mode.
The high-resolution data does not extend as high, but provides more details. Thus the
high-resolution data were combined with the low resolution data before processing by
CALMET.
73
-------
., ^ .-
****'*\ • **.,' . «.
:\ ,'t:
\ ' ~'
\
\
.... \
\
Figure 15. The Project MOHAVE tracer field study collected tracer samples at Dolan
Springs (DOSP), Las Vegas Wash (LVWA), Meadview (MEAD), and Overton Beach
(OVBE). The tracer was released from the stack of the Mohave Power Plant
(MOPP). The bold solid lines are the CALMET "barriers" that were employed in
developing the CALMET analyses.
A number of diagnostic model runs were made, and some general conclusions
were reached. The initial CALPUFF runs with no complex terrain treatment grossly
underestimated concentrations, especially at the nearest monitoring site at DOSP. The
inclusion of the partial plume path adjustment terrain treatment option in CALPUFF
improved the correspondence of the calculated concentration with the measurements,
which is not surprising since several of the monitoring sites are more than 600 m higher
than the base of the Mohave Power Plant stack. It was also concluded that the
monitoring sites were too sparse to be used in a conventional manner. The narrowness
of the dispersing plume meant that even a very small shift of a few kilometers in the
plume position would result in major (in some cases orders of
74
-------
6,00
5.00
4,00
3,00
2,00
1.00
0,00
•IT"
V-V--
y>
5
S?
CO
(•-
•3?
5
r--
a>
C!
r--
H?
h~
Cj
r--
H?
5
r--
a?
C4
o5
fP
y?
c*
g)
cB
CP
CO
2*
o
r^
c3
8,00 -
5.00
4.00
3,00 --
2,00
1,00
0,00
i*
4 i
q^-f-t >••+••< t-4-l )-4H t I t-4H t--H H-< fr-M t H t I \ I
f-l f-l f-l M M «N >N >M >M >M
S-
CO
c:
5 C? ci C
C1
Figure 16 Tracer concentrations (femtoliters/liter) predicted and observed for the Las
Vegas Wash (LVW) monitoring site, (A) are results obtained without use of barriers,
and (B) are results obtained using barriers in the CALMET processing.
magnitude) changes in the simulated concentration values. To counter this deficiency,
eight receptors were placed 15 km around each monitor's location. The estimated
concentration in closest correspondence to that measured was selected for use in
assessing model performance.
When CALMET was initialized using the wind and temperature profiles from all
three sites, the CALPUFF tracer simulation results were in poor agreement with the
observations. It was determined that the 1/(distance)2 initialization used in CALMET
was not capturing the fact that each of these sites were located in very unique settings.
The TRUX profiler is located at a high elevation in a valley with a southwest to
75
-------
northeast orientation. The MOPP profiler is located in the Colorado River Valley, which
encompasses the DOSP location and continues up to the MEAD monitoring site. The
MEAD site is located near the west entrance of the Grand Canyon, which runs east-
west. CALMET was modified to allow 'barriers' to be placed in the Step 1 initialization
process, so that the influence of observations could be limited to better reflect the
unique settings of the profiler locations. The bold solid lines in Figure 15 show the
placement of the barriers. Figures 16a and 16b illustrate the differences seen in the
simulated concentrations without and with the barriers in the CALMET processing. The
correspondence of the simulated tracer concentrations with those monitored seems to
have been improved by allowing the barriers to influence the development of the
CALMET Step 1 wind fields.
4.7 Comparisons of CALPUFF With ISC3
A sensitivity study comparing the ground level concentration values of CALPUFF
(Scire et al, 1995b) with those of the Industrial Source Complex Short Term (ISC3)
model (U.S. EPA, 1995c) for steady state and nonsteady state conditions was
performed. The study (U.S. EPA, 1998; Eckhoff and Coulter, 1998) was divided into
two parts. First, specific CALPUFF model input settings were tested for the best setting
for emulating ISC3 under steady-state conditions. In the second part, the same input
settings were then used to compare CALPUFF to ISC3 results under nonsteady state
conditions.
For the first part of the study, CALPUFF (4.0) was compared with the latest
version of ISC3 (Version 96113). ISC3 was implemented in the 'Regulatory Default'
mode and the input file for CALPUFF was configured so as to emulate ISC3 as closely
as possible. Point sources were simulated for rural environments free of obstacles and
with stack heights of 2 m, 35 m, 100 m and 200 m above ground level. 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 the domain. A line of 62 receptors
out to 100 km was placed along the 360° radial, aligned with the transport wind flow,
and spaced at increasing intervals from the source.
For each pair of model runs (CALPUFF and ISC3), a residual concentration was
computed at each of the 62 receptors. From the 62 residuals (one for each receptor), a
mean, standard deviation (OR), and sum of residuals squared were computed. The
mean provides an indication (sign) of bias along the receptor radial. The variance of
the residuals provides general indication of the variance along the receptor radial.
Because many of the absolute residuals were quite small, the sum of the residuals
squared was also computed to provide a relatively robust indicator of accord along the
receptor radial. A Fractional Bias equation, FB = 2*(CALPUFF-ISC)/(CALPUFF+ISC)
was also computed. The steady state agreement between CALPUFF and ISC3 was
very good. Minimum and maximum fractional bias values of-0.18 and 0.06 are an
76
-------
indication of how well CALPUFF, with the slug model invoked, can emulate ISC3 in a
steady-state environment using point source input data.
Having confirmed that the dispersion algorithms within CALPUFF do emulate
quite well those within ISC3, the second part of the study was to compare modeling
results when the meteorological conditions were allowed to varying each hour. Hourly
meteorological data processed for input into ISC3 were selected from three stations to
simulate three climatically different regions of the United States: 1991 Boise, Idaho;
1990 Medford, Oregon; and 1964 Pittsburgh, Pennsylvania. Each set contains a year
of hourly averaged values of wind speed, wind direction, ambient temperature, stability
class, and mixing heights. In these simulations, wind fields were not generated for the
CALPUFF simulations. Instead, the ISC option was used, which allows CALPUFF to
operate using meteorological data processed for input to ISC3. This means that the
meteorology for these simulations is assumed to be everywhere the same, over the
entire CALPUFF computational domain. This admittedly artificial assumption for the
CALPUFF meteorology at least insured that the only reason for differences in the
modeling results, resulted from the fundamental differences in the treatment of
transport between a plume and a puff model.
The Boise data were selected because they were obtained in a river valley which
has highly directional wind flows. More than 33% of the recorded winds have a
northwesterly component and more than 33% of the winds have a southeasterly
component with the majority of those winds having speeds greater than two m/s. Some
of the modeled puffs were expected to be transported directly to the most distant
receptors. The Medford data were selected because they were obtained in an area
surrounded by mountains with a high number of calm wind hours. In 1990, 22% of the
Medford Oregon winds were calm winds (<1 m/s). This compares to the average of
6.5% occurrence of calm winds for the other two sites. During calm wind events,
CALPUFF calculates concentrations while ISC3 does not calculate concentrations.
ISC3 treats the hour as missing when determining concentration averages. There was
an expectation that there would be large differences in concentration averages because
of the high number of calm winds. The Pittsburgh data were selected because the
recording site is located on an open plain above a river valley and that the data have
been used as a standard test set for a number of years. The wind directions and wind
speeds are fairly well distributed, although there is a tendency for southwesterly winds.
The main receptor placement consisted of 15 rings of 36 receptors each, with
receptors spaced every 10 degrees starting at 360 degrees. The rings were spaced at
distances of 0.5, 1, 2, 3, 5, 10, 15, 20, 30, 50, 100, 150, 200, 250, and 300 km from the
77
-------
Table 8. Characteristics for point sources used in the
CALPUFF/ISC3 comparisons.
Stack
Height
(m)
2
35
100
200
Emission
Rate
(g/s)
100
100
100
100
Exit
Velocity
(m/s)
10.0
11.7
18.8
26.5
Stack Diameter
(m)
0.5
2.4
4.6
5.6
Exit
Temperatur
e(K)
300
432
416
425
source. Four point sources (Table 8) were used in these ISC3/CALPUFF comparisons
with hourly varying meteorology. For each year of hourly meteorology, and for each of
the 15 rings of receptors, the highest and second-highest concentration was
determined for four averaging times (1 -hour, 3-hour, 24-hour, and annual) for each
model. It was anticipated that some differences would be seen in the simulated
maxima, since ISC3 ignores hours with wind speed less than or equal to 1 m/s (calms),
for which CALPUFF continues to process, and because ISC3 can not treat the
consequences of a flow reversal between hours, a situation easily interpreted by a puff
model. Even though such differences were anticipated, the magnitude of these
differences, and sensitivity to release height was not known.
As a case study, a ten-hour period was examined to determine the cause of
large differences detected in the concentrations simulated by the two models using the
Boise meteorological data. The large differences were detected 5 to 15 km downwind
from the 2 -m point source for Hour 62 of the simulation, following a 4 hour period of
calm winds and then a wind reversal. Figure 17a illustrates the distribution of the
pollutant mass, in which emissions from particular hours, preceding Hour 62 have been
depicted so that it is possible to see how the sequence of events preceding Hour 62
has affected the results seen. Starting with Hour 52 there was a 5-hour period for
which the winds were from the east-southeast. Then there was a period of 4 hours of
calm winds, which was followed by a 180-degree wind shift for 2 hours. Thus the
pollutant mass which had started out moving towards the west-northwest, for the last
two hours is seen moving back towards the southwest. As illustrated in the figure, the
result is that a broad area upwind (in terms of Hour 62) is being affected by emissions
that were released during Hours 52-60. The subsequent superposition of all of the
puffs is shown in Figure 17b. For receptors 5 to 12 km southwest of the stack, the
superposition of all the puffs results in concentration values that are more than a factor
of two greater than that simulated by ISC3 for Hour 62. During one of the calm hours,
Hour 57, the emissions were simulated to penetrate a low-level inversion. CALPUFF
78
-------
(A)
lenterline at Hour 62
Calpuff Emissions
From Hours 61-62
Calpuff Emissions
From Hours 57-60
Calpuff Emissions
From Hours 52-56
ISCST Emissions
From Hour 62 Only
Distance (km)
(B)
\
-20 0 20
Centerline Distance Downwind From Source (km)
Figure 17. Case study of a ten-hour period of simulated dispersion by CALPUFF.
(A) depicts impacts for Hour 62, showing contributions from emissions released prior
to Hour 62, (B) depicts centerline concentrations from CALPUFF and ISC.
tracked the puffs and the subsequent fumigation to the ground once the inversion grew
to encompass the puffs which were aloft.
The ten-hour example just described is illustrative of how differences can arise.
It also serves notice that large differences can arise, and that these differences may
arise from a rather complicated history of events. Both the sequence of events, as well
as the dispersiveness of the atmosphere are important in understanding the puff
simulation results. Unlike in plume simulations, the concentration results obtained for a
given hour are unlikely to be understood simply by knowing the meteorology for the
given hour. The sequence of the meteorological conditions leading up to the hour in
question may be all important.
For each of the three sites, the results for each averaging time were summarized
by plotting on a common graph the percent difference, PD = 100(C-I)/I, where C equals
the highest (or second-highest) CALPUFF concentration along a receptor ring, and I
equals the highest (or second-highest) ISC3 concentration along the same receptor ring
for the given averaging time. Figure 18 is an example of some of the CALPUFF/ISC3
comparison results obtained by this sensitivity study. Illustrated in the figure are results
for the Medford meteorology, and the results shown are for the 1-, 3-, 24-hour and
annual average comparisons of the second-highest concentration values for each of the
15 receptor rings. As each receptor ring is at a prescribed distance, the results
illustrate as a function of downwind distance and source-type whether CALPUFF is
providing higher (PD>0) or lower (PD<0) concentrations than ISC3. The Medford
79
-------
meteorological data had the highest occurrence of calm wind conditions, and some of
the largest differences. CALPUFF was seen to provide concentration values greater
than ISC3, for this site for all averaging times. It is instructive to note that CALPUFF
can produce nonzero concentration values at the base of an elevated point source
release (when flow reverses and pollutants previously released are transported back
towards the source). ISC3 can never produce such an effect. It is also instructive to
note that CALPUFF can yield much greater concentrations than ISC3 at most any
distance downwind and for most any averaging time. The exact timing and distance
downwind of such occurrences is dependent on the sequence of meteorological events,
but typically can occur whenever there is a period of calm winds of extended length
(say more than three hours in duration).
Figure 19 provides a sense of the differences to be seen between sites. Here
annual averages are compared for each site. Typically, the greatest differences seen
between the two models are seen in the shorter averaging times. The comparisons of
the annual averages most clearly illustrate that as release height increases, CALPUFF
tends to increasingly provided higher concentration impacts in comparison to ISC3.
Also the comparisons of the annual averages best illustrate the occurrences of nonzero
concentration values being simulated by CALPUFF and the lack thereof by ISC3. The
effect on the model simulations and resulting comparison of concentration values of the
high incidence of calm wind conditions (~23%) at Medford is apparent. The Pittsburgh
and Boise meteorological data had similar incidences of calm wind conditions (7% and
6% respectively), but the Boise data was highly directional, with 33% of the wind
towards the southeast and 33% of the winds towards the northwest (suggestive of high
incidences of flow reversals). Thus it is that the Boise comparisons show CALPUFF
yielding higher percentage difference in the annual averages than those seen using the
Pittsburgh meteorological data.
In the above comparisons discussed, it is important to remember that no terrain
effects on the meteorology were modeled (other than that which is inherent in the
climatology of the surface weather observations). Hence, the percentage differences
seen could have occurred at any of the sites, given the correct sequence of
meteorological events. In this sensitivity analysis, one year of data was used from each
site. The site-to-site differences is seen to be large. In the comparisons to be
discussed in the next section, several years of data are used from one site.
80
-------
(A)
Distance Downwind (km)
(B)
Distance Downwind (km)
(C)
(D)
Distance Downwind (km)
Distance Downwind (km)
Figure 18. Percent differences (CALPUFF versus ISC3) as a function of downwind
distance for the second-highest concentrations; (A) 1-hr, (B) 3-hr, (C) 24-hr, and (D)
annual averages. Data are for Medford, Oregon.
81
-------
(A)
Distance Downwind (km)
(B)
Distance Downwind (km)
(C)
Distance Downwind (km)
Figure 19. Percent differences (CALPUFF versus ISC3) as a function of downwind
distance for the annual average: (A) Boise, (B) Medford, and (C) Pittsburgh.
82
-------
4.8 CALPUFF SCREEN
By far, one of the most demanding tasks in performing refined puff model
simulations is to successfully develop a valid time and space varying characterization of
the meteorological conditions for use by the CALPUFF puff dispersion model. The
processors that format and organize the input data to CALMET are not presently user-
friendly and demand strong computer skills. Often special custom routines are needed
to format available data into the acceptable formats for use by CALMET. Even if the
logistics of manipulating the data were simplified, developing realistic characterizations
of the time-varying three-dimensional wind fields will always require specialized skills.
Developing mesoscale and microscale meteorological characterizations requires having
not only the specialized understanding of micrometeorological wind effects, but also the
experience and expert judgement to know when a characterization developed by the
meteorological processor is unreasonable. Furthermore, to review and critique the
CALMET results requires strong computer skills for visualization of the CALMET
results, or for listing out for special inspection portions of the CALMET results.
In this section, we summarize the results from a study in which a methodology
was tested whereby CALPUFF could be used with a simplified set of meteorological
data, for the purpose of providing screening estimates of concentration and deposition
impacts (U.S. EPA, 1998b). The methodology was tested in two ways: 1) five years of
hourly meteorology were used to develop data for assessing the year-to-year variability,
and 2) one year of hourly meteorology was fully processed through CALMET to assess
whether the screening methodology devised did indeed provide concentration impacts
greater than would be developed using a fully developed set of meteorology.
4.8.1 Screening methodology
As a design goal, it was our intention to minimize the effort needed to create the
meteorological input for use by CALPUFF. CALPUFF has a built-in mode whereby it
can use the meteorological data file generated by PCRAMMET for the Industrial Source
Complex Short Term (ISCST3) model, thus bypassing the need to run CALMET.
The following approach was devised for running CALPUFF in a screening mode
to estimate ground-level concentrations over a large area:
1) generate five years of ISCST3 input meteorology using PCRAMMET,
2) generate an ISCST3 control file and use the ISC2PUF conversion program
to create the CALPUFF control file,
3) use the CALPUFF Graphical User Interface (GUI) to finalize the CALPUFF
control file before running the CALPUFF model,
83
-------
4) run CALPUFF with the ISCMET.DAT data option, and
5) pick the maximum concentration for each pollutant, for each distance and
averaging time modeled (see discussion below on receptor placement).
Generating ISC3 input meteorology
PC RAM MET can be used to generate the meteorological data files for
CALPUFF. Using hourly surface observations and twice-daily mixing heights,
PCRAMMET computes atmospheric stability in the form of Pasquill-Gifford (PG)
categories and rural and urban mixing heights. These data, along with the wind
direction, wind speed, and temperature, are read directly by CALPUFF without any
modification to the data file. It is recognized that this characterization of the
meteorology will not vary spatially as in a refined modeling effort, and the consequent
differences will be significant as the terrain becomes more rugged and complicated. As
is known by those experienced in puff dispersion modeling, one can make conservative
choices and assumptions, but it is difficult to guarantee that the results obtained will
always be more conservative (but not onerously so) than that derived using fully
developed time and space varying meteorological input data. In order to encourage the
resulting concentration estimates to be higher than would be obtained using a refined
set of meteorology, we have made conservative assumptions as to how the receptors
will be placed, and which concentration values will be selected for use.
To perform dry deposition calculations, the surface roughness, friction velocity,
and Monin-Obukhov length are required. These parameters can be computed by
PCRAMMET and written to an output file for use. In order to estimate these additional
parameters, several additional input values are required by PCRAMMET, including:
surface roughness at the site where the wind measurements are taken (usually an
airport), surface roughness at the site where the model is to be applied, noon-time
albedo, Bowen ratio, and fraction of the net radiation absorbed by the ground.
For the test results to be discussed, we applied the screening methodology to
the region about Oklahoma City depicted in Figure 12. We had previously developed
the CALMET terrain and land-use files as discussed in Section 4.6.3. A surface
roughness length at the measurement site (Oklahoma City airport) of 0.10 meters was
assumed. An average roughness length of 0.34 meters was computed from the
GEO.DAT file created for the entire modeling domain, as well as, an average noon-time
albedo of 0.15 and Bowen ratio of 1.00. A value of 0.15 was assumed for the fraction
of net radiation absorbed by the ground.
84
-------
ISC3 control file settings
Using this proposed methodology, the control file and meteorology input can be
created as if preparing for an application of the ISCST3 dispersion model. This
approach has the advantage that many dispersion modelers are familiar with both the
meteorological data and control file structures. The rural dispersion coefficients with
the regulatory default settings were selected, which include use of stack-tip downwash,
buoyancy induced dispersion, final plume rise, default wind speed profile exponents,
and default vertical potential temperature gradient. Averaging times for the model runs
were 1-hr, 3-hr, 24-hr, and annual averages.
A polar grid was used to define the receptor locations. CALPUFF does not
accept a polar grid directly. Receptors must be a subset of the CALPUFF
computational domain (which is a subset of the meteorological domain), i.e., a
Cartesian grid, or they must be discrete receptors. The ISC2PUF program can convert
a polar grid and associated terrain to discrete receptors with terrain. The ISC2PUF
limits the number of receptors that can be processed to 1200. Hence very large
applications, may require two or more runs of this program followed by merging the
results into a single CALPUFF control file.
For the test results to be discussed, the following polar receptor networks were
used:
Distance (km)
1,2, 3, 5, 10, 15,20, 30, 50
75, 100, 150,200,250, 300
Receptor Spacing
Alona Each Rina
every 10°
every 5°
Number of
Receptors
324
432
Distance Along Ring
Between Receptors
0.2 -8.7 km
6.5 -26.2 km
These two polar networks result in 756 receptors, well below the limit of about 1200.
CALPUFF control file settings
The CALPUFF system includes a program (ISC2PUF) that translates an
ISCST3 control file to a CALPUFF control file. The converted control file must be
edited prior to running CALPUFF. This can be accomplished in one of two ways: using
a text editor to modify the control file directly, or using the CALPUFF graphical user
interface (GUI) to guide the user through the options.
The CALPUFF control parameters that must be set by editing the ISC2PUF
generated CALPUFF control file include: the use of puffs or slugs, and the type of
dispersion - Pasquill-Gifford or internally-computed o's. Also to be considered are the
pollutants of interest, modeling of chemical transformations, and modeling dry
deposition processes.
85
-------
Pollutants: The transformation pathways for five active pollutants are treated by
the MESOPUFF II scheme in CALPUFF: S02, S04=, NOX, HN03, and N03-. Since haze
and visibility are of concern in areas such as national parks, CALPUFF is most likely to
be applied to model sulfates and nitrates. For the testing results to be presented, the
focus was on S02 and S04=.
Slug model versus puff model: The slug model was used, with the set of
default options for slugs.
Dispersion coefficients: Pasquill-Gifford dispersion coefficients for both the
screening and refined CALPUFF modeling were used.
Concentration estimates: CALPUFF was run initially for both the refined (one
year) and screening modes (five years) without any chemistry or deposition. This
provided 'baseline' concentration estimates without the effects of chemistry and
deposition. CALPUFF was then run for one year with chemistry and deposition
activated.
Chemical transformations: There are two transformation options in CALPUFF:
1) MESOPUFF II mechanisms and 2) a file with a diurnal cycle of transformation rates.
The MESOPUFF II option requires relative humidity as one of the input variables for
chemical transformations. However, this variable was not present in the current
ISCST3 meteorological data file. The second mode requires a file of diurnal
transformation rates specified by the user. In this mode, transformation rates are
spatially uniform but provides for some temporal variability. The second method, with
the file of transformation rates, was used in this modeling effort.
For the test results to be discussed, we used a 3.0 %/hr transformation rate for
daylight hours and 0.2 %/hr at night for the S02 to S04= transformation. These rates
were used for both CALPUFF refined and screening modes. The daytime period was
defined as 0700 to 2000 LST, which biases the daylight period towards a summer day
and should result in production of more sulfates.
Dry deposition: In CALPUFF, deposition can be modeled as either particle or
gas, depending on the pollutant. To estimate deposition, the surface roughness length,
surface friction velocity, and Monin-Obukhov length are estimated by CALMET and vary
temporally and spatially when CALPUFF is run in a refined mode. In the screening
mode, though, these variables are specified for each hour on the 'extended' data record
in the meteorological file. PCRAMMET was run to generate these extended data
records for the CALPUFF screening mode.
To compute dry deposition of particles or gases, CALPUFF requires one of the
following: 1) a file of the diurnal variation of deposition velocities for each pollutant
modeled, or 2) specification of the mass mean diameter, geometric standard deviation,
86
-------
and number of particle size intervals to evaluate the effective particle deposition
velocity. For the latter option, CALPUFF has default parameters for several pollutants,
including S02 and S04=. For this modeling effort, we used the second option with the
gaseous S02 default parameters and the particulate S04= default parameters.
4.8.2 Year to year variability
Five years of meteorological data were processed through PC RAM MET to create
the necessary input meteorology for the CALPUFF screening model runs. Oklahoma
City was used for the hourly surface observations and Oklahoma City/Norman for the
upper air data. Data from the Solar and Meteorological Surface Observation Network
(SAMSON) compact disc (see Appendix C) were used to obtain the hourly surface data.
The twice-daily mixing heights were retrieved from EPA's Support Center for Regulatory
Air Models (SCRAM) system. The mixing height data on SCRAM restricted the period
of choice from 1984 through 1991. Oklahoma City for 1985-1988 were used for this
modeling effort. There were no mixing height data for 1989 on SCRAM. The upper air
station changed from Oklahoma City to Norman in 1989. Since Norman is only 25-30
kilometers from Oklahoma City, data from Norman for 1990-1991 was used. There
were no periods of missing data for Oklahoma City that required filling. There were five
2-hour periods of unfilled mixing heights for the two years of data at Norman. Following
EPA guidance, the mixing heights were filled by linearly interpolating between the hours
before and after the missing periods to fill in the mixing heights.
The 2-m, 35-m and 200-m point sources listed in Table 8 were used in this
assessment. The rings of receptors (as listed earlier) were placed around each source,
and the five years of ISC-type meteorology were processed by CALPUFF. For this
assessment of year to year variations, inspection was made of the variations seen in
the highest 1-hr, 3-hr, 24-hr and annual simulated S02 concentration, along each
receptor ring. Table 9 summarizes the maximum and minimum value seen for C/Cavg,
where C is the maximum concentration seen in any one year along a receptor ring for a
particular source and averaging time, and Cavg is the average of the five values seen
for this source along this ring for the five years simulated. Rather than list the results
separately for each of the 15 receptor rings, we have compiled the results for three
groups, with group 1 having four receptor rings at distances 1-, 2-, 3-, and 5-km; group
2 having five receptor rings at distances 10-, 15-, 20-, 30-, and 50-km; and group 3
having six receptor rings at distances 75-, 100-, 150-, 200-, 250-, and 300-km. These
three groups subjectively relate to near-field impacts (group 1), mesoscale impacts
(group 2), and longer-range impacts (group 3).
87
-------
Table 9. Summary for three hypothetical sources and for four averaging times in the
variation seen in the ratio C/Cavg, where Cavg is the average of the 5 maximum S02
concentrations for a receptor ring and averaging time, and C is the maximum S02
concentration simulate during one of the five years. The variations listed in C/Cavg
reflect both variations seen with the distance range and variations over the five year
period of the analysis.
1-Hour
3-Hour
24-Hour
Annual
Range
1-5 km
10-50 km
75-300 km
1-5 km
10-50 km
75-300 km
1-5 km
10-50 km
75-300 km
1-5 km
10-50 km
75-300 km
2-m Source
Min
0.79
0.65
0.56
0.83
0.58
0.59
0.85
0.64
0.61
0.90
0.88
0.81
Max
1.52
1.86
1.66
1.26
1.43
1.87
1.16
1.55
2.13
1.11
1.13
1.29
35-m Source
Min
0.89
0.76
0.72
0.89
0.80
0.60
0.80
0.79
0.59
0.88
0.90
0.89
Max
1.14
1.47
1.23
1.21
1.48
1.52
1.26
1.45
1.52
1.13
1.12
1.25
200-m Source
Min
0.68
0.47
0.44
0.80
0.58
0.71
0.75
0.77
0.68
0.83
0.88
0.86
Max
2.05
2.32
1.74
1.49
2.19
1.51
1.29
1.60
1.41
1.26
1.17
1.17
As averaging time increased, there was less variation between years. For
instance, for the 75 to 300 km receptor rings, the ratio of the one-hour maxima to the
five-year average one-hour maxima varied from 0.45 to 2.30. Whereas the ratios for
the annual-averages ranged from 0.80 to 1.30. For these results it is concluded that we
might expect variation in the maximum concentrations of at least 20 to 30 percent.
4.8.3 SO2 concentrations
One of the comments received at the Sixth Modeling Conference was the
perception that the Phase 1 (Level I) screening procedure (U.S. EPA, 1993) was too
conservative to be of practical use. Thus a second part of developing a new screening
methodology was to test the results obtained with the new methodology with those that
would be obtained in a refined modeling analysis. Table 10 summarizes the maximum
88
-------
Table 1 0. Summary for three hypothetical sources and for four averaging times in
the variation seen in the ratio Cs/Cr, where Cs is the S02 maximum concentration for
a receptor ring and averaging time using CALPUFF with ISC meteorology, and Cr is
the second highest S02 concentration simulated for the same ring using CALPUFF
with CALMET meteorology. Results were generated for only 1990, so the minimum
and maximum values for Cs/Cr, reflect variations seen over the receptor rings
included in the range.
1-Hour
3-Hour
24-Hour
Annual
Range
1-5 km
10-50 km
75-300 km
1-5 km
10-50 km
75-300 km
1-5 km
10-50 km
75-300 km
1-5 km
10-50 km
75-300 km
2-m Source
Min
0.59
0.74
0.99
0.82
0.88
0.96
0.95
0.88
1.57
1.28
1.49
1.82
Max
1.31
1.30
3.20
1.42
1.43
3.89
1.00
1.21
4.48
1.37
2.28
3.47
35-m Source
Min
1.03
0.87
1.55
1.02
0.80
1.17
0.86
0.86
1.11
1.41
1.59
1.85
Max
1.51
1.86
2.09
1.08
1.53
3.35
0.98
1.28
2.55
1.47
2.29
2.89
200-m Source
Min
0.86
1.27
0.68
0.77
0.72
0.73
0.73
0.92
0.96
0.58
1.16
1.07
Max
3.07
3.29
1.09
1.56
2.71
1.04
0.99
1.54
1.50
1.13
1.30
1.27
and minimum value seen for Cs/Cr, where Cs is the maximum S02 concentration seen
for 1990 along a receptor ring for a particular source and averaging time using the
CALPUFF with ISC meteorology (a screening estimate), and Cr is the second-highest
S02 concentration seen along the same receptor ring resulting from using the
CALPUFF with fully developed CALMET meteorology (a refined estimate). Comparing
the CALPUFF screening estimate of the maximum concentration along a ring versus
the second-highest concentration from a refined model run, provides an assessment of
whether the new screening methodology is a conservative screening estimate.
In each of the three ranges, instances can be found where the maximum S02
concentration obtained using ISC meteorology as input to CALPUFF (screening
estimate), Cs, was not as high as the second-highest concentration obtained from using
CALMET meteorology as input to CALPUFF (refined analysis), Cr. It is important to
89
-------
remember that these comparisons are between the maximum screening estimate found
anywhere along the receptor ring and the refined modeling results for the maximum of
the second-highest concentrations found anywhere along the ring. If an actual source
were being modeled for a particular Class I area, we would anticipate that it is quite
likely that the refined model's estimate of the highest of the second-high concentrations
for the segment of a receptor ring passing through the Class I area would be lower than
what would be found anywhere around the receptor ring.
For the receptor rings at 75-km to 300-km, the Cs/Cr ratios for the 1-hour
comparisons range from 1.00 to 3.0 for the 2-m source, and from 0.70 to 1.10 for the
200-m source. As averaging time increases, the screening estimate of the maximum
concentration tends to be greater than the second-highest concentration obtained in the
refined analysis. The IWAQM concludes the new screening method (ISC meteorology
as input to the CALPUFF model) provided conservative estimates of maximum S02
concentration values, that were not overly conservative for practical use. Although
there is a finite possibility for the refined analyses to develop higher concentration
impacts, the likelihood of this result is low for specific source-receptor pairings.
4.8.4 SO4= concentrations
Table 11 summarizes the maximum and minimum value seen for Cs/Cr, where
Cs is the maximum sulfate concentration seen for 1990 along a receptor ring for a
particular source and averaging time using the CALPUFF with ISC meteorology, and Cr
is the maximum sulfate concentration seen along the same receptor ring resulting from
using the CALPUFF with fully developed CALMET meteorology. In each of the three
ranges in Table 11, instances can be found where the screening estimate for sulfate
concentration, Cs, is not as high as that obtained from the refined analysis, Cr. The 24-
hour ambient sulfate concentrations are typically used in long-range haze impact
assessments. For the 24-hour averaging time, the 2-m source maximum sulfate
concentrations obtained with CALPUFF using ISC meteorology are typically greater
than were obtained for this site in 1990 using CALPUFF with CALMET meteorological
data. For the 200-m release, the CALPUFF screening estimates are less than that
obtained using CALPUFF with CALMET meteorology.
90
-------
Table 1 1 . Summary for three hypothetical sources and for four averaging times in
the variation seen in the ratio Cs/Cr, where Cs is the maximum sulfate
concentration for a receptor ring and averaging time using CALPUFF with ISC
meteorology, and Cr is the maximum sulfate concentration simulated for the same
ring using CALPUFF with CALMET meteorology. Results were generated for only
1990, so the minimum and maximum values for Cs/Cr, reflect variations seen over
the receptor rings included in the range.
1-Hour
3-Hour
24-Hour
Annual
Range
1-5 km
10-50 km
75-300 km
1-5 km
10-50 km
75-300 km
1-5 km
10-50 km
75-300 km
1-5 km
10-50 km
75-300 km
2-m Source
Min
0.52
0.44
0.76
0.42
0.50
0.56
0.62
0.61
0.58
1.10
1.20
2.00
Max
0.89
0.73
2.57
0.76
0.72
3.25
1.04
0.91
3.07
1.17
2.02
3.20
35-m Source
Min
0.80
0.65
0.58
0.91
0.38
0.54
0.55
0.36
0.38
1.23
1.16
1.50
Max
2.12
1.74
2.11
1.02
1.15
2.16
0.83
0.50
1.51
1.43
1.52
1.88
200-m Source
Min
0.56
0.72
0.33
0.42
0.65
0.25
0.30
0.32
0.16
0.47
0.82
0.68
Max
1.72
1.16
0.51
1.27
1.55
0.50
0.70
1.39
0.33
0.89
0.93
0.80
4.8.5 SO2 deposition
Table 12 summarizes the maximum and minimum value seen for Cs/Cr, where
Cs is the maximum S02 deposition seen for 1990 along a receptor ring for a particular
source and averaging time using the CALPUFF with ISC meteorology, and Cr is the
maximum S02 deposition seen along the same arc resulting from using the CALPUFF
91
-------
Table 12. Summary for three hypothetical sources and for four averaging times in
the variation seen in the ratio Cs/Cr, where Cs is the maximum S02 deposition for a
receptor ring and averaging time using CALPUFF with ISC meteorology, and Cr is the
maximum S02 deposition simulated for the same ring using CALPUFF with CALMET
meteorology. Results were generated for only 1990, so the minimum and maximum
values for Cs/Cr, reflect variations seen over the receptor rings included in the range.
1-Hour
3-Hour
24-Hour
Annual
Range
1-5 km
10-50 km
75-300 km
1-5 km
10-50 km
75-300 km
1-5 km
10-50 km
75-300 km
1-5 km
10-50 km
75-300 km
2-m Source
Min
0.95
0.69
0.70
0.73
0.52
0.70
0.65
0.80
1.04
1.16
1.32
1.50
Max
1.29
1.47
1.18
0.81
1.06
1.05
0.70
1.14
1.25
1.00
1.76
2.54
35-m Source
Min
0.81
0.48
0.85
0.73
0.59
0.71
0.50
0.55
0.58
0.90
1.14
1.48
Max
0.96
1.16
1.27
0.78
0.91
1.47
0.55
0.63
1.11
0.99
1.71
2.13
200-m Source
Min
0.33
0.28
0.23
0.26
0.27
0.27
0.16
0.44
0.41
0.30
0.70
0.89
Max
0.99
1.03
0.62
0.60
0.87
0.47
0.55
0.62
0.77
0.64
0.91
1.02
with fully developed CALMET meteorology. In each of the three ranges in Table 12,
instances can be found where the screening estimate of S02 deposition, Cs, is not as
high as that obtained from the refined analysis, Cr. For S02 deposition, the annual or
seasonal average deposition is of most interest. For the lower release heights of 2-m
and 35-m, the maximum annual S02 deposition obtained with CALPUFF using ISC
meteorology are consistently greater than were obtained for this site in 1990 using
CALPUFF with CALMET meteorological data. For the 200-m release, the CALPUFF
screening estimates of S02 deposition are similar to those obtained using CALPUFF
with CALMET meteorology.
92
-------
Table 1 3. Summary for three hypothetical sources and for four averaging times in
the variation seen in the ratio Cs/Cr, where Cs is the maximum sulfate deposition for
a receptor ring and averaging time using CALPUFF with ISC meteorology, and Cr is
the maximum sulfate deposition simulated for the same ring using CALPUFF with
CALMET meteorology. Results were generated for only 1990, so the minimum and
maximum values for Cs/Cr, reflect variations seen over the receptor rings included in
the range.
1-Hour
3-Hour
24-Hour
Annual
Range
1-5 km
10-50 km
75-300 km
1-5 km
10-50 km
75-300 km
1-5 km
10-50 km
75-300 km
1-5 km
10-50 km
75-300 km
2-m Source
Min
3.60
1.39
1.32
2.05
1.63
1.21
0.93
2.32
1.50
1.41
1.61
1.98
Max
6.79
3.90
2.76
3.68
3.77
3.47
2.38
2.79
4.50
1.51
2.05
4.07
35-m Source
Min
2.66
1.21
0.94
1.54
0.96
0.83
0.98
0.78
0.90
1.54
1.54
1.38
Max
4.09
3.43
2.63
1.97
1.72
2.26
1.06
1.35
1.58
1.90
1.87
2.63
200-m Source
Min
1.20
1.01
0.45
0.59
0.96
0.34
0.30
0.72
0.31
0.80
1.11
0.81
Max
2.20
1.67
1.04
1.31
2.06
0.62
0.74
1.38
0.68
1.39
1.37
1.27
4.8.6 SO4= deposition
Table 13 summarizes the maximum and minimum value seen for Cs/Cr, where
Cs is the maximum sulfate deposition seen for 1990 along a receptor ring for a
particular source and averaging time using the CALPUFF with ISC meteorology, and Cr
is the maximum sulfate deposition seen along the same arc resulting from using the
CALPUFF with fully developed CALMET meteorology. In each of the three ranges in
Table 13, instances can be found where the screening estimate of sulfate deposition,
Cs, is not as high as that obtained from the refined analysis, Cr. For sulfate deposition,
the annual or seasonal average deposition is of most interest. For the lower release
heights of 2-m and 35-m, the maximum annual sulfate deposition obtained with
CALPUFF using ISC meteorology are consistently greater than were obtained for this
site in 1990 using CALPUFF with CALMET meteorological data. For the 200-m
93
-------
release, the CALPUFF screening estimates of sulfate deposition are generally higher
than those obtained using CALPUFF with CALMET meteorology.
4.8.7 Old screen versus new screen estimates
Ambient S02 concentrations
In Section 4.7, it was shown that when both CALPUFF and ISC use the ISC
meteorology with assumed flat terrain, there was for these comparisons a clear
tendency for CALPUFF to develop higher maximum concentration estimates the ISC for
all averaging times and for all distances downwind. Whether this will happen for a
particular source-receptor pairing, is dependent on whether calms and wind reversal
are existent in the proper sequence. The more frequently calm conditions occur for a
site, the more likely CALPUFF can develop higher concentration impacts. It is
concluded from comparisons results as summarized in Section 4.7 that it would be a
misconception to consider the Level I screening estimates for ambient S02
concentration values as being overly conservative (i.e., too high). CALPUFF is capable
of providing higher S02 concentration values than ISC simply due to the fact that calms
and wind reversals are not ignored.
Ambient sulfate concentrations
In the IWAQM Phase 1 and Phase 2 recommendations, ambient sulfate
concentrations are used in assessing long-range transport haze impacts. And in this
context, the 24-hour averages are used (as explained in Section 3) to address the
'regional' nature of the haze assessment. In the Phase 1 recommendations, the
screening estimate (Level 1) of sulfate concentration was obtained by multiplying the
S02 concentration by 1.5 (to account for the difference in molecular weight between
S04=and S02).
Shown in Figure 20 is a comparison of the 24-hr maximum sulfate
concentrations derived using the Phase 1 screening procedure and using CALMET
meteorology as input to CALPUFF. The Phase 1 screening procedure is seen to be
considerably higher (on the order of 30 times higher) than that derived using CALPUFF.
For the 200-m source, The Level 1 screening estimates are seen to be higher than that
derived using CALPUFF for receptor rings of 50-km or less. Beyond 50-km, the Level 1
24-hr sulfate concentration estimates are similar to that derived using CALPUFF.
94
-------
10000.00
24-hr Sulfate
Level 1, 2-m Source
Level 1, 35-m Source
Level 1, 200-m Source
CALPUFF, 2-m Source
CALPUFF, 35-m Source
CALPUFF, 200-m Source
0.10
10 100
Downwind Distance (km)
1000
Figure 20. Comparison of Phase 1 Level 1 screening estimates of maximum 24-hour
ambient sulfate concentrations versus maximum 24-hour ambient sulfate
concentration derived using CALMET meteorology as input to CALPUFF.
Total sulfur deposition
In the IWAQM Phase 1 and Phase 2 recommendations, sulfur deposition is used
in assessing air quality related values associated with forest health. In this context, the
seasonal or annual averages are used (as explained in Section 3). In the Phase 1
recommendations, the screening estimate (Level 1) of total sulfur deposition flux was
obtained by multiplying the S02 concentration by an assumed deposition velocity of
0.005 m/s, which was then multiplied by 0.5, since each gram of S02 deposited
contributes 0.5 grams of sulfur. In the Phase 1 screening estimates, the sulfur
deposition was assumed to be mostly from S02 gaseous deposition. In CALPUFF we
can simulate the dry deposition of both S02 and sulfate. The total sulfur flux is then
computed as 0.5 times the S02 deposition flux plus 0.33 times the S04= deposition flux
(to account for amount of sulfur provided by S04= and S02).
95
-------
100.00
1.00
s
o
0.10
Annual Average
Total Sulfur
2-m Source
35-m Source
200-m Source
10 100
Distance Downwind (km)
1000
Figure 21. Comparison of annual average total sulfur flux as derived using the Phase
1 Level screening procedure and using CALMET as input to CALPUFF. Shown in
the figure is the ratio formed by dividing the Level 1 result by the CALPUFF result.
Shown in Figure 21 is a comparison of the annual average total sulfur flux,
derived using the Phase 1 Level 1 screening procedure and using CALMET as input to
CALPUFF. In Figure 21, the comparison is shown by dividing the Phase 1 screening
estimate by the CALPUFF result. For the more distant receptor rings, the Phase 1
screening procedure is seen to provide higher (by a factor of 2 to more than a factor of
10 times higher) than that derived using CALPUFF.
4.8.8 Findings and conclusions
In this section, we have summarized the results from a study in which a
methodology was tested whereby CALPUFF was used with a simplified set of
meteorological data, for the purpose of providing screening estimates of concentration
and deposition impacts. It was seen that there were reasonably large variations in the
S02 concentration maxima from one year to the next. There are limitations to the
conclusions that can be reached, due to the limited nature of the testing that has been
thus far accomplished. Comparisons of results obtained using the new screening
methodology versus results obtained using fully developed CALMET meteorology has
only been conducted for one location and for one year. In all cases examined, cases
could be found where the CALPUFF screening results underestimated the maximum
impacts simulated using more fully developed (CALMET) meteorology as input to
CALPUFF. Thus IWAQM concludes that the screening method that has been tested
96
-------
does not guarantee that the pollutant impacts will always be greater than that obtained
using refined meteorology. Whether this precludes its use is a judgement decision.
There is a certain degree of conservatism inherent in the screening procedure tested.
This arises because the screening procedure requires use of receptor rings that
completely surround the source being assessed, and it requires use of the maximum
impact found anywhere along the receptor ring. In an actual situation, it is unlikely that
the Class I area will completely surround the source being analyzed. It is more likely
that the actual Class I area is limited to a small segment of a receptor ring. Thus if
actual refined (fully developed) meteorology were developed and used, with actual
source locations and receptors limited to the Class I area, one is likely to find the
impacts simulated within the Class I may be considerably lower than that derived from
the screening procedure for receptors that encircle the source.
4.9 CALMET/CALPUFF Enhancements
4.9.1 Use of FDDA-MMdata with CALMET
Results of an ongoing investigation were reported by Sherwell and Garrison
(1997). The study was being conducted by the Maryland Department of Natural
Resources Power Plant Research Program (PPRP), in cooperation with EPA, to
investigate the results that might be obtained using the CALMET/CALPUFF modeling
system to simulate the magnitudes, sources, and possible reductions of NOx deposition
to the Chesapeake Bay. In this study, the Penn State MM4 gridded 1990 meteorology
data (discussed in Section 4.4) were employed. The EPA 1990 National Emissions
Inventory for NOx has been used to derive source inputs. In the presentation by
Sherwell and Garrison (1997), an overview was presented of the experiences gained in
preparing and running the CALMET/CALPUFF system, and on the preliminary results of
the analysis of NOx deposition to the Chesapeake Bay. A large part of their effort was
developing the NOx inventory, due to its sheer size, as the raw inventory contained
almost 90,000 entries for point sources and over 1100 counties in the defined domain.
The selection of their modeling domain was guided in part by a desire to
compare results obtained using the CALMET/CALPUFF modeling system with those
available from modeling runs from the Regional Acid Deposition Model (RADM). The
RADM contains a more detailed treatment of the atmospheric chemical transformations
and removal processes, and thus provides an interesting 'check' of the
CALMET/CALPUFF modeling results. The final domain selected extended west
through Illinois, north almost to the northern border of New York, south through most of
97
-------
Figure 1: Calpuff Modeling Domain
Figure 22. CALMET computational domain.
South Carolina, and covers most of southern New England (Figure 22). The MM4 data
has a grid interval of 80 km. CALMET was used to produce 40 km gridded wind and
meteorological fields. The fields had seven layers in the vertical: 10, 50, 150, 300, 750,
1500 and 3000 meters. Precipitation data were derived from the stations available on
the Samson CD-ROM obtained from the National Climatic Data Center. Given the large
computational domain, a lambert conformal projection scheme was selected. Figure 22
shows the location of the MM4 profile points throughout the domain, contours of terrain
elevations based on the MM4 grid points, and the locations of the surface stations from
which precipitation measurements were obtained and used.
To reduce the CALPUFF simulation times, a series of overlapping receptor grids
were constructed, with the density of receptor increasing in the Chesapeake Bay area.
The meteorological grid is displayed in Figure 22. The computation domain was sized
to be somewhat larger than the domain of the sources being modeled. They had
divided the source inventory into four 'zones', with zone 1 including sources within 50
km of the Chesapeake Bay, zone 2 including all sources out to 210 km, zone 3
including all sources out to 500, and zone 4 including all sources from 500 out to the
edge of the domain shown in Figure 22. The CALPUFF simulations where configured
to use the MESOPUFF II fine species chemical transformation scheme, and wet and
98
-------
dry fluxes for HN03 and N03" were calculated and stored. Constant background
concentrations of ozone (80 ppb) and ammonia (10 ppb) were specified.
CALPUFF was run in one-month segments on a 200 MHZ Pentium Pro
computer. Storage requirement per month were approximately 220 MB for the raw
MM4 data for the selected domain, 75 MB for the CALMET output file used to drive
CALPUFF, and 40 MB for the output files of wet and dry fluxes for two species (HN03
and N03"). Each month took approximately 10 hours to run.
Total deposition due to dry and wet processes were roughly equal, with predicted
wet deposition dominated by deposition of HN03 and dry deposition dominated by N03".
On a domain-wide basis , the average total Nitrogen deposition rate was calculated to
be about 3.64 kg/(hectare«year). This can be compared to the RADM results for 1990
(Dennis, 1997), which estimated the total Nitrogen deposition rate to be approximately
10 kg/hectare/year in the Chesapeake Bay area. In the most recent RADM simulations
(R. Dennis, personal communication), the observations in the vicinity of the
Chesapeake Bay and the current RADM seem to be in accord, with most of the wet and
dry deposition resulting from HN03. In the Demonstration Assessment (Section 4.1),
the low amount of wet deposition of HN03 estimated by CALPUFF was considered a
surprise, given the high aqueous phase solubility of HN03. Given these more recent
results and comparisons of RADM with observations, there is further cause to suspect
the wet deposition results for HN03 to be underestimated by CALPUFF.
Of interest is that this analysis has in fact used the FDDA-MM meteorological
data with the CALMET/CALPUFF modeling system successfully. They provide an
independent check of the resources needed. The results are consistent with those
found in performing the demonstration analyses, discussed in Section 4.1.
4.9.2 Use of CALMET to Develop Wind Fields
Scire and Robe (1997) reported on a series of enhancements to CALMET for
improving the characterization of wind fields in the presence of topographical features
that might be anticipated to induce strong upslope and downslope winds. Many
industrial facilities are located in river valleys where the terrain effects can have a
dominant influence on pollutant transport and dispersion. Furthermore, it is quite likely
that the surface wind observations available may not be representative of the flow in the
near vicinity of the facility (or conversely may only be representative in the near vicinity
and not representative of flow conditions downwind). Attempting to resolve the fine-
scale wind effects prognostic meteorological modeling is not only computationally
demanding, but often beyond the state-of-the-art. CALMET uses diagnostic (empirical)
models of these local-scale effects (upslope and downslope winds), which is
computationally more cost-effective and provides a pragmatic solution to a complex
problem.
99
-------
The original wind field module in CALMET, was based on the parameterizations
in the Diagnostic Wind Model (DWM) of Douglas and Kessler (1988). This module
computed a slope flow (a vector oriented in the drainage direction) which was added to
the initialized wind field components (Step 1 initialization). Time of day was used to
deduce whether the slope flow was upwind or downwind. The magnitude of the slope
flow was based on the local steepness of the terrain, the terrain elevation, the vertical
temperature lapse rate, and the maximum terrain height within a radius of influence
about the local point under consideration. An undesirable trait of this algorithm was that
it would yield larger slope flows for higher terrain elevations, all other factors being
equal. The original module used the ambient temperature lapse rate, whereas the
potential temperature lapse rate provides the proper characterization of the static
stability of the atmosphere. To address these concerns, a new downslope flow module
was developed based on the shooting flow parameterization of Mahrt (1982) and a new
upslope module was develop based on surface drag concepts. These new modules
are no longer restricted to the first layer within CALMET, but instead can affect flow in
upper layers.
The local sensible heat flux is an important parameter in deducing the magnitude
and direction of the slope flows. This module in CALMET was upgraded to account for
the angle of the terrain relative to the sun. For the sensible heat flux computations, the
original algorithms assumed the terrain was horizontal to the earth. An east facing
terrain slope will "feel" the effects of the sun's heating in the morning and cooling in the
afternoon sooner than a west facing terrain slope. To address this concern, algorithms
by Whiteman and Allwine (1986 ) were added to CALMET that compute the solar
radiation for a surface of arbitrary inclination and azimuth.
No matter how much skill is added in the computation of the slope flows, if the
initial guess field in the Step 1 initialization is too far off, CALMET will poorly
characterize the local wind flows. To address this concern, two enhancements were
added to CALMET. First, instead of relying on only the upper-air observations to
characterize the local vertical temperature profiles, an option was added to CALMET
that allows the vertical temperature profiles using similarity theory (van Ulden and
Holtslag, 1985). Second, a modification was added to CALMET that allows the user to
control the influence of the upper-air observations and local surface observations in the
construction of the Step 1 initialization wind fields. The user is allowed to specify height
dependent factors, B, ranging from -1 (no influence by upper air observations) to +1 (no
influence by surface observations). If the local surface observations are within the
valley, they might be restricted to influencing the Step 1 initialization only for heights
below the top of the valley, above which the upper-air observations could be given
dominance.
Scire and Robe (1997) illustrated the enhancements discussed above by
calculating wind fields in the Columbia River Valley. The region of interest included the
City of Wenatchee, WA and the region to the southeast of the city. Wenatchee is
located on both sides of the river and includes a relatively flat elevated plain. Figure 23
100
-------
shows the CALMET modeling domain. The terrain variation is substantial, extending
from under 200 meters elevations above mean sea level (MSL) to over 1700 meters
MSL at the higher peaks in the southwest part of the domain. The Columbia River runs
to the southeast, turns roughly east, and then turns again to the south over a distance
of 20 kilometers. The high terrain features change orientation from east-west to north-
south in the southern part of the domain.
Terrain Elevations
100 m contours
5250-
5245-
5240-
700
705
710
715
720
725
UTM
East
(km)
Figure 23. Terrain contours for the Wenachee, WA domain. Contour internal is 100
m, and the location of the Wenatchee (Pangborn) Airport is indicated by the filled
circle. The box inset shows the subdomain for which slope flows are presented.
To characterize the terrain, a grid resolution of 250 meters was used in
CALMET. The modeling domain consisted of 108 by 84 horizontal grid cells. Nine
levels were used in the vertical: 10, 50, 120, 230, 450, 800, 1250 and 2600 meters
above ground. The lower four levels are within the valley, the upper three levels are
above most of the terrain features. The local wind observations were available at the
location shown with a filled circle in Figure 23. This location is within the valley, and the
wind rose for the location shows the influence for the local winds to channel and align
with the local orientation of the valley. The height dependent factors, B, were set at -1,
-1, -1, -1, +0.5, +0.8, +1, +1, +1. The nearest upper air observation station is Spokane,
WA, which is over 200 km to the east of the CALMET domain.
101
-------
Downslope Flow
July 1, 1994-1 AM
UTM Jf
North
(km)
5252H
5242H
722 724
100m terrain contours
UTM
East
(km)
Figure 24. Down slope flow components for 1:00 AM July 1, 1994
for Wenatchee, WA . Wind vectors are plotted every 500 meters
(at every second grid point).
102
-------
UPSLOPE FLOW
July 1,1994 -3PM
UTM
North
(km)
5252^
5250-
5248^
524&v
5244^
5242^
716
718 720 722 724
2.5 m/s 100 m terrain contours
UTM
East
(km)
Figure 25. Upslope flow components for 3:00 PM July 1 , 1 994 for
Wenatchee, WA . Wind vectors are plotted every 500 meters (at
every second grid point).
Figure 24 shows the computed CALMET layer 1 downslope flows and Figure 25
shows the computed CALMET layer 1 upslope flows. The region shown in these
figures is the box inset of Figure 23. The moderately strong nighttime drainage flows
(1:00 AM, July 1, 1994) are shown in Figure 24. Figure 25 shows the computed
upslope flows, which are (as expected) weaker than the downslope flows. Upslope
flows are not expected to accelerate as rapidly as downslope flows.
103
-------
The pattern and magnitude of the CALMET winds were considered consistent
with expectations. These winds were developed using a single surface station in the
valley and an upper air station located well outside of the valley. The computed wind
fields show the expected diurnal cycle of nighttime drainage followed by daytime
upslope flows. The patterns are consistent with the strong terrain channeling of the
main valley and within the smaller side canyons. These wind fields would be expected
to better characterize the transport of the pollutants. It was anticipated that the proper
characterization of the local terrain influences on the wind field would provide a more
realistic characterization of the transport around major terrain features rather than
directly impacting apon such features.
These results are relevant to the Phase 2 recommendations. They demonstrate
that realistic wind fields can be produced using diagnostic techniques. They illustrate
that in some situations, the proper characterization of the terrain's influence on the wind
fields will require gridded terrain heights, with a grid resolution of order 250 m. This
would not be expected for resolving transport and dispersion of plumes that have
already traveled large distances and are already quite broad relative to the local terrain
features. But it is reasonable to expect that sources within such terrain would require
more care to be taken in the characterization of the wind fields, otherwise the transport
and resulting dispersion would be poorly characterized. These results also illustrate the
feasibility of performing local-scale impacts assessments of the transport and
dispersion of emissions from sources located in severe terrain, even when available
meteorological observations are sparse.
4.9.3 Kincaid SF6 and Lovett SO2 Comparisons
The goal of this study (Strimaitis et al., 1997) was to conduct an evaluation of
CALPUFF in comparison with ISCST3 (U.S. EPA, 1995c) for the Kincaid data set, and
CTDMPLUS (Perry et al., 1989) for the Lovett data set. Also, enhancements were
tested in a special version of CALPUFF (4.07), in which convective scaling
parameterizations and concepts were implemented to better characterize dispersion
from tall stacks during unstable conditions. Earlier versions of CALPUFF use a simple
Gaussian distribution to characterize the vertical distribution of puff material within the
convective boundary layer (CBL). As the puff grew, it soon filled the layer between the
ground and the mixing height, resulting in a uniform vertical distribution. Within this
framework, the primary effect of the convective motions in the mixed layer was to cause
a rapid growth in the vertical size of a puff. Depending on the CALPUFF dispersion
option selected, this growth rate is either parameterized by the stability class, scaled by
measured turbulence intensity, or scaled by a turbulence intensity computed from the
surface layer parameters.
In the last decade, modeling techniques that recognize the asymmetry of the
vertical dispersion process in the CBL have matured. These techniques explicitly
account for the differences between the distribution and strength of updrafts and
downdrafts in the layer as they relate to the ensemble-mean concentration distribution.
104
-------
One such technique that is simple yet effective is the "p.d.f." approach that relates the
probability density function of the vertical position of puff mass in the layer to the
skewed probability density function of vertical velocity. Using the superposition of two
Gaussian distributions to characterize the skewed p.d.f of the vertical velocity, the p.d.f.
model produces a "dual plume" formulation that maps the evolution of one plume that is
initially carried toward the ground in a (mean) downdraft, and a second plume that
initially rises toward the top of the mixed layer in a (mean) updraft. Subsequent
"reflections" from both the ground and the lid are simulated using image sources.
Because each of these plumes has its own mean vertical velocity and rate of spread,
the resulting vertical distribution of mass is skewed in much the same way as the
observed distributions. The initial downdraft plume is called the direct source because
it travels directly from an elevated source (accounting for the plume rise velocity) to the
ground, while the initial updraft plume is called the indirect source because it reaches
the ground only after traversing the full depth of the mixed layer.
U.S. EPA (1995d) extended the p.d.f. CBL formulation to include a simple way of
simulating the tendency of highly buoyant plumes to "loft" at the top of the mixed layer,
remaining there for some time before the convective eddies are able to overcome the
buoyancy and mix their mass to the surface. This formulation forms the basis of the
CBL component of AERMOD, U.S. EPA (1995d), and has been adapted for use in
CALPUFF (4.07). Strimaitis et al. (1997) adopted AERMOD's CBL parameterizations
for obtaining the mean updraft and downdraft properties, and also their novel simulation
of the lofting plume by means of the effective rise of the indirect source. However, the
CALPUFF algorithms for treating partial penetration and subsequent entrainment into
the mixed layer as the layer grows remain unchanged.
The Kincaid Generating Station is a coal-fired electric generating station with two
606 megawatt (MW) units vented through a single 187 meter stack. It is located in
Kincaid, Illinois, approximately 25 kilometers southeast of Springfield. The power plant
is in an area of relatively flat terrain surrounded by farmland. As part of an intensive
monitoring program sponsored by the Electric Power Research Institute (EPRI), SF6
tracer was released as a gas continuously through the stack for approximately 30
experiments of 6-9 hours in length during 1980 and 1981. A network of 200 samplers
located from 0.5 to 50 km from the stack measured hourly SF6 concentrations at the
ground level. The SF6 samplers were located approximately in arcs at distances of 0.5,
1, 2, 3, 5, 7, 10, 15, 20, 30, 40, and 50 km from the stack. In addition to
measurements of winds and turbulence made at four levels on a 100-m tower, other
meteorological data collected include vertical profile data from balloons, solar and net
radiation, and cloud cover.
In this study the meteorological data files were developed by emphasizing the
tower winds and temperatures, the "observed" mixing heights, and by computing the
surface layer parameters using the RAMMET (U.S. EPA, 1995e) preprocessor. One
major element in preparing the data files for CALPUFF relates to the transport time
from the stack to the outermost receptor arc, which is frequently greater than 1 hour.
105
-------
While plume models neglect this feature of the study, puff models do not. The tracer
gas was typically started 1 to 4 hours before the sampling network, so emissions during
the initial transport period are known. Therefore, the full meteorological data set was
used (not just the meteorological hours for when tracer data were available) to extract
data for complete days (24 hours) for each of the tracer experiments. This allowed
development of all of the data needed to properly simulate each period.
The Lovett Power Plant is located in the Hudson River Valley approximately 70
km north of New York City. The terrain in the area rises to approximately 330 meters
above the stack base elevation along a ridge 2.5 km north of the stack. The highest
point in the area is 340 meters above stack base at about 3 km north of the facility.
S02 emissions were released from a 145 meter stack containing two flues. Hourly
stack parameters (temperatures and flows) were derived from continuously reported
load data. The monitoring period is one full year. Twelve continuous S02 monitors
were located to the west, south, and north of the facility. Ten of the twelve monitors
were located in complex terrain from 2 km to 3.5 km from the stack. These ten
monitors were all located at or above the elevation of the stack top. Two monitors were
located south of the facility at distances of 2 km and 8.5 km for purposes of estimating
background concentrations.
Meteorological data were collected on a 100 meter tower located in the valley
south-southwest of the stack. The tower was instrumented at three levels (10 m, 50 m,
and 100 m). Wind speed, wind direction, and horizontal turbulence (oe) were measured
at all three levels. In addition, four 10 meter meteorological towers collecting wind
speed, wind direction and oe were located in an arc on the high terrain to the west,
west-northwest, north-northwest, and north of the stack. Much of the data needed to
apply CALPUFF to this site was prepared by Paumier et al. (1992) for their evaluation
of CTDMPLUS. These data files were used in this evaluation, and all but the hourly
emissions data could be applied directly to CALPUFF.
For the Kincaid data, quality indicators of 1, 2, and 3 had been assigned to the
data. A quality indicator of 1 meant that sampling was very incomplete and no
maximum concentration could reliably be determined for the arc. A quality indicator of
3 meant that a distinct pattern could be seen in the data for the arc, and a maximum
concentration could be determined reliably. Model performance results for the Kincaid
data set were presented for two subsets of the data. The first were comprised of all
arc-hours for which the peak concentration on the arc was given a quality indicator (Ql)
of 2 or 3; the second was the subset of these arc-hours with a Ql of 3.
106
-------
400
Legend
ISC3
CALPUFF(4.0)
CALPUFF(4.07)
100 200 300
Observed SF6(g/m3)x109
400
Figure 26. Q-Q plots for Kincaid, comparing observed surface concentration values
of SF6 with simulation results by CALPUFF(4.0), CALPUFF(4.07) and ISC3 for
Quality 2 and 3 data.
Figure 26 shows the Quantile-Quantile (Q-Q) plots for CALPUFF(4.07),
CALPUFF(4.0) and ISCST3 compared to the observations, for data of quality 2 and 3.
Although all models display a tendency to predict more zero-impacts (see inset) than is
observed, both versions of CALPUFF predict fewer zeros than ISCST3. This is
traceable to the fact that for all CALPUFF model runs, the option for partial plume
penetration was enabled. ISC3 sets all surface concentrations to zero whenever the
computed centerline of the plume is detected to be above the mixing height. CALPUFF
would do the same, except when the partial plume penetration option is enabled.
When this option is enabled, a computation is made based on the size of the dispersing
plume, and any mass below the mixing height is allowed to continue to disperse and
possibly impact surface receptors. While fewer observed concentrations are
underpredicted by both versions of CALPUFF (106 out of 586 values) than are
107
-------
underpredicted by ISCST3 (236 out of 586 values), there remains a substantial number
of large observed concentrations for which both models miss (zero is predicted)
250
O)
CD
LJ_
c/3
CD
CO
<
O
100
50
50
100
150
200
250
ISC3 Estimated SF6 (g/m3)x109
Figure 27. Scatter plot for Kincaid, comparing CALPUFF(4.0) and ISC3 simulation
results for surface concentration values of SF6 for Quality 2 and 3 data.
The CALPUFF(4.0) results in Figure 26 appear to be similar to the ISC3 results.
This is true in the sense that the computed maxima are of a similar value. But the
CALPUFF(4.0) runs were conducted using dispersion parameters based on similarity
theory, and a scatter plot of CALPUFF(4.0) versus ISC3, Figure 27, reveals that the two
models are not providing as similar of results as might be deduced from inspection of
Figure 26.
CALPUFF(4.07) is seen to overpredict the upper range of observed
concentrations, while CALPUFF(4.0) and ISCST3 underpredict throughout, and the
ranked distribution for CALPUFF(4.07) lies nearer the 1:1 line in Figure 26. How much
of this improved performance might be attributed to the p.d.f. algorithms for the CBL?
108
-------
As can be seen in Figure 26, the lower end of the ranked distributions for both versions
of CALPUFF appear to coincide, which implies that reasons for the elevated frequency
of zero-impact predictions are not related to the treatment of dispersion in the CBL.
The character of the higher end is quite different, with CALPUFF(4.0) producing
concentrations more like ISCST3 than CALPUFF(4.07). This suggests that the p.d.f. is
responsible for the improvements at the high end noted in Figure 26.
800
CALPUFF(4.07)
CTDMPLUS
600
CM
O
C/3
CD
03
400 —
200 —
200 400 600
Observed S02 (g/rri ) x10b
800
Figure 28. Q-Q plots for Lovett, comparing observed surface concentration values of
S02with simulation results CALPUFF(4.07) and CTDMPLUS
For the Lovett data set, Paumier et al. (1992) found that CTDMPLUS and RTDM
predictions were poorly correlated with the observed peak hourly concentrations. Given
that CALPUFF was applied with the same meteorological data, similar results were
expected and found. Figure 28 shows the Q-Q plot for CALPUFF(4.07) and
CTDMPLUS compared to the observations. Although both models tend to overpredict
the observed ranked distribution, CALPUFF predictions lie nearer the 1:1 line. The
109
-------
dashed line in Figure 28 indicates an overprediction of two, which is close to the results
depicted for CTDMPLUS.
The evaluation results presented suggests that the p.d.f. formulations tested in
CALPUFF(4.07) eliminates the tendency seen in the CALPUFF(4.0) and ISC3 to
underestimate the surface concentration values for the Kincaid data set, and
CALPUFF(4.07) overpredicts surface concentration values less than CTDMPLUS for
the Lovett data set. Correlation of hourly predicted and observed concentrations for the
Lovett data set are poor for both CALPUFF(4.07) and CTDMPLUS, suggesting that
more diagnostic analyses of this data set may be fruitful. For the Kincaid data,
CALPUFF (4.0) performance is similar to ISCS's. And for the Kincaid evaluation,
performance during periods with observed mixing heights less that about 600m
suggests that more mass remains in the mixed layer than is predicted by the models,
even with partial penetration enabled.
110
-------
5.0 CONCLUSIONS
This report provides a summary and status report of the activities sponsored and
initiated by the Interagency Workgroup on Air Quality Modeling (IWAQM). The IWAQM
was formed to provide a focus for development of air quality models for assessment of
pollutant source impacts on Federal Class I areas and wilderness areas. In particular,
IWAQM has focused attention on providing modeling techniques for assessing possible
adverse air quality impacts resulting from long-range transport of pollutants, as required
by the prevention of significant deterioration (PSD) program.
In 1993, IWAQM provided interim Phase 1 recommendations that provided the
best approach from existing "off-the-shelf-techniques." The MESOPUFF II puff
modeling system was recommended for use. This model provided the ability to
simulate the influence of time and space varying meteorological conditions on transport
and dispersion. It provided a first-order approximate characterization of the formation
of sulfate and nitrate during transport downwind. It provided characterizations for the
removal of pollutants by dry and wet deposition. The meteorology processing had
limited capabilities, as it was not able to characterize geographical terrain slope flows.
The puff dispersion model had no capabilities to address effects associated with
variation of terrain heights. Furthermore, due to limitations in the dispersion treatments,
the puff model was not considered appropriate for use in characterizing impacts
associated with transport distances of less than 50 km.
As discussed in Appendix D, IWAQM provided a status report of ongoing
activities at the Sixth Modeling Conference, which was held in Washington, D.C. August
9-10, 1995. As a result of comments received, a series of investigations were
undertaken. Comparisons were made of CALPUFF simulated dispersion with near-
surface concentrations collected during several tracer field studies, where the transport
distances were of the order 50 to 300 km. Comparisons were made of CALPUFF
simulated dispersion with simulation results obtained using the Industrial Source
Complex (ISC3) plume dispersion model. Initially, the focus was on whether CALPUFF
and ISC3 gave similar results for steady-state meteorological conditions. Once this was
confirmed, the focus shifted to investigating what difference might result between a puff
and plume model, having identical meteorology and dispersion. It became clear that
wind reversals and buildup of pollutant mass during calm wind conditions could result in
large differences in simulation results. As a consequence of comments received and of
the puff/plume model comparisons, a new screening analysis was tested that relied on
use of the CALPUFF dispersion model with a greatly simplified characterization of the
hourly meteorological conditions.
Based on the findings of the various investigations summarized in Section 4,
IWAQM is providing a Phase 2 recommendation to replace the interim Phase 1
recommendation. The CALPUFF modeling system is recommended in place of the
MESOPUFF II modeling system for a number of reasons. A primary consideration is
that the CALMET meteorological processor is capable of diagnostically characterizing
111
-------
geographic terrain slope flows, and has been updated to allow use of sophisticated
output from modern mesoscale meteorological processors. Another important
consideration is that the CALPUFF puff dispersion algorithms have been fashioned to
allow characterization of both local-scale and long-range transport and dispersion. This
allows use of one model for all sources with a consistent treatment of the chemistry and
fate of the pollutants. Using CALPUFF within the new screening analysis provides
consistent treatment of the chemistry and fate, and some of the 'causality' effects of
hourly varying meteorology that a puff dispersion model can treat more directly and
easily than a standard plume dispersion model.
This Phase 2 recommendation provides a major improvement in the treatment
and characterization of the meteorological conditions and the mesoscale transport.
There is yet room for improvement in the characterization of the chemistry and fate of
the pollutants. The CALPUFF dispersion model represents a major improvement over
the MESOPUFF II model in its flexibility for treating a variety of source types, time
varying emission rates, and of dispersion situations. The IWAQM recommends the
CALPUFF modeling system for use as a refined long-range transport and dispersion
modeling technique for characterizing reasonably attributable pollutant impacts from
one or a few sources.
112
-------
6.0 REFERENCES
Anthes, R.A., E.Y. Hsie, and Y.H. Kuo, 1987: Description of the Penn State/NCAR
mesoscale model version 4 (MM4). NCAR Tech Note, NCAR/TN-282+STR, Available
from the National Center for Atmospheric Research, PO Box 3000, Boulder CO 80307,
66 pp.
Bowne, N.E., R.J. Londergan, D.R. Murray, and H.S. Bornstein, 1983: Overview,
results, and conclusions for the EPRI plume model validation project: plains site. EA-
3074, EPRI, Palo Alto, CA.
Bullock, O.R., 1993: A Workstation Concept For The Production of Dynamically-
Constrained Meteorological Characterizations For Use in Air-Quality Modeling.
Preprints to the Ninth International Conference on Interactive Information and
Processing Systems For Meteorology, Oceanography, and Hydrology. January 17-22,
1993, Anaheim, CA. American Meteorological Society. Boston MA. pp 9-14.
Carhart, R.A., A.J. Policastro, M. Wastag, and L Coke, 1989. Evaluation of Eight
Short-term Long-range Transport Models using Field Data. Atmospheric Environment,
23(1):85-105.
Chen, F., Z. Janjio and K. Mitchell, 1997. Impact of atmospheric surface-layer
parameterizations in the new land-surface scheme of the NCEP mesoscale ETA model.
Boundary-Layer Meteorology, (85):391-421.
Cimorelli, A.J., S.G. Perry, R.F. Lee, R.J. Paine, A. Venkatram, J.C. Weil, and R.B.
Wilson, 1996: Current progress in the AERMIC model development program. AWMA
89th Annual Meeting and Exhibition, Nashville, TN, June 23-28.
Clements, W.E (ed.)., 1979: Experimental Design and Data of the April 1977
Multitracer Atmospheric Experiment at the Idaho National Engineering Laboratory, Los
Alamos Scientific Laboratory Informal Report, LA-7795-MS/UC-11, Los Alamos, NM
87545, U.S.A., 100pp.
Davis, W.E., A.R. Olsen, B.T. Didierand D.W. Damschen, 1986: A comparison of
Meso-Alpha MLAM Model Results with Observations from CAPTEX. 5th Joint
Conference On Applications of Air Pollution Meteorology, American Meteorological
Society, Chapel Hill, NC, 69-72.
Dennis, R.L., 1997: Using the Regional Acid Deposition Model to Determine the
Nitrogen Deposition Airshed of the Chesapeake Bay Watershed. Atmospheric
Deposition to the Great Lakes and Coastal Waters. Society of Environmental
Toxicology and Chemistry. (In Press).
113
-------
Douglass, S., and R. Kessler, 1988: User's Guide to the Diagnostic Wind Field Model
(Version 1.0). Systems Applications, Inc., San Rafael, CA, 48 pp.
Draxler, R.R., 1976: Determination of atmospheric diffusion parameters. Atmospheric
Environment, 10, 99-105.
Draxler, R.R., 1979: Modeling the results of two recent mesoscale dispersion
experiments. Atmospheric Environment, 13, 1523-1533.
Draxler, R.R., 1987: Sensitivity of a Trajectory Model to the Spatial and Temporal
Resolution of the Meteorological data during CAPTEX. Journal of Climate and Applied
Meteorology, 26, 1577-1588.
Eckhoff, P.A., and C.T. Coulter, 1998: CALPUFF vs. ISCST3 Sensitivity Study.
Proceedings of the 10th Joint Conference on the Applications of Air Pollution
Meteorology with the Air and Waste Management Association. January 11-16, 1998,
Phoenix, AZ.
Ferber, G., Telegadas, K., Heffter, J., Dickson, C., Dietz, R., and Krey, P., 1981:
Demonstration of a Long-range Transport Atmospheric Tracer System Using
Perfluorocarbons. NOAA Technical Memorandum, ERLARL-101.
Ferber, G.J., J.L. Heffter, R.R. Draxler, R.J. Lagonmarsino, F.L. Thomas, R.N. Dietz,
and C.M. Benkovitz, 1986: Cross-Appalachian Tracer Experiment (CAPTEX '83) Final
Report. NOAA Technical Memorandum ERL ARL-142, Air Resources Laboratory,
NOAA Environmental Research Laboratories, Silver Spring, MD 20910, 60 pp.
Fox, D., 1981: Judging Air Quality Model Performance. Bulletin of the American
Meteorological Society, 62, 599-609.
Godowitch, J.M. 1989a. Evaluation and Sensitivity Analyses Results of the
MESOPUFF II Model with CAPTEX Measurements. EPA-600/3-89-056. U.S.
Environmental Protection Agency, Atmospheric Research and Exposure Assessment
Laboratory, Research Triangle Park, NC.
Godowitch, J.M., 1989b: Evaluation and Testing of the MESOPUFF II Model System
with CAPTEX Measurements. 6th Joint Conference On Applications of Air Pollution
Meteorology, American Meteorological Society, Anaheim, CA, 56-59.
Hoecker, W., 1963: Three Southerly Low-level Jet Systems Delineated by the Weather
Bureau Special Pibal Network of 1961. Monthly Weather Review, 91, 573-582.
Irwin, J.S., and J.O. Paumier, 1990: Characterizing the Dispersive State of Convective
Boundary Layers for Applied Dispersion Modeling. Boundary-Layer Meteorology.
(53):267-296.
114
-------
Irwin, J.S., J.S. Scire, and D.G. Strimaitis, 1996: A Comparison of CALPUFF Modeling
Results with CAPTEX Field Data Results. Air Pollution Modeling and Its Application XI.
Edited by S.E. Gryning and F.A. Schiermeier. Plenum Press, New York, NY., pg 603-
611.
Irwin, J.S., 1997: A Comparison of CALPUFF Modeling Results with 1977 INEL Field
Data Results. Air Pollution Modeling and Its Application, XII. Edited by S.E. Gryning
and N. Chaumerliac, Plenum Press, New York, NY. 8 pp. (In Press).
Langford, A.O., F.C. Fehsenfeld, J. Zachariassen, and D.S. Schimel, 1992: Gaseous
Ammonia fluxes and Background Concentrations in Terrestrial Exosystems of The
United States. Global Biogeochemical Cycles. Vol. 6(4):459-483.
Larsen, R.I., A.S. Heagle, and W.W. Heck, (1983): An Air Quality Data Analysis
System for Interrelating Effects, Standards, and Needed Source Reductions: Part 7. An
03-S02 Leaf Injury Mathematical Model. Journal of Air Pollution and Control
Association. Vol. 33(3): 198-207.
Malm, W.C., J.V. Molenar, R.A. Eldred, J.F. Sisler, (1996): Examining the Relationship
Among Atmospheric Aerosols and Light Scattering and Extinction in the Grand Canyon
Area. Journal of Geophysical Research. Vol. 101(014): 19251-19265.
Moran, M.D., and R.A. Pielke, 1994: Delayed Shear Enhancement in Mesoscale
Atmospheric Dispersion. 8th Joint Conference On Applications of Air Pollution
Meteorology, American Meteorological Society, Nashville, TN, 96-103.
Moran, M. and Pielke, R., 1995a: Evaluation of a Mesoscale Atmospheric Dispersion
Modeling System with Observations from the 1980 Great Plains Mesoscale Tracer Field
Experiment. Parti: Datasets and Meteorological Simulations. Journal of Applied
Meteorology, 35, 281-307.
Moran, M. and Pielke, R., 1995b: Evaluation of a Mesoscale Atmospheric Dispersion
Modeling System with Observations from the 1980 Great Plains Mesoscale Tracer Field
Experiment. Part I: Dispersion Simulations. Journal of Applied Meteorology, 35, 308-
329.
Morris, R.E., R.C. Kessler, S.G. Douglas, K.R. Styles, and G.E. Moore. 1988: Rocky
Mountain Acid Deposition Model Assessment: Acid Rain Mountain Mesoscale Model
(ARMS). EPA-600/3-88-042. U.S. Environmental Protection Agency, Atmospheric
Research and Exposure Assessment Laboratory, Research Triangle Park, NC.
National Climatic Data Center (NCDC), 1995: MM4 -1990 Meteorology Data, Federal
Building, 151 Patton Avenue, Asheville, NC 28801-5001.
115
-------
Paumier, J.O., S.G. Perry, and D.J. Burns, 1992: CTDMPLUS: A dispersion model for
sources near complex topography. Part II: Performance characteristics. Journal of
Applied Meteorology, 31, 646-660.
Perry, S.G., D.J. Burns, L.H. Adams, R.J. Paine, M.G. Dennis, M.T. Mills, B.C.
Strimaitis, R.J. Yamartino, E.M. Insley, 1989. User's Guide to the Complex Terrain
Dispersion Model Plus Algorithms for Unstable Situations (CTDMPLUS) Volume 1:
Model Description and User Instructions. EPA/600/8-8/041. U.S. Environmental
Protection Agency, Research Triangle Park, NC.
Pielke, R.A., and M. Uliasz, 1997: Use of Meteorological Models as Input to Regional
and Mesoscale Air Quality Models - Limitations and Strengths. Atmospheric
Environment. (In Press).
Pielke, R.A., T.J. Lee, J.H. Copeland, J.L. Eastman, C.L. Ziegler, and C.A. Finley,
1997: Use of USGS- Provided Data to Improve Weather and Climate Simulations.
Ecological Applications. Vol 7 (1 ):3-21.
Sherwell, J., and M. Garrison, 1997: An Investigation Into Using the
CALMET/CALPUFF Modeling System for Assessing Nitrogen Deposition in the
Chesapeake Bay. Proceedings of the Air & Wast Management Association's 90th
Annual Meeting & Exhibition, June 8-13, Toronto, Ontario, Canada. Paper 97-TA2A.05.
20 pages.
Schulze, R.H., and D.B. Turner, 1998: Potential use of NOAA-archeived meteorological
observations to improve air dispersion model performance. Environmental Manager.
Air & Wast Management Association, March Issue, pg 12-21.
Scire, J.S., R.J. Yamartino, G.R. Carmichael and Y.S. Chang, 1989: CALGRID: A
mesoscale photochemical grid model. Volume II. Sigma Research Corporation,
Concord,
Scire, J.S., E.M. Insley and R.J. Yamartino, 1990a: Model evaluation and user's guide
for the CALMET meteorological model. California Air Resource Board, Sacramento,
CA.
Scire, J.S. D.G. Strimaitis, and R.J. Yamartino, 1990b: Model formulation and user's
guide for the CALPUFF dispersion model. California Air Resources Board,
Sacramento, CA.
Scire, J.S., D.G. Strimaitis, R.J. Yamartino, and E.M. Insley, 1995a: User's Guide for
the CALPUFF Dispersion Model. Earth Tech, Inc., Concord, MA
Scire, J.S., E.M. Insley, and R.J. Yamartino, and M.E. Fernau, 1995b: User's Guide for
the CALMET Meteorological Model. Earth Tech, Inc., Concord, MA.
116
-------
Shi, B., J.D. Kahl, Z.D. Christidis, and P.J. Samson, 1990: Simulation of the Three-
Dimensional Distribution of Tracer During the Cross-Appalachian Tracer Experiment.
Journal of Geophysical Research, 95, No. D4, 3693-3703.
Sisler, J.F. (Editor), 1996: Spatial and Seasonal Patterns and Long Term Variability of
the Composition of the Haze in the united States: An Analysis of data from the
IMPROVE Network. ISSN: 0737-5352-32. Cooperative Institute for Research in the
Atmosphere. Colorado State University, Fort Collins, CO.
Stauffer, D.R. and N.L. Seaman, 1989: Use of four dimensional data assimilation in a
limited-area mesoscale model. Part I: Experiments with synoptic-scale data. Monthly
Weather Review, 118, 1250-1277.
Stauffer, D.R., N.L. Seaman, and F.S. Binkowski, 1990: Use of four dimensional data
assimilation within the planetary boundary layer. Monthly Weather Review, 119, 734-
754.
Strimaitis, D.G., J.S. Scire and J.S. Chang, 1997: Evaluation of the CALPUFF Model
with the Lovett (S02) and Kincaid (SF6) Power Plant Data Sets. Contract Number 700-
96-001. California Energy Commission, Sacramento, CA and Jersey Central Poser and
Light, Trenton, NJ, Project report (Document 21861-03) available from Earth Tech, Inc.
Concord, MA.
Turner, D.B., and J. DeToro, 1998: What air pollution meteorologists should know
about data resulting from NOAA forecast models. 10th Joint Conference On
Applications of Air Pollution Meteorology, American Meteorological Society, Phoenix,
AZ, 584-587.
U.S. Department of Energy, 1978: Heavy Methane-SF6 Tracer Test Conducted at the
Savannah River Laboratory, December 10, 1975. DP-1469. Prepared by E.I. du Pont
de Nemours and Company, Savannah River Laboratory, Aiken, South Carolina.
U.S. Environmental Protection Agency, 1977: Guidelines for the Regional Evaluation of
State and Local New Source Review Program. EPA-450/2-77-027. U.S. Environmental
Protection Agency, Research Triangle Park, NC. (NTIS PB-275053).
U.S. Environmental Protection Agency, 1980: Prevention of Significant Deterioration
Workshop Manual. EPA-450/2-80-081. U.S. Environmental Protection Agency,
Research Triangle Park, NC. (NTIS PB-81-201410).
U.S. Environmental Protection Agency, 1992a. Interagency Workgroup on Air Quality
Modeling (IWAQM) Workplan Rationale. EPA-454/R-92-001. U.S. Environmental
Protection Agency, Office of Air Quality Planning and Standards, Research Triangle
Park, NC.
117
-------
U.S. Environmental Protection Agency, 1992b. User's Guide for the Industrial Source
Complex (ISC2) Dispersion Models. EPA-454/4-92-008. U.S. Environmental
Protection Agency, Office of Air Quality Planning and Standards, Research Triangle
Park, NC.
U.S. Environmental Protection Agency, 1993. Interagency Workgroup on Air Quality
Modeling (IWAQM) Phase 1 Report: Interim Recommendation for Modeling Long
Range Transport and Impacts on Regional Visibility. EPA-454/R-93-015. U.S.
Environmental Protection Agency, Office of Air Quality Planning and Standards,
Research Triangle Park, NC.
U.S. Environmental Protection Agency, 1994: Revised User's Guide to the
MESOPUFF II (V5.1) EPA-454/B-94-025. U.S. Environmental Protection Agency,
Atmospheric Research and Exposure Assessment Laboratory, Research Triangle Park,
NC.
U.S. Environmental Protection Agency, 1995a: Inter-agency Workgroup on Air Quality
Modeling (IWAQM): Assessment of Phase 1 Recommendations Regarding The Use of
MESOPUFF II. EPA-454/R-95-006. U.S. Environmental Protection Agency,
Atmospheric Research and Exposure Assessment Laboratory, Research Triangle Park,
NC.
U.S. Environmental Protection Agency, 1995b: Testing of Meteorological and
Dispersion Model for Use in Regional Air Quality Modeling. EPA-454/R-95-005. U.S.
Environmental Protection Agency, Atmospheric Research and Exposure Assessment
Laboratory, Research Triangle Park, NC.
U.S. Environmental Protection Agency, 1995c: User's Guide for the Industrial Source
Complex (ISC3) Dispersion Models. EPA-454/B-95-003a. U.S. EPA, Research
Triangle Park, NC.
U.S. Environmental Protection Agency, 1995d: AERMIC model (AERMOD) evaluation
design and implementation. U.S. EPA document dated July 27, 1995. U.S. EPA,
Research Triangle Park, NC.
U.S. Environmental Protection Agency, 1995e: PCRAMMET User's Guide (Draft).
U.S. EPA document dated July, 1995. U.S. EPA, Research Triangle Park, NC.
U.S. Environmental Protection Agency, 1996: CALMET, CALPUFF and CALPOST
Modeling System, PB 96-502-083INC (CD ROM and diskette), National Technical
Information Service, U.S. Department of Commerce, Springfield, VA, 22161.
U.S. Environmental Protection Agency, 1997. Appendix W to 40 CFR Part 51:
Guideline on Air Quality Models. Code of Federal Regulations, Office of the Federal
118
-------
Register National Archives and Records Adminsitration, U.S. Government Printing
Office, Washington, DC. Pg 346-438
U.S. Environmental Protection Agency, 1998a: A Comparison of CALPUFF Modeling
Results to Two Tracer Field Experiments. EPA-454/R-98-009, U.S. EPA, Research
Triangle Park, NC.
U.S. Environmental Protection Agency, 1998b: Analysis of the CALMET/CALPUFF
Modeling System in a Screening Mode. EPA-454/R-98-030. U.S. EPA, Research
Triangle Park, NC.
Van Ulden, A.P., and A.A.M. Holtslag, 1985: Estimation of Atmospheric Boundary Layer
Paramters for Diffusion Applications. Journal of Climate and Applied Meteorology,
(24): 1196-1207.
Vimont, J.C., 1998: Evaluation of the CALMET/CALPUFF Modeling System in the
Southwestern U.S. 10th Joint Conference on the Applications of Air Pollution
Meteorology with the A&WMA, American Meteorological Society, Phoenix, AZ, 30-34.
Whiteman, C.D., and K.J. Allwine, 1986: Extraterrestiral Solar Radiation on Included
Surfaces. Environmental Software. Vol. 1(3): 164-169.
119
-------
APPENDIX A
CALMET RECOMMENDATIONS
In the following, a listing is provided of the defaults currently assumed in
CALMET for long-range transport analyses in involving assessments of not on
concentration impacts, but also deposition flux impacts and visibility impacts. Some of
the variables have the Value' is listed in bold. This is meant to indicate that these likely
will need to be tailored for a given application.
A-l
-------
Variable
GEO.DAT
SURF. DAT
PRECIP.DAT
NUSTA
UPn.DAT
IBYR
IBMO
IBDY
IBHR
IBTZ
IRLG
IRTYPE
LCALGRD
NX
NY
DGRIDKM
XORIGKM
YORIGKM
XLATO
YLONO
IUTMZN
LLCONF
XLAT1
XLAT2
Description
Name of Geophysical data file
Name of Surface data file
Name of Precipitation data file
Number of upper air data sites
Names of NUSTA upper air data files
Beginning year
Beginning month
Beginning day
Beginning hour
Base time zone
Number of hours to simulate
Output file type to create (must be 1 for
CALPUFF)
Are w-components and temperature needed?
Number of east-west grid cells
Number of north-south grid cells
Grid spacing
Southwest grid cell X coordinate
Southwest grid cell Y coordinate
Southwest grid cell latitude
Southwest grid cell longitude
UTM Zone
When using Lambert Conformal map
coordinates, rotate winds from true north to map
north?
Latitude of 1st standard parallel
Latitude of 2nd standard parallel
Value
GEO.DAT
SURF. DAT
PRECIP.DAT
User Defined
UPn.DAT
User Defines
User Defines
User Defines
User Defines
User Defines
User Defines
1
T
User Defines
User Defines
User Defines
User Defines
User Defines
User Defines
User Defines
User Defines
F
30
60
A-2
-------
Variable
RLONO
RLATO
NZ
ZFACE
LSAVE
IFORMO
NSSTA
NPSTA
ICLOUD
FORMS
IFORMP
IFORMC
IWFCOD
IFRADJ
IKINE
IOBR
ISLOPE
IEXTRP
ICALM
BIAS
Description
Longitude used if LLCONF = T
Latitude used if LLCONF = T
Number of vertical layers
Vertical cell face heights (NZ+1 values)
Save met. data fields in an unformatted file?
Format of unformatted file (1 for CALPUFF)
Number of stations in SURF. DAT file
Number of stations in PRECIP.DAT
Is cloud data to be input as gridded fields? (0 =
No)
Format of surface data (2 = formatted)
Format of precipitation data (2 = formatted)
Format of cloud data (2 = formatted)
Generate winds by diagnostic wind module? (1 =
Yes)
Adjust winds using Froude number effects? (1 =
Yes)
Adjust winds using kinematic effects? (1 = Yes)
Use O'Brien procedure for vertical winds? (0 =
No)
Compute slope flows? (1 = Yes)
Extrapolate surface winds to upper layers? (-4 =
use similarity theory and ignore layer 1 of upper
air station data)
Extrapolate surface calms to upper layers? (0 =
No)
Surface/upper-air weighting factors (NZ values)
Value
90
40
User Defines
User Defines
T
1
User Defines
User Defines
0
2
2
2
1
1
0
0
1
-4
0
NZ*0
A-3
-------
Variable
IPROG
LVARY
RMAX1
RMAX2
RMAX3
RMIN
RMIN2
TERRAD
R1
R2
DIVLIM
NITER
NSMTH
NINTR2
CRITFN
ALPHA
IDIOPT1
ISURFT
IDIOPT2
IUPT
ZUPT
Description
Using prognostic or MM-FDDA data? (0 = No)
Use varying radius to develop surface winds?
Max surface over-land extrapolation radius (km)
Max aloft over-land extrapolation radius (km)
Maximum over-water extrapolation radius (km)
Minimum extrapolation radius (km)
Distance (km) around an upper air site where
vertical extrapolation is excluded (Set to -1 if
IEXTRP = ±4)
Radius of influence of terrain features (km)
Relative weight at surface of Step 1 field and obs
Relative weight aloft of Step 1 field and obs
Maximum acceptable divergence
Max number of passes in divergence
minimization
Number of passes in smoothing (NZ values)
Max number of stations for interpolations (NA
values)
Critical Froude number
Empirical factor triggering kinematic effects
Compute temperatures from observations (0 =
True)
Surface station to use for surface temperature
(between 1 and NSSTA)
Compute domain-average lapse rates? (0 =
True)
Station for lapse rates (between 1 and NUSTA)
Depth of domain-average lapse rate (m)
Value
0
F
User Defines
User Defines
User Defines
0.1
4
User Defined
User Defines
User Defines
5.E-6
50
2, 4*(NZ-1)
99
1
0.1
0
User Defines
0
User Defines
200
A-4
-------
Variable
IDIOPT3
IUPWND
ZUPWND
IDIOPT4
IDIOPT5
CONSTB
CONSTE
CONSTN
CONSTW
FCORIOL
IAVEXZI
MNMDAV
HAFANG
ILEVZI
DPTMIN
DZZI
ZIMIN
ZIMAX
ZIMINW
ZIMAXW
IRAD
TRAD KM
Description
Compute internally inital guess winds? (0 = True)
Upper air station for domain winds (-1 = 1/r**2
interpolation of all stations)
Bottom and top of layer for 1st guess winds (m)
Read surface winds from SURF. DAT? ( 0 =
True)
Read aloft winds from UPn.DAT? (0 = True)
Neutral mixing height B constant
Convective mixing height E constant
Stable mixing height N constant
Over-water mixing height W constant
Absolute value of Coriolis parameter
Spatial averaging of mixing heights? (1 = True)
Max averaging radius (number of grid cells)
Half-angle for looking upwind (degrees)
Layer to use in upwind averaging (between 1
and NZ)
Minimum capping potential temperature lapse
rate
Depth for computing capping lapse rate (m)
Minimum over-land mixing height (m)
Maximum over-land mixing height (m)
Minimum over-water mixing height (m)
Maximum over-water mixing heigh (m)
Form of temperature interpolation (1 = 1/r)
Radius of temperature interpolation (km)
Value
0
-1
1, 1000
0
0
1.41
0.15
2400
0.16
1.E-4
1
1
30
1
0.001
200
50
3000
50
3000
1
500
A-5
-------
Variable
NUMTS
IAVET
TGDEFB
TGDEFA
JWAT1
JWAT2
NFLAGP
SIGMAP
CUTP
SSn
USn
PSn
Description
Max number of stations in temperature
interpolations
Conduct spatial averaging of temperature? (1 =
True)
Default over-water mixed layer lapse rate (K/m)
Default over-water capping lapse rate (K/m)
Beginning landuse type defining water
Ending landuse type defining water
Method for precipitation interpolation (2 = 1/r**2)
Precip radius for interpolations (km)
Minimum cut off precip rate (mm/hr)
NSSTA input records for surface stations
NUSTA input records for upper-air stations
NPSTA input records for precipitation stations
Value
5
1
-0.0098
-0.0045
999
999
2
100
0.01
User Defines
User Defines
User Defines
A-6
-------
APPENDIX B
CALPUFF RECOMMENDATIONS
In the following, a listing is provided of the defaults currently assumed in
CALPUFF for long-range transport analyses in involving assessments of not on
concentration impacts, but also deposition flux impacts and visibility impacts. Some of
the variables have the Value' is listed in bold. This is meant to indicate that these likely
will need to be tailored for a given application.
B-l
-------
Variable
METDAT
PUFLST
CONDAT
DFDAT
WFDAT
VISDAT
METRUN
IBYR
IBMO
IBDY
IBHR
IRLG
NSPEC
NSE
M RESTART
METFM
AVET
MGAUSS
MCTADJ
MCTSG
MSLUG
Description
CALMET input data filename
Filename for general output from CALPUFF
Filename for output concentration data
Filename for output dry deposition fluxes
Filename for output wet deposition fluxes
Filename for output relative humidities (for
visibility)
Do we run all periods (1 ) or a subset (0)?
Beginning year
Beginning month
Beginning day
Beginning hour
Length of run (hours)
Number of species modeled (for MESOPUFF II
chemistry)
Number of species emitted
Restart options (0 = no restart), allows splitting
runs into smaller segments
Format of input meteorology (1 = CALMET)
Averaging time lateral dispersion parameters
(minutes)
Near-field vertical distribution (1 = Gaussian)
Terrain adjustments to plume path (3 = Plume
path)
Do we have subgrid hills? (0 = No), allows
CTDM-like treatment for subgrid scale hills
Near-field puff treatment (0 = No slugs)
Value
CALMET.DAT
CALPUFF. LST
CONC.DAT
DFLX.DAT
WFLX.DAT
VISB.DAT
0
User Defined
User Defined
User Defined
User Defined
User Defined
5
3
0
1
60
1
3
0
0
B-2
-------
Variable
MTRANS
MTIP
MSHEAR
MSPLIT
MCHEM
MWET
MDRY
MDISP
MTURBVW
MDISP2
MROUGH
MPARTL
MTINV
MPDF
MSGTIBL
MREG
CSPECn
Specie
Names
Specie
Groups
NX
NY
NZ
Description
Model transitional plume rise? (1 = Yes)
Treat stack tip downwash? (1 = Yes)
Treat vertical wind shear? (0 = No)
Allow puffs to split? (0 = No)
MESOPUFF-II Chemistry? (1 = Yes)
Model wet deposition? (1 = Yes)
Model dry deposition? (1 = Yes)
Method for dispersion coefficients (3 = PG & MP)
Turbulence characterization? (Only if MDISP = 1
or 5)
Backup coefficients (Only if MDISP = 1 or 5)
Adjust PG for surface roughness? (0 = No)
Model partial plume penetration? (0 = No)
Elevated inversion strength (0 = compute from
data)
Use PDF for convective dispersion? (0 = No)
Use TIBL module? (0 = No) allows treatment of
subgrid scale coastal areas
Regulatory default checks? (1 = Yes)
Names of species modeled (for MESOPUFF II,
must be S02, S04, NOX, HN03, N03)
Manner species will be modeled
Grouping of species, if any.
Number of east-west grids of input meteorology
Number of north-south grids of input meteorology
Number of vertical layers of input meteorology
Value
1
1
0
0
1
1
1
3
3
3
0
1
0
0
0
1
User Defined
User Defined
User Defined
User Defined
User Defined
User Defined
B-3
-------
Variable
DGRIDKM
ZFACE
XORIGKM
YORIGIM
IUTMZN
XLAT
XLONG
XTZ
IBCOMP
JBCOMP
IECOMP
JECOMP
LSAMP
IBSAMP
JBSAMP
IESAMP
JESAMP
MESHDN
ICON
IDRY
IWET
MS
LCOMPRS
Description
Meteorology grid spacing (km)
Vertical cell face heights of input meteorology
Southwest corner (east-west) of input
meteorology
Southwest corner (north-south) of input
meteorology
UTM zone
Latitude of center of meteorology domain
Longitude of center of meteorology domain
Base time zone of input meteorology
Southwest X-index of computational domain
Southwest Y-index of computational domain
Northeast X-index of computational domain
Northeast Y-index of computational domain
Use gridded receptors? (T = Yes)
Southwest X-index of receptor grid
Southwest Y-index of receptor grid
Northeast X-index of receptor grid
Northeast Y-index of receptor grid
Gridded recpetor spacing = DGRIDKM/MESHDN
Output concentrations? (1 = Yes)
Output dry deposition flux? (1 = Yes)
Output west deposition flux? (1 = Yes)
Output RH for visibility calculations (1 = Yes)
Use compression option in output? (T = Yes)
Value
User Defined
User Defined
User Defined
User Defined
User Defined
User Defined
User Defined
User Defined
User Defined
User Defined
User Defined
User Defined
F
User Defined
User Defined
User Defined
User Defined
1
1
1
1
1
T
B-4
-------
Variable
ICPRT
IDPRT
IWPRT
ICFRQ
IDFRQ
IWFRQ
IPRTU
IMESG
Output
Species
LDEBUG
Dry Gas Dep
Dry Part. Dep
RCUTR
RGR
REACTR
MINT
IVEG
Wet Dep
MOZ
BCK03
BCKNH3
RNITE1
RNITE2
RNITE3
Description
Print concentrations? (0 = No)
Print dry deposition fluxes (0 = No)
Print wet deposition fluxes (0 = No)
Concentration print interval (1 = hourly)
Dry deposition flux print interval (1 = hourly)
West deposition flux print interval (1 = hourly)
Print output units (1 = g/m**3; g/m**2/s)
Status messages to screen? (1 = Yes)
Where to output various species
Turn on debug tracking? (F = No)
Chemical parameters of gaseous deposition
species
Chemical parameters of particulate deposition
species
Reference cuticle resistance (s/cm)
Reference ground resistance (s/cm)
Reference reactivity
Number of particle-size intervals
Vegetative state (1 = active and unstressed)
Wet deposition parameters
Ozone background? (1 = read from ozone.dat)
Ozone default (ppb) (Use only for missing data)
Ammonia background (ppb)
Nighttime S02 loss rate (%/hr)
Nighttime NOx loss rate (%/hr)
Nighttime HN03 loss rate (%/hr)
Value
0
0
0
1
1
1
1
1
User Defined
F
User Defined
User Defined
30.
10.
8
9
1
User Defined
1
80
10
0.2
2
2
B-5
-------
Variable
SYTDEP
MHFTSE
JSUP
CONK1
CONK2
TBD
IURB1
IURB2
Description
Horizontal size (m) to switch to time dependence
Use Heffter for vertical dispersion? (0 = No)
PG Stability class above mixed layer
Stable dispersion constant (Eq 2.7-3)
Neutral dispersion constant (Eq 2.7-4)
Transition for downwash algorithms (0.5 = ISC)
Beginning urban landuse type
Ending urban landuse type
Value
550.
0
5
0.01
0.1
0.5
10
19
Use Following Only For Single-Point Meteorological Input (CALPUFF Screen)
ILANDUIN
ZOIN
XLAIIN
ELEVIN
XLATIN
XLONIN
ANEMHT
ISIGMAV
IMIXCTDM
Land use type (20 = Unirrigated agricultural land)
Roughness length (m)
Leaf area index
Met. Station elevation (m above MSL)
Met. Station North latitude (degrees)
Met. Station West longitude (degrees)
Anemometer height of ISC meteorological data
(m)
Lateral turbulence (Not used with ISC
meteorology)
Mixing heights (Not used with ISC meteorology)
20
0.25
3
0
User Defined
User Defined
10.0
1
0
End of Single Point Meteorology Input Variables
XMXLEN
XSAMLEN
Maximum slug length in units of DGRIDKM
Maximum puff travel distance per sampling step
(units of DGRIDKM)
1
1
B-6
-------
Variable
MXNEW
MXSAM
SL2PF
PLXO
WSCAT
PGGO
SYMIN
SZMIN
SVMIN
SWMIN
CDIV
WSCALM
XMAXZI
XMINZI
PPC
NSPLIT
IRESPLIT
ZISPLIT
ROLDMAX
Description
Maximum number of puffs per hour
Maximum sampling steps per hour
Maximum Sy/puff length
Wind speed power-law exponents
Upper bounds 1st 5 wind speed classes (m/s)
Potential temperature gradients PG E and F
(deg/km)
Minimum lateral dispersion of new puff (m)
Minimum vertical dispersion of new puff (m)
Array of minimum lateral turbulence (m/s)
Array of minimum vertical turbulence (m/s)
Divergence criterion for dw/dz (1/s)
Minimum non-calm wind speed (m/s)
Maximum mixing height (m)
Minimum mixing height (m)
Plume path coefficients (only if MCTADJ = 3)
Number of puffs when puffs split
Hours when puff are eligible to split
Previous hour's mixing height (minimum), (m)
Previous Max mixing height/current mixing
height ratio, must be less then this value to
allow puff split
Value
99
99
10
0.07,0.07,0.10,0.
15,0.35,0.55
1.54,3.09,5.14,8.
23.10.8
0.020, 0.035
1.0
1.0
6*0.50
0.20, 0.12, 0.08,
0.06, 0.03, 0.016
0.01
0.5
3000
50
0.5,0.5,0.5,0.5,0.
35,0.35
3
User Defined
100
0.25
B-7
-------
Variable
EPSSLUG
PESAREA
NPT1
IPTU
NSPT1
NPT2
Point Sources
Area Sources
Line Sources
Volume
Sources
NREC
Receptor Data
Description
Convergence criterion for slug sampling
integration
Convergence criterion for area source
integration
Number of point sources
Units of emission rates (1 = g/s)
Number of point source-species combinations
Number of point sources with fully variable
emission rates
Point sources characteristics
Area sources characteristics
Buoyant lines source characteristics
Volume sources characteristics
Number of user defined receptors
Location and elevation (MSL) of receptors
Value
1.0E-04
1.0E-06
User Defined
1
0
0
User Defined
User Defined
User Defined
User Defined
User Defined
User Defined
B-8
-------
APPENDIX C
COMPACT DISK DATA RESOURCES
Solar and Meteorological Surface Observation Network (SAMSON), 1961 -1990
Version 1.0, September 1993
Available from U.S. Department of Commerce, National Climatic Data Center,
Federal Building, 37 Battery Park Ave., Asheville, NC 28801
Radiosonde Data of North America, 1946 -1992
Version 1.0, August 1993
Available from U.S. Department of Commerce, National Climatic Data Center,
Federal Building, 37 Battery Park Ave., Asheville, NC 28801
CALMET, CALPUFF, and CALPOST Modeling System
Version 1.0
Available from U.S. Department of Commerce, National Technical Information
Service, 5285 Port Royal Rd., Springfield, VA 22161. NTIS PB 96-502 083.
Hourly United States Weather Observations (HUSWO) 1990-1995
Available from U.S. Department of Commerce, National Climatic Data Center,
Federal Building, 37 Battery Park Ave., Asheville, NC 28801
MM4 -1990 Meteorological Data, Volumes 1-12
Available from U.S. Department of Commerce, National Climatic Data Center,
Federal Building, 37 Battery Park Ave., Asheville, NC 28801
NCDC Precipitation data CD's
Available from U.S. Department of Commerce, National Climatic Data Center,
Federal Building, 37 Battery Park Ave., Asheville, NC 28801
C-l
-------
APPENDIX D
SIXTH MODELING CONFERENCE
Section 320 of the 1990 CAA amendments requires the EPA to conduct public
conferences on air quality modeling at least every three years. These conferences are
to give special attention to appropriate modeling necessary for addressing prevention of
significant deterioration of air quality. The sixth of these modeling conferences was
held August 9-10, 1995 in Washington, D.C. One of the main topics discussed at the
sixth conference was a review of the status of work by IWAQM in the development of a
Phase 2 recommendation. At the conference comments were presented by several
attendees regarding their experiences in applying the Phase 1 interim
recommendations, and thoughts and recommendations were presented on future
needs. The presentation results, comments received and conclusions reached at the
Sixth Modeling Conference are fundamental to the development of the Phase 2
recommendations, and will be discussed in the following sections.
D.1 Presentation Summary
The IWAQM presentation at the Sixth Modeling Conference began with a review
of the work plan (U.S. EPA, 1992a) and the Phase 1 interim recommendations (U.S.
EPA, 1993), which have briefly been discussed above. The presentation then provided
a summary of work accomplishments in five areas: 1) results from a case study
conducted using the Phase 1 interim recommendations; 2) adaptations made to a
Lagrangian puff modeling system called CALPUFF; 3) MESOPUFF II and CALPUFF
simulation results for three of the 1983 Cross Appalachian Tracer Experiment
(CAPTEX) tracer releases; 4) conclusions reached regarding use of sophisticated
mesoscale meteorological analyses; 5) a proposed process for modeling specific Class
I areas. In the following five subsections, brief summaries are presented for each of
the topics listed. More extensive summaries are presented in Section 4.
D.1.1 MESOPUFF II Implementation Assessment
The objective was to learn by experience where the difficulties are in the process
of applying the MESOPUFF II air quality modeling system following IWAQM Phase 1
interim recommendations (U.S. EPA, 1993) and when possible, to provide a means for
resolving these difficulties. It was not an objective to provide a meaningful assessment
of PSD, NAAQS or AQRV impacts for the Class I areas considered in the study. As
part of this study the following tasks were carried out:
The MESOPUFF II model and associated processors were tested using the
example problem intended for Support Center for Regulatory Air Models
bulletin board (SCRAM BBS) distribution. The SCRAM BBS example
problem computer files were evaluated and some suggested improvements
were implemented.
D-l
-------
A five-year meteorological data set suitable for input to the MESOPUFF II
model was developed for a multi-state area surrounding Shenandoah
National Park (SNP), including the James River Face (JRF) wilderness area.
Model simulations were performed using three years of the developed five-
year meteorological data set to demonstrate the assessment of visibility,
acidic deposition, and PSD increments for a set of real sources in the states
surrounding Shenandoah National Park.
Model simulations were performed to test the sensitivity of concentrations to
the distance between sources and receptors using a set of "pseudo" sources
placed in successive rings around Shenandoah National Park.
Several conclusions were reached as a result of the demonstration assessment.
First, tailoring a puff simulation analysis to assess PSD and AQRV impacts at SNP and
JRF was a difficult task. A primary goal is to develop a reasonably good
characterization of the spatially-varying, time-varying three dimensional wind field. This
goal alone requires an air pollution modeler with strong dispersion meteorology
experience, having expert judgement and finesse. The differing goals between PSD
permitting and AQRV assessment required frequent consultation with FLMs. Since the
modeling necessarily required case-specific judgement decisions, strong collaboration
and review was required by the various regulatory reviewing authorities. All these
considerations have caused IWAQM to believe in and recommend a Regional
Approach, as discussed in Section 4.5.
Even with anticipated improvements in the software, it is likely that the analyses
will require programming special routines to convert data into appropriate formats, or to
assist in the analysis and summarization of the data. Also even though the modeling
software can be executed on personal computers, it is likely that a workstation would
prove more convenient and useful. Thus it is concluded that mesoscale analyses using
Lagrangian puff dispersion models must be viewed as more involved and difficult than
using conventional plume dispersion models. This is especially true in consideration
that most plume dispersion modeling assessments use a single-station's hourly
meteorological observation, whereas puff dispersion modeling is founded on using a
three-dimensional wind field, that is consistent with the terrain and land-use.
The PSD and AQRV impacts developed in the demonstration assessment from
the ten aggregated sources exhibited strong year-to-year variations. Furthermore,
some of the impacts were close to the PSD increments. These results suggest that at
least three years of modeling is needed in order to assess the likely maximum impacts.
Part of the reason for recommending a Regional Approach is to address concerns that
more attention needs to be given to comparing PSD and AQRV impacts generated by
all the sources as a the cumulative impact assessment.
The ring source analysis illustrates some of the effects of source-receptor
distance on air quality and deposition impacts. For the primary species, S02 and NOX,
D-2
-------
peak impacts declined rapidly with distance. MESOPUFF II results suggest that
sources of the size used for this analysis (183 g/s S02), located 50 km from SNP, are
capable of producing 3-hour S02 impacts close to the allowable PSD Class I increment.
For the secondary species, S04= and N03", impacts did not show a decreasing trend for
sources between 50 km and 200 km from SNP. Although modeled PMio
concentrations were well below allowable PSD Class I increments for all rings, the lack
of a clear trend suggests that sources beyond 200 km may need to be considered in
some cases in order to assess the impact upon PMio and related parameters, such as
visibility.
D.1.2 Revisions to CALMET and CALPUFF
In the course of completing the Phase 1 recommendations, IWAQM became
aware of the CALMET/CALPUFF modeling system (Scire et al., 1990ab), which was
actively under development. Building from lessons learned from the
MESOPAC/MESOPUFF II modeling system, IWAQM felt further enhancements were
needed in two areas. With a view towards allowing one model to be used for all
sources (which might include source-receptor distances of less than 50 km), the first
area for enhancement was to include within CALPUFF dispersion additional algorithms,
so that CALPUFF simulation results would be consistent with the Industrial Source
Complex (ISC) model and the Complex Terrain Dispersion Model (CTDMPLUS) (Perry
et al., 1989) modeling results for steady-state meteorological conditions. Both ISC and
CTDMPLUS are recommended in the Guidelineior use for source-receptor distances of
less than 50 km. The second area for enhancement to the CALMET/CALPUFF
modeling system was to include provisions within CALMET to allow use of mesoscale
meteorological modeling results created using data assimilation techniques, for
example Stauffer and Seaman (1989) and Stauffer et al. (1990). The characterization
of the time-varying three-dimensional wind field is one of the most challenging issues
for long-range transport.
Consistency with local-scale plume dispersion models.
At the Sixth Modeling Conference, IWAQM reported on the inclusion within
CALPUFF of dispersion algorithms to provide results consistent with ISC and
CTDMPLUS. At the time of the conference, there were only preliminary sensitivity
testing results to show that the modifications to CALPUFF would be successful in
mimicking ISC. There were no comparison results available showing consistency
between CALPUFF and CTDMPLUS. It was recognized and mentioned by IWAQM at
the Sixth Modeling Conference that more testing was needed to test whether the code
modifications implemented in CALPUFF would replicate dispersion results as would be
simulated by ISC and CTDMPLUS.
D-3
-------
Enhancement of CALMET for mesoscale applications.
The wind field module in CALMET is based on the Diagnostic Wind Model
(DWM). In anticipation of using CALMET and CALPUFF for long-range transport
distances, a series of modifications were made (U.S. EPA, 1995b). The DWM in
CALMET uses a two step procedure in developing the final wind fields. An initial guess
field is developed based on a domain-average wind profile, and this domain-average
profile of winds is adjusted for terrain effects and divergence minimization to produce a
"Step 1" wind field. The second step in the processing of the wind field is the
introduction of the observational data into the terrain adjusted Step 1 wind field.
The adaptations needed to allow use of meteorological wind fields as analyzed
by sophisticated mesoscale meteorological models (hereafter referred to as FDDA-MM
data) involved more than simply providing a new data input option. (For further
discussion of Four Dimensional Data Assimilation (FDDA) and how such data sets are
constructed, see Section 4.4.) IWAQM was confident that since the surface and upper-
air observations are included as part of the data assimilation process, the mesoscale
meteorological analyses from such analyses could be treated as data. But IWAQM was
aware that inherently such data are representative of a certain scale, as defined by the
physics included in the numerical equations and by the grid scale of the results.
Therefore, CALMET was modified to allow introduction of the FDDA-MM data 1) as
input in the creation of the CALMET Step 1 wind fields; 2) as the CALMET Step 1 wind
fields; or 3) as input in the creation of the CALMET Step 2 wind fields.
At the Sixth Conference, results from a series of sensitivity analyses were
reported investigating how to combine the FDDA-MM data with observations, and
investigating the impact of FDDA-MM data on simulated trajectory results. The
sensitivity analyses were conducted for two episodes in eastern United States, one
summer episode (August 1-6, 1988) and one winter episode (December 3-10, 1988).
The summer episode was characterized by light wind, stagnant conditions. The winter
episode was characterized as an active period that included the passage of a front and
low-pressure system through the domain. Penn State Mesoscale Meteorological
(version 4) modeling results (MM4) were available employing four dimensional data
assimilation (Stauffer and Seaman, 1989) for both episodes and for three different grid
resolutions, 18-, 54- and 80-km. The CALMET model was run using hourly weather
observations from 119 surface stations for both episodes, and used twice-daily
observations from 25 upper-air stations for the summer episode and 23 upper-air
stations for the winter episode. In general, the introduction of the 80-km MM4 winds
improved the ability of CALMET to reproduce the reference 54- and 18-km MM4 wind
fields. Slightly better agreement was achieved when the 80-km MM4 winds were
brought in after the diagnostic terrain adjustment procedures (i.e., as the Step 1 wind
fields or as "observations"). This was conjectured to occur due to the fact that 1) the
80-km MM4 are already close to the 54- and 18-km MM4 results, and 2) the CALMET
diagnostic adjustments may duplicate terrain effects that were already accounted for in
the development of the 80-km MM4 winds. From the numerical sensitivity tests,
D-4
-------
IWAQM concluded that using FDD/A-MMwind fields adds a noticeable and significant
improvement in the characterization of trajectories of the dispersing material within the
atmosphere.
D.1.3 Trajectory Comparisons
As discussed in the previous section, modifications were made to CALMET in
anticipation of using CALMET and CALPUFF for long-range transport distances (U.S.
EPA, 1995b). It was anticipated that use of FDDA-MM winds would improve
CALPUFF's characterization of trajectories of dispersing pollutants. To further
investigate this, two trajectory studies were conducted: 1) comparisons with simulated
trajectories, and 2) comparisons with trajectories derived from surface tracer
monitoring.
CALMET trajectory comparisons.
Trajectories were computed from four release locations at three levels (10m,
200 m, and 400 m) for each of the wind fields discussed in Section 4.1.2 for the
summer episode (U.S. EPA, 1995b). Trajectories were generated at each location
every 4, 6 and 12 hours from the beginning of the simulation, for up to 24 hours before
the end of the simulation. A statistical analysis was conducted on the trajectories to
assess the effect of the different wind fields.
Trajectory statistics were computed from each release time and for all three
levels. In general, the introduction of the 80 km MM4 winds into CALMET to develop
either 54 km or 18 km gridded wind fields significantly improved the comparisons with
the trajectories developed from the 54 km and 18 km MM4 wind fields directly, versus
using only the routine hourly weather observations and twice-daily upper air
observations as input to CALMET.
CAPTEX comparisons
One of the primary objectives of this study was to assess whether use of
mesoscale dynamic wind fields developed using FDDA-MM data exhibited improved
spatial and temporal resolution versus typical mesoscale wind fields determined
diagnostically from the available hourly surface and twice-daily upper air observations,
would improve the quality of the characterization of the transport and dispersion.
Results were presented for CAPTEX releases 3, 5 and 7 (Irwin et al., 1996). The
Cross-APpalachian Tracer Experiment (CAPTEX) is a unique series of tracer releases
which, besides testing a particular tracer technology, was conducted for the purpose of
providing insight into the mechanisms involved in long-range transport and dispersion
(Ferber et al., 1986). A three-hour ground-level release of perfluoromonomethyl-
cyclohexan (C7H14, PMCH) was made five times near Dayton, Ohio and twice from near
Sudbury, Ontario when winds were expected to transport the tracer over the ground-
level sampling network. Samplers were operated at 86 sites in Ohio, Pennsylvania,
D-5
-------
New Jersey, New York, New England and southern Canada at distances from 300 to
1100 km from the release site. Air concentrations were collected for 3 and 6 hour
durations for several days following each release.
Meteorological data available for use in developing the CAPTEX wind fields
consisted of 122 National Weather Service (NWS) surface locations reporting hourly
and 13 upper-air locations reporting twice-daily (0000 GMT and 1200 GMT) throughout
the region. Furthermore, mesoscale wind fields developed using Version 8 of the Penn
State/NCAR Mesoscale Model - Generation 4 (MM4) were available on an 80-km grid.
Three wind field models were used to obtain a gridded field of meteorological data with
a horizontal resolution of 18-km: MESOPAC II, CALMET, and CALMET using the
mesoscale wind fields as STEP-1 inputs. The latter modeled wind fields were
developed using FDDA and are referred to as CALMET/MM4.
The comparison results presented were conducted with simplified puff dispersion
model assumptions, hence the model-to-model differences were minimized. It was
concluded there was a noticeable improvement in the simulation of the puff centroid
trajectories, when the wind fields were developed using FDDA-MM data, versus
developing the wind fields diagnostically from the available hourly surface and twice-
daily upper air observations. The analysis of the concentration maxima and lateral
dispersion values suggest that the simulation assumptions employed in these results
consistently underestimate the horizontal extent of the tracer puff as it is transported
downwind. The centroid maximum surface concentration was found to be
correspondingly overestimated and relatively insensitive to the mesoscale wind
characterization. In these simulations, no provisions were made to address delayed
shear enhancement of the dispersion as described by Moran and Pielke (1994) and Shi
et al. (1990). Inclusion of some sort of puff splitting is obviously warranted, but the
computational demands are not trivial if one is attempting to develop an operational
model for routine use. Furthermore, in those cases where the puff model dynamics
have been enhanced, for example Draxler (1987) and Davis et al. (1986), there was a
tendency to underestimate the surface maximum concentrations.
D.1.4 Constructing FDDA-MM Data Sets Assessment
To foster use of mesoscale meteorological (MM) data processed using Four
Dimensional Data Assimilation (FDDA) in routine air pollution modeling assessments
and to learn what problems might be associated with such a project, a one-year
meteorological data set for 1990 was fabricated that spans the contiguous United
States, southern Canada and northern Mexico. The 1990 data set (NCDC, 1995) was
somewhat dated when it was discussed at the Sixth Modeling Conference. Yet it still
represents a significant advancement to meteorological characterizations employed in
many current routine air pollution assessments. These data were made available to
allow investigators to have access to this type of data that they might otherwise not
have access to (NCDC, 1995), with the hope that this would stimulate investigations,
exploring the strengths and weaknesses of these data in a variety of ways.
D-6
-------
The 1990 data set consist of hourly profiles of wind, temperature and moisture at
23 levels in the atmosphere on an 80-km grid spacing. The Penn State mesoscale
meteorological model (MM4) with FDDA was used in developing these data. The
horizontal grid spacing of the MM4 simulation was 80 km in both dimensions, with a grid
array size of 85 by 56 centered on 90 W longitude and 40 N latitude to cover most of
the North American continent and adjacent oceanic areas (Bullock, 1993). The domain
of the model's vertical coordinate system extended from the earth's surface to the 100
millibar pressure level (approximately 15 km above sea level). It's 15-level structure
provides height-resolved information similar to that routinely obtained by the National
Weather Service at 12-hour intervals from approximately 80 rawinsonde balloon
sounding locations across North America. However, the data set obtained from the
MM4 simulation provides synthetic soundings at 1-hour intervals for 4080 model
grid-point locations, or about 600 times more information than is available from routine
observational networks. The 1990 data set is comprized of over 20 billion bytes of
information.
One of the lessons learned in developing the 1990 data set was that the science
of mesoscale analysis using data assimilation is rapidly developing. Major
advancements have been occurring every several months during the period from 1995
through 1997. There are various research groups with active development programs
investigating mesoscale meteorological modeling employing data assimilation, e.g.
Pielke et al., (1997). This suggests that developing and maintaining a multi-year data
set would likely involve a substantial committment in resources. An alternative to
commissioning special one-year runs would be to use results from an existing
operational ongoing activity (Pielke and Uliasz, 1997). One such activity is a product
that currently is under development and refinement by the NOAA Environmental
Modeling Center Mesoscale Modeling Branch. They are employing a mesoscale
numerical weather prediction model, known as the NCEP ETA Model, with data
assimilation. This mesoscale meteorological model produces analyses of the vertical
profiles of wind, temperature, pressure and moisture on a 48 km grid resolution that
covers a very large domain (e.g., as far west as the Hawaiian Islands, all of the
contiguous United States, as far north as all of Alaska). The current operational output
provides information for 38 layers in the vertical. The model employs a surface
moisture balance model (Chen et al., 1997), and hence is capable of reporting the
surface heat and moisture fluxes, soil moisture, and precipitation (amounts for both wet,
frozen and snow).
The Environmental Modeling Center (EMC) Home Page is: http://nic.fb4.noaa.
gov:8000/. The ETA model outputs are produced by the Mesoscale Modeling Branch
(MMB). The Mesoscale Modeling Branch Home Page is: http://nic.fb4.noaa.gov:8000
/research/mesoscale2.html.
A major obstacle is access to these data. The 1990 MM4 data set in a
compressed format (providing only profiles of wind, temperature and moisture) requires
12 Compact Diskettes. An operational means for gaining easy access to
D-7
-------
comprehensive mesoscale meteorological data sets, as alluded to here, has yet to be
developed. But IWAQM believes that the future of air quality modeling requires a
solution being found to gain routine inexpensive access to such data. And IWAQM
believes that it is only a matter of time (perhaps less than three years) that such a
solution could be found.
D.1.5 Regional Approach
The effective treatment of mesoscale transport of pollutants requires treatment of
the time and space variations of the three-dimensional meteorological conditions. This
in turn involves consideration of the variations in terrain heights and land-use, and their
consequent effects on the dispersive characteristics of the atmosphere. A major issue
in assessing mesoscale impacts to specific Class I areas from specific sources is giving
due consideration to the proper characterization of the transport (trajectory) of the
pollutants. As alluded to in the discussion of Section 4.1, where MESOPUFF II was
used to assess impacts on the Shenandoah National Park, and in Section 4.3, where
the characterization of the transport trajectories was investigated, this is not a process
that lends itself to a cookbook approach. Tailoring the wind field analysis will require
technical judgement and discretion, and IWAQM has come to the opinion that these
case-specific decisions are best achieved using a team of experts and development of
a consensus.
The discretion and case-specific decision making does not apply soley to the
implementation of the dispersion modeling. Not all wilderness areas have the same
flora and fauna. The AQRVs of interest will be specific to each wilderness area.
Developing an inventory with agreed upon emission rates is not trivial. The inventory
could differ depending on whether the analysis is addressing NAAQS assessments
(which typically address maximum allowable emission rates from PSD sources) versus
AQRV assessments (which typically address actual current emission rates from all
sources).
Therefore at the Sixth Modeling Conference, IWAQM suggested that Class I air
quality modeling assessments be designed for each Class I area (or cluster of Class I
areas), rather than for each permit as is the case for the Class II program. The
cornerstone of this approach is an up-front comprehensive increment and AQRV
analysis of the area. This "initialization" study would be accomplished outside the
context of a permit application, and could involve technical experts from private and
public groups. Perhaps most importantly, it provides future applicants with up-front
information needed for planning and assurance of what is expected for the given
situation.
D.1.6 Sixth Modeling Conference IWAQM Recommendations
At the Sixth Modeling Conference, IWAQM recommended that the CALPUFF
modeling system (Scire et al., 1995a,b) replace the MESOPUFF II modeling system.
D-8
-------
Based on sensitivity studies and the CAPTEX comparisons of results obtained by the
two modeling systems, IWAQM had concluded that the CALPUFF modeling system
was at least as good as the MESOPUFF II modeling system, and offered several
improvements. For one, the CALPUFF modeling system code was better documented
and was believed to be better engineered. Graphical User Interface (GUI) applications
were under development to assist and make easier the use of the CALPUFF modeling
system. CALMET had been enhanced to allow use of FDDA-MM mesoscale
meteorology. CALPUFF had been enhanced to allow better correspondence with local-
scale (less than 50 km transport) models recommended for use in the Guideline.
These latter enhancements provided the basis for allowing one model to be used for all
sources in a mesoscale assessment, even though for some of the sources the transport
might be less than 50 km. At the time of the Sixth Modeling Conference, the CALPUFF
modeling system was not available from the EPA Support Center for Regulatory Models
(SCRAM) electronic Bulletin Board Service (BBS), but it was anticipated that it would
soon be available in an area designated "Topics for Review and Comment."
Even though such data were not available, nor was it likely that IWAQM would
be able to develop such data, IWAQM recommended that use of FDDA-MM mesoscale
meteorological data be approved in regulatory assessments. The comparisons
between alternative methods for developing CALMET wind fields, and the comparisons
of the resulting trajectories, suggested to IWAQM that the added skill in the assessment
was dramatic and desirable. There were (and still remain) serious logistical issues to
development and use of such data. But in each case where such data have been used,
the characterization of the trajectories has improved.
Finally, given the myriad of decisions needing to be made in most mesoscale
assessments of transport and dispersion impacting Class I areas, IWAQM
recommended that a Regional Approach be used to resolve the various decisions. This
involves assembling a committee of the various public and private stakeholders, and
gaining consensus that the committee will be proactive in developing a plan to chart a
course through the specific decisions needing to be made for the Class I area they are
concerned with, and following the developed plan. For those seeking permits, being
able to go to such a committee for information is of great benefit. It insures that all
applicants are treated equitably, and it allows packaging of the results from each
applicant in a manner that supports cumulative impact tracking and assessment. If the
committee sponsors development of meteorological data sets, these can be made
available to the applicants to save costs and time.
D.2 Summary Of Sixth Modeling Conference Comments
One commenter pointed out the practical deficiencies with straight line plume
transport models and strongly urged the development and adoption of a trajectory
model. Another commenter pointed out the technical superiority of CALPUFF over the
best Gaussian models EPA supports. The commenter urged that the advantages and
applicability of CALPUFF be discussed among state and regional modeling contacts,
D-9
-------
that evaluation and comparison data be amassed, and that early consideration be given
to case-by-case regulatory applications of CALPUFF.
Another commenter noted that "[The IWAQM] apparently did not address the
question of whether or not [an air quality analysis beyond 50 km] can be done with the
hope of reaching any reasonable answer." The commenter argued that there is no
accurate way to do this. The commenter also noted that IWAQM did not address the
accuracy of the Phase I screening analysis (Level I) or refined analysis (Level II). Also,
the results were not qualified (i.e., level of conservatism was not stipulated). It was
recommended that IWAQM qualify the accuracy of the results. For Level I analyses, it
was recommended that IWAQM remove the stagnation and recirculation wording and
that IWAQM specify with "reasonable particularity" the Guideline models to be used,
that some reasonable yet conservative and consistent conversion rates for both visibility
and deposition be assumed, and that the Level I analysis be a true screen. For Level II
analyses, the commenter recommended that, if a preapproved protocol is required, the
regulatory and FLM agencies must agree to abide by the results and that on-site
meteorological data not be required. The commenter believes that in some cases the
regulatory need to analyze long-range impacts is beyond the state-of-the-practice, and
that IWAQM failed to portray how conservative such calculations really may be.
Several figures were included to illustrate the commenter's contention that CALPUFF
would have severe difficulty simulating accurate concentrations on the back side of
terrain along certain transects near the Rockies.
Aside from the more specific comments mentioned above, the commenters all
seemed in agreement that the Regional Approach outlined by IWAQM was desirous.
There seemed general agreement that the Level I screen suggested in the Phase I
interim recommendations was not working well and needed to be improved. There was
also general agreement that the CAPTEX comparisons were for transport distances of
300 to 1000 km, well beyond the anticipated range of PSD Class I impact assessments.
Therefore, more comparisons with tracer data studies for transport distances of order
50 to 200 km were needed. Finally, although only mentioned by the last commenter,
there was a sense that mesoscale transport assessments may be very difficult
analyses, involving discretion and expert judgment. In some instances, the modeling
results would be so uncertain as to preclude them of being any use.
D.3 Response to Sixth Modeling Conference Comments
The comments received at the Sixth Modeling Conference can be summarized
into several general areas: 1) there seemed to be agreement that the Regional
Approach outlined by IWAQM was desirous; 2) there was general agreement that as
the CAPTEX comparisons were for transport distances of 300 to 1000 km, well beyond
the anticipated range of PSD Class I impact assessments, that more comparisons with
tracer data studies for transport distances of order 50 to 200 km were needed; 3) there
was general agreement that the Level I screen suggested in the Phase I interim
recommendations was not working well and needed to be improved; 4) there was
D-10
-------
agreement that comparisons were needed to assess whether CALPUFF can provide
results similar to ISC and CTDMPLUS for steady-state meteorological conditions; and
5) finally there was a sense that mesoscale transport assessments may be difficult
analyses (perhaps impossible in some cases), requiring discretion and expert judgment.
Regional Approach
It has been IWAQM's experience that most long-range transport dispersion
assessments involve expert judgement and finesse in order to provide a reasonable
characterization of the transport and dispersion with the information and resources
available. The process has yet to be successfully accomplished using "cookbook"
procedures. To further complicate matters, the AQRV's of interest are typically specific
to the Class I area and climate, and thus the modeling endpoints need to be
customized to address the site-specific issues of concern. It is therefore
understandable that most people concur with the recommendation by IWAQM to
employ committees of technical experts, whenever possible, to sort through the various
decisions and tradeoffs. It also understandable that most people would endorse having
a panel of experts dedicated to a particular Class I area (or areas) that could provide
some consistency in the modeling assessments and assist in addressing site-specific
questions as they arise. But just as some of the options in an air quality modeling
system can not be automatically prescribed, the same can be said for use of technical
committees. Hence, while IWAQM endorses use of such committees, IWAQM does
not recommend Federal agencies mandate or require their use.
Model evaluations for the 50 to 200 km range
There are very few tracer dispersion field studies with sufficient sampling to
depict with some certainty the relative location of the receptors to the puff of dispersing
tracer as it was transported downwind. In fact, the number of studies is so small that
comparison results that can be developed only provide anecdotal evidence regarding
model performance. In response to the suggestion that comparison results be
developed for transport of order 50 to 200 km, IWAQM commissioned studies for four
tracer dispersion experiments. A summary of the findings for each of the four tracer
studies is provided in Section 4.6. Three of these field studies involved 3 to 4 hour
releases that were then sampled along arcs of receptors downwind. In some cases,
shorter-term concentration values were available, such that the transport of the puff
past an arc could be seen. Differences on the order of 10 to 20 degrees were seen
between the simulated and observed center of mass of the puff as it passed the
receptor arc. Most of the simulated centerline concentration maxima along each arc
were within a factor of two of those observed. In those instances when large over- or
underpredictions occurred, there was insufficient information available to ascertain the
physical processes that resulted in the observed concentration values. Without a better
understanding of the physical dispersion processes affecting these instances, no
simulation has (or likely will) show skill.
D-ll
-------
It was concluded from these admittedly anecdotal case studies, that CALPUFF is
performing in a reasonable manner, and has no apparent bias towards over- or under-
prediction. This is in contrast to the CALPUFF comparison results with the CAPTEX
tracer field data, where CALPUFF consistently underestimated the 'footprint' of the puff
in contact with the surface, and correspondingly overestimated the puff concentration
maxima by a factor of 2 or more. This suggests CALPUFF (as configured for the
CAPTEX comparisons) was poorly characterizing dispersion processes (such as
delayed shear enhancement) which become important for transport distances of order
300 km or more. It is concluded that CALPUFF will need to be further developed
before it can more realistically characterize transport and dispersion for more than one
diurnal cycle.
The IWAQM concludes that CALPUFF can be recommended as providing
unbiased estimates of concentration impacts for transport distances of order 200 km or
less, and for transport times of order 12 hours or less. For larger transport times and
distances, our experience thus far is that CALPUFF tends to underestimate the
horizontal extent of the dispersion and hence tends to overestimate the surface-level
concentration maxima. This does not preclude the use of CALPUFF for transport
beyond 300 km, but it does suggest that results in such instances be used cautiously
and with some understanding.
Comparisons with ISC and CTDMPLUS
At the time of the Sixth Modeling Conference, the algorithms installed in
CALPUFF to allow it to mimic ISC and CTDMPLUS had not been thoroughly tested. In
response to the need for further comparisons, a thorough investigation and comparison
with ISC was completed. A similar investigation is needed but has yet to be
accomplished with CTDMPLUS. A summary of the comparison results with ISC is
presented in Section 4.7.
As expected, when the meteorological conditions were steady-state (fixed wind
direction, wind speed, mixing height, and stability), there were only minor differences
seen between ISC and CALPUFF. For all of the point source and volume source
comparisons, the differences seen between ISC and CALPUFF were less than one
percent. The largest differences were seen for receptors within and near the downwind
edge of a simulated area source. For the area source comparisons, the differences on
average were less than four percent, but individual cases were seen with larger
differences. These larger differences seen with the area source are traceable to the
difficulty of efficiently replicating the plume model's characterization of an area source.
The IWAQM concluded that, for all practical purposes, although improvements might be
made, the algorithms within CALPUFF are capable of replicating the results of ISC for
steady-state meteorological conditions.
When the meteorology was allowed to vary from one hour to the next, but not in
space, large differences were seen in the simulated highest and second-highest
D-12
-------
concentrations for all averaging times examined (1-hour, 3-hour, 24-hour, and annual)
and for all distances downwind out to 300 km. In general, CALPUFF was seen to
provide higher surface-level concentration impacts than ISC in those special situations
where the preceding hours had several hours of calm winds, or involved a dramatic
wind reversal and caused previously released material to be simulated to combine with
newly released material. Given that ISC ignores impacts during calm wind conditions,
and can not characterize the 'causality' resulting from a buildup of material during an
extended period of calm winds or a reversal in the transport winds, such differences
were expected.
The IWAQM concludes that CALPUFF does reproduce, for all practical
purposes, the results that would be obtained using the ISC plume dispersion model,
when the meteorological conditions are steady-state. For situations involving complex
winds conditions (frequent periods of extended calms, routine and periodic wind
reversals, complex topographical wind effects and channeling, etc.), where a local-scale
plume dispersion model (such as ISC) is the recommended modeling approach,
IWAQM recommends acceptance on a case-by-case basis the results obtained using
the CALPUFF modeling system. In these situations, the three-dimensional time-varying
wind field and the 'causality' from one hour to the next are most important in deriving an
assessment of the impacts associated with the transport and dispersion of pollutants.
In conducting the comparisons of CALPUFF with ISC, minor inconsistencies
were found and corrections were made to the CALPUFF algorithms, so that the
CALPUFF results would replicate (to the extent a puff model can) the results that would
be obtained by a plume model for steady-state meteorological conditions. A similar
intensive inspection of the CTDMPLUS algorithms within CALPUFF has yet to be
accomplished. Once such an investigation has been completed (assuming that
CALPUFF can be made to provide results acceptably similar to those obtained with
CTDMPLUS for steady-state meteorological conditions), then IWAQM would endorse
use of results from the CALPUFF modeling system on a case-by-case basis, in lieu of
those obtained from the CTDMPLUS model, for complex wind situations.
Develop a better screen technique
At the Sixth Modeling Conference, the 'Level I' technique suggested in IWAQM
Phase I interim recommendations (EPA, 1993), of using as a screening model results
for the ISC dispersion model, was generally considered inadequate. The Level I
screening estimates of the AQRV impacts were considered so conservative that only
sources with very small emissions were 'screened' from further consideration. In
response to these comments, a study was conducted in an attempt to construct a more
meaningful and useful screening technique, a summary of which is presented in
Section 4.8.
By far the most demanding task to successfully accomplish is to develop a valid
time and space varying characterization of the meteorological conditions for use by the
D-13
-------
CALPUFF puff dispersion model. Thus IWAQM concluded that if a simplification could
be made in specifying the meteorological conditions, perhaps CALPUFF could be used
to develop the screening estimates of pollutant impacts. This would allow the chemistry
and fate to be appropriately characterized, and at least some of the 'causality' effects
would be captured. It was with this background and understanding that IWAQM
investigated developing a screening estimate using CALPUFF with a highly simplified
characterization of the meteorological conditions.
In the course of the development of the CALPUFF modeling system, options
were implemented that allow the meteorological conditions to be specified using either
the meteorological input required for the ISC model, or the meteorological input
required for the CTDMPLUS model. These optional formats for specifying the
meteorology to CALPUFF were implemented initially to facilitate testing of ISC and
CTDMPLUS algorithms in CALPUFF. The IWAQM decided to pursue development of a
screening estimate that could be made using the meteorological input for the ISC
model, as input to CALPUFF.
One of the consequences of simplifying the meteorological input to that of the
ISC model, is all terrain effects (channeling, slope flows, etc.) are lost. The
consequence of these effects can become important in assessing pollutant impacts for
specific source-receptor combinations. To counter for the loss of properly
characterizing the spatial variations in the meteorological conditions, it was decided
that rings of receptors would be used that completely surround the source under
analysis. The rings would be placed to pass through the Class I area(s) of concern,
and would be spaced to provide suitable coverage, using expert judgement and in
agreement with applicable reviewing authorities. The maximum pollutant impacts
(increment, AQRV, etc.) found anywhere on any of the rings of receptors would be used
as the screening impact estimate of the pollutant impact. Selecting only receptors
within the particular Class I area(s) is not appropriate, as the screening analysis as
designed does not address the spatial variations in the meteorological conditions.
As summarized in Section 4.7, use of ISC meteorological input does not
guarantee that the maximum S02 concentration found on a ring of receptors is
consistently estimated to be larger than what would be found if a complete three-
dimensional time-varying wind field and meteorological conditions were used, all other
factors being equal. In the sensitivity tests conducted, the longer the averaging time of
the S02 concentration, the more likely the maximum for a ring of receptors will be
similar, whether ISC or CALMET meteorology input is used. The CALPUFF screening
estimates of sulfate concentrations and deposition were consistently much less than
would be estimated using IWAQM Phase 1 Level 1 ISC procedures. The CALPUFF
screening estimates of sulfate concentrations and deposition were sometimes higher
and sometimes lower than that seen using full CALMET input to CALPUFF.
It is concluded by IWAQM that using five years of ISC meteorology as input to
CALPUFF, and selecting the maximum concentration from rings of receptors that
D-14
-------
completely surround the source under analysis, provides a reasonable basis for
developing screening estimates. There is a finite possibility that using fully developed
CALMET input to CALPUFF, somewhat higher maxima might be simulated, but whether
these would be at receptors of interest in the Class I area(s) is considered less likely. In
fact, it is considered highly likely that lower pollutant impacts would be simulated if use
was made of fully developed CALMET input to CALPUFF, and only impacts to
receptors in the Class I area(s) were considered. The IWAQM recommends use of the
new screening technique, as it provides a more realistic assessment of the fate and
transport of the pollutants than the Phase 1 screening (Level 1) technique. The new
screening technique provides estimates of maximum impacts, that are similar to those
derived using fully developed CALMET input. Although, the air dispersion modeling
community has less experience using the CALPUFF puff dispersion model than the ISC
plume dispersion model, the operation of CALPUFF (with ISC meteorological input) is
not anticipated to be unduly difficult or onerous.
Realistic expectations
At the Sixth Modeling Conference, some of the concerns expressed were that
estimating pollutant impacts is so uncertain in some circumstances to perhaps preclude
meaningful use of the estimated impacts. An example that was used to illustrate this
point was described where the source emissions being characterized were separated
from the receptors of interest by two mountain ridges, whose tops exceeded the
effective release height of the example source. The situation described was neither
impossible nor unrealistic. The IWAQM agrees with the commenter's conclusion that in
such a situation highly uncertain pollutant impacts would be generated, regardless of
the air quality model employed. Whether such estimates would be of little practical use
would depend on the exact nature of the assessment being attempted. Even though
the estimates might be uncertain and depend on the physics incorporated in the air
dispersion model simulations, inspection of the simulation results might suggest that
impacts from the source in question could reasonably be argued to be inconsequential
or highly unlikely. As in any air quality simulation, the usefulness of the results obtained
depends mostly on the expertise brought to the analysis in characterizing the situation,
and on the experience applied in interpreting the results obtained. The IWAQM agrees
with the sentiments of the commenters at the Sixth Modeling Conference, that as the
terrain and land-use induced mesoscale circulations become more dominant, the
expertise and integrity of the modeler will define the usefulness of the results.
It would be convenient if objective criteria could be developed that would identify
when air quality simulation results are of questionable integrity. It would be convenient
if objective criteria and "cookbook" procedures could be constructed that would
preclude inappropriate application of air dispersion models. This has proved to be
troublesome for local-scale modeling, and even more problematic for mesoscale and
long-range transport modeling.
D-15
-------
In response to the comments received, IWAQM has attempted to warn the
modeling community in the summary of its Phase 2 recommendations (Section 2), that
conducting a long-range transport assessment requires competent individuals. The
IWAQM have also tried to warn the modeling community that application of the
CALPUFF modeling system to any situation will require expert judgment, it will likely
involve site-specific decisions, and it will require strong interaction and coordination with
the applicable reviewing authorities.
D-16
-------
TECHNICAL REPORT DATA
(Please read Instructions on reverse before completing)
1. REPORT NO.
EPA-454/R-98-019
3. RECIPIENT'S ACCESSION NO.
4. TITLE AND SUBTITLE
5. REPORT DATE
December 1998
Interagency Workgroup on Air Quality Modeling
(IWAQM) Phase 2 Summary Report and
Recommendations for Modeling Long Range Transport
Impacts
6. PERFORMING ORGANIZATION CODE
7. AUTHOR(S)
John S. Irwin
8. PERFORMING ORGANIZATION REPORT NO.
9. PERFORMING ORGANIZATION NAME AND ADDRESS
Office of Air Quality Planning and Standards
Emissions, Monitoring, and Analysis Division
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
10. PROGRAM ELEMENT NO.
11. CONTRACT/GRANT NO.
12. SPONSORING AGENCY NAME AND ADDRESS
13. TYPE OF REPORT AND PERIOD COVERED
14. SPONSORING AGENCY CODE
15. SUPPLEMENTARY NOTES
16. ABSTRACT
This Phase 2 Report is published by the Interagency Workgroup on Air Quality Modeling
(IWAQM) in an effort to provide the sponsoring agencies and other interested parties
information on appropriate "off-the-shelf methods for estimating long range transport
impacts of air pollutants on Federal Class I areas.
17.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
b. IDENTIFIERS/OPEN ENDED TERMS
c. COSATT Field/Group
Air Pollution
Meteorological Data
Air Dispersion Models
New Source Review
Air Pollution Control
18. DISTRIBUTION STATEMENT
Release Unlimited
19. SECURITY CLASS (Report)
Unclassified
21. NO. OF PAGES
151
20. SECURITY CLASS (Page)
Unclassified
22. PRICE
EPA Form 2220-1 (Rev. 4-77)
PREVIOUS EDITION IS OBSOLETE
------- |