&EPA
Unitoda**
EnviroimrU PiutecBmi
Agency
Assessment of the "VISTAS" Version of
the CALPUFF Modeling System
-------
-------
EPA-454/R-08-007
August 2008
Assessment of the "VISTAS" Version of the
CALPUFF Modeling System
U. S. Environmental Protection Agency
Office of Air Quality Planning and Standards
Air Quality Assessment Division
Air Quality Modeling Group
Research Triangle Park, North Carolina
-------
PREFACE
This report summarizes the process undertaken and documents the results of EPA's assessment
of the "VISTAS" version of the CALPUFF modeling system, which lead to EPA's approval of
CALPUFF (v5.8), CALMET (v5.8) and CALPOST (v5.6394) as the "EPA-approved" version,
announced on June 29, 2007.
11
-------
ACKNOWLEDGMENTS
The assessment of the "VISTAS" version of the CALPUFF modeling system required the
involvement of several members of the Air Quality Modeling Group in the Air Quality
Assessment Division at EPA's Office of Air Quality Planning and Standards, the developers of
the CALPUFF modeling system, and the Visibility Improvement State and Tribal Association of
the Southeast (VISTAS). The efforts of all contributors are gratefully acknowledged.
in
-------
CONTENTS
PREFACE ii
ACKNOWLEDGMENTS iii
1.0 INTRODUCTION 1
2.0 CALPUFF ASSESSMENT TOOL OVERVIEW 2
3.0 SUMMARY OF CALPUFF ASSESSMENT RESULTS 3
3.1 PRELIMINARY ASSESSMENT OF VISTAS VERSION 3
3.2 RESULTS OF ADDITIONAL TESTS 5
3.3 SUMMARY OF SIGNIFICANT FINDINGS 9
4.0 DESCRIPTION OF TECHNICAL ISSUES 11
4.1 IS SUE WITH NEW DEFAULT MIXING HEIGHT PARAMETERS 11
4.2 POTENTIAL RAMIFICATIONS OF TECHNICAL ENHANCEMENT 12
4.3 RESOLUTION OF ISSUES 16
5.0 REFERENCES 17
IV
-------
FIGURES
Figure
1. Contour plot of percent differences in 4th-highest 24-hour averages for Scenario 2
from Test 4 8
2. CALMET convective mixing heights for one time step, Test 4 14
3. Time series of convective mixing height from point identified in Figure 1 15
4. Time series of sensible heat flux and convective mixing height for 2 days from
Figure 3 15
5a. Contour plot of convective mixing heights (m) with default THRESFIL for Scenario 4,
10/19/92, 15:00 LST 21
5b. Contour plot of convective mixing heights (m) with default THRESFIL for Scenario 4,
10/19/92, 16:00 LST 22
5c. Contour plot of convective mixing heights (m) with default THRESFIL for Scenario 4,
10/19/92, 17:00 LST 23
-------
TABLES
Table
1. Range of Maximum Absolute Percent Differences by Scenario for Initial Application of
Assessment Tool 3
2. Results for Test No. 1: Range of Maximum Absolute Percent Differences by Scenario 6
3. Results for Test No.4: Range of Maximum Absolute Percent Differences by Scenario 7
4. Range of Maximum Absolute Percent Differences by Scenario 9
5. Range of Maximum Absolute Percent Differences by Scenario 18
6. CALPUFF Comparison Results for Scenarios 1-5 for Base v5.71 la vs. Beta v5.8 19
7. CALPUFF Comparison Results for Scenarios 6-10 for Base v5.71 la vs. Beta v5.8 20
VI
-------
1.0 INTRODUCTION
The CALPUFF Modeling System, consisting of the CALPUFF dispersion model, CALMET
meteorological processor, and CALPOST postprocessor, was promulgated by EPA as the
preferred model for long-range transport (LRT) regulatory modeling applications for purposes of
demonstrating compliance with Class I PSD increments.1 As with any modeling system,
periodic updates are anticipated as part of the standard software life cycle to address bugs that
are identified, as well as enhancements that may be needed to address new data formats or other
needs that may arise. To address the need for a systematic process to assess impacts of
modifications to the CALPUFF modeling system, EPA established a standard "Protocol for
Updating the CALPUFF Modeling System" and developed a "CALPUFF Assessment Tool" to
support that process.2 Such a process is vital to preserving the integrity of the preferred status of
models recommended by EPA in the Guideline on Air Quality Models (40 CFRPart 51,
Appendix W).1
This report summarizes the process undertaken, and documents the results of EPA's assessment
of the "VISTAS" version of the CALPUFF modeling system, which lead to EPA's approval of
CALPUFF (v5.8), CALMET (v5.8) and CALPOST (v5.6394) as the "EPA-approved" version,
announced on June 29, 2007. As part of this assessment, EPA performed several tests to
compare modeled impacts based on the then-current VISTAS versions of CALMET (v5.726)
and CALPUFF (v5.756) to impacts based on the previous EPA-approved versions of CALMET
(v5.53a) and CALPUFF (v5.71 la), for the purpose of assessing whether to update the EPA-
approved version of the modeling system.
-------
2.0 CALPUFF ASSESSMENT TOOL OVERVIEW
The CALUFF Assessment Tool consists of 11 scenarios designed to test the modeling system
across a range of possible applications in terms of modeling domain, meteorological data
options, and source types. The CALPUFF Assessment Tool prepares summaries of differences
in predicted concentrations between two versions of the CALPUFF Modeling System, the
"Base" version referring to the current EPA-approved version, and the "Beta" version referring
to the updated version of the modeling system that is the subject of the assessment. The
CALPUFF Assessment Tool was successfully applied to support EPA's adoption of CALPUFF
(v5.71 la) and CALMET (v5.53a) as the EPA-approved versions in June 2006.3
The following list provides a brief description of the 11 scenarios included in the CALPUFF
Assessment Tool:
1. Large-scale domain in Pacific NW - NWS met data only
2. Same as Scenario 1 with MM5 NOOBS option
3. Medium-scale domain in Pacific NW (subset of 1) with MM5&NWS data
4. Medium-scale domain near Shenandoah NP - NWS met data only
5. Small-scale complex flow with deep valley - NWS met data only
6. Idealized hill with steady-state met, with similarity dispersion
7. Same as Scenario 6, with PG dispersion
8. Flat terrain with steady-state met, with similarity dispersion
9. Same as Scenario 8, with PG dispersion
10. Idealized hill with simulated wind shear, profile met data
11. Same as Scenario 3, with chemistry and deposition included (optional)
The following list provides a brief description of the sources included in the different scenarios:
Four core sources included in all scenarios:
> Ground-level area source (20m by 200m)
> 10m volume source
> 30m non-buoyant point source
> 65m buoyant point source
Two sets of core sources included in Scenarios 1, 2, and 5 at different locations
within domain
Scenario 3 also includes 99m buoyant stack near coast
Scenario 4 also includes buoyant area source
Scenarios 6 and 7 also include three PRIME downwash sources (35m buoyant,
35m capped, 50m buoyant)
More details regarding the scenarios included in the CALPUFF Assessment Tool are provided in
a presentation submitted at EPA's 8th Modeling Conference held in RTF, NC in September,
2005.4
-------
3.0 SUMMARY OF CALPUFF ASSESSMENT RESULTS
3.1 PRELIMINARY ASSESSMENT OF VISTAS VERSION
EPA initiated an assessment of the latest "Beta" version of the CALPUFF modeling system in
June 2006, based on the version being distributed at that time by TRC as the "VISTAS" version.
The initial application of the CALPUFF Assessment Tool to the "VISTAS" version indicated
significant differences between the VISTAS version and the then-current EPA-approved version
across all scenarios, except for Scenario 10, which does not utilize CALMET. A brief summary
of the initial comparison results are provided in Table 1, which shows the range of absolute
maximum percent differences, computed as [100*(Beta-Base)/Base], for the high ranked values
by scenario across all sources and averaging periods by scenario. Note that Scenario 11 is not
included in the summary of comparison results presented here. Scenario 11 is based on Scenario
3, but also includes chemistry and deposition. The magnitude of differences found between the
VISTAS and prior EPA-approved versions necessitated a thorough assessment of the factors
contributing to those differences before EPA could make a determination of whether to approve
the VISTAS version as an update to the previous regulatory version. The CALPUFF
Assessment Tool also attributed most of the differences in Table 1 to changes in CALMET,
rather than CALPUFF. The only scenario that showed differences due to CALPUFF was
Scenario 5, with a maximum percent difference due to CALPUFF of about 7 percent.
Table 1: Range of Maximum Absolute Percent
Differences by Scenario for Initial Application of
Assessment Tool
Scenario 1
Scenario 2
Scenario 3
Scenario 4
Scenario 5
Scenario 6
Scenario 7
Scenario 8
Scenario 9
Scenario 10
-22.0 to +23. 3
-0.2 to +83.4
-17.7 to +60.8
-13.6 to +28.1
-46.0 to +2 1.1
-10.3 to +6.3
-1.7 to +1.0
-10.0 to +5.4
-1.2 to +1.0
No differences
Due to limited documentation available at that time, it was impossible to determine which model
changes were contributing to the differences. Under the auspices of VISTAS, TRC provided
additional documentation regarding the changes between the previous EPA-approved versions of
CALMET (v5.53a) and CALPUFF (v5.71 la) and the then-current VISTAS versions of
CALMET (v5.726) and CALPUFF (v5.756). This additional documentation included:
(1) tables derived from in-code documentation of changes to each model component,
annotated to indicate the category of each change and whether each change may
-------
affect results;
(2) Model Change Bulletin B (MCB-B), for CALPUFF v5.71 la/ v5.71 Ib and
CALMET v5.53a/v5.53b (available from TRC CALPUFF website);
(3) Model Change Bulletin C (MCB-C), for CALPUFF v5.71 Ib/v5.756 and
CALMET v5.53b/v5.726;
(4) Model Change Bulletin D (MCB-D), for CALMET v5.726/v5.730; and
(5) a summary of model tests performed by TRC.
The changes identified by the annotated tables of in-code documentation were classified as
follows:
(1) bug fixes;
(2) non-optional technical enhancements;
(3) optional technical enhancements;
(4) non-technical enhancements;
(5) enhancement adjustments; or
(6) coordinate conversion fixes.
In order to facilitate isolating the potential impact of changes identified as (1) bug fixes, and (2)
non-optional technical enhancements, TRC provided to EPA the following interim versions of
CALMET and CALPUFF:
CALMET (5.53c) = CALMET (5.53a) + bug fixes
CALMET (5.53c2) = CALMET (5.53c) + non-optional technical enhancements
CALPUFF (5.7lie) = CALPUFF (5.711 a) + bug fixes
Another potential source of differences identified by TRC in the summary of their comparison
test related to new CALMET parameters associated with optional technical enhancements for
which new default values had been specified in the VISTAS versions that differed from values
that would be needed to maintain equivalence with the previous EPA-approved version. As
implemented in the VISTAS version, these options required the user to manually override the
new default values in the appropriate input file for the VISTAS version to maintain consistency
with the EPA-approved version. The following list identifies the new default parameters within
the VISTAS version of CALMET, along with the non-default value required to manually
override the default in the CALMET.INP file for consistency with the EPA-approved version:
(1) IMIXH = -1 for convective mixing height option based on Maul-Carson for land
cells only and Original OCD mixing height overwater (default value of 1 would
be selected to use Maul-Carson for land and water cells if IMIXH is omitted);
(2) THRESHL = 0.0 for threshold buoyancy energy flux per meter of boundary layer
required for Mixing Height Growth over land (default value of 0.05 W/m2/m
would be selected if THRESHL is omitted);
(3) THRESHW = 0.0 for threshold buoyancy energy flux per meter of boundary layer
required for Mixing Height Growth over water (default value of 0.05 W/m2/m
would be selected if THRESHW is omitted); and
(4) ICOARE = 0 for option to use delta-T method for the Coupled Ocean Atmosphere
-------
Response Experiment (COARE) bulk flux model for computing wave heights and
boundary layer parameters overwater (default value of 10 for COARE method
with COARE wave option 0 would be selected if ICOARE is omitted).
The following new defaults in the VISTAS version of CALPUFF were also identified, requiring
manual override with the specified value in the CALPUFF.INP file to maintain consistency with
the then-current EPA-approved version:
SVMIN(6:12) = 0.50 for minimum sigma-v value (m/s) for over water cells for stability
classes 1 through 6 (default value of 0.37 m/s would be used for each stability class for
over water grid cells if these parameters are omitted).
3.2 RESULTS OF ADDITIONAL TESTS
Based on the interim versions of CALMET and CALPUFF provided by TRC, and the additional
clarification regarding input file modifications needed for the VISTAS version to maintain
consistency with the EPA-approved version for certain parameters, EPA conducted the following
initial set of test runs in order to compare and isolate affects of the Bug Fixes, the Non-optional
Technical Enhancements, and the New Defaults for Optional Technical Enhancements that
require manual override in the CALPUFF/CALMET input files:
Test 1. CALPUFF (5.71 Ic) and CALMET (5.53c) vs. VISTAS (CALPUFF v5.756 and
CALMET v5.726) without modified inputs for new default parameters;
Test 2. CALPUFF (5.71 Ic) and CALMET (5.53c2) vs. VISTAS (5.756/5.726) without
modified inputs for new default parameters;
Test 3. CALPUFF (5.71 Ic) and CALMET (5.53c) vs. VISTAS (5.756/5.726) with
modified inputs to override new default parameters; and
Test 4. CALPUFF (5.71 Ic) and CALMET (5.53c2) vs. VISTAS (5.756/5.726) with
modified inputs to override new default parameters.
TRC asserted, citing results of their comparison tests5, that "the VISTAS code is equivalent to
the EPA-approved code once the coding errors in the EPA code are corrected." Based on this
assertion, the results for Test No. 1 should demonstrate equivalency. Section 3.2.2 of Appendix
W establishes a maximum threshold of 2 percent differences between the maximum or highest,
second-highest (H2H) results as the criterion for equivalence between an alternative model and
the preferred model. This 2 percent threshold has often been used as a benchmark for assessing
differences between models or model versions, and is useful in the context of this assessment.
Summaries of comparison results obtained from EPA's preliminary tests are presented below in
order to provide an initial indication of whether the criterion of equivalence was met. It is
important to note in assessing the results from these comparisons that the current regulatory
niche for the CALPUFF modeling system in Appendix W is for modeling compliance with
NSR/PSD permitting requirements for long range transport (> 50km) applications at Class I
areas. In light of this, differences between various versions of the CALPUFF modeling system
for design values representative of such regulatory applications are a necessary part of any
-------
comparison study intended to support adoption of an updated version of the model. The
summaries presented here are limited in scope, focusing on the percent differences in high
ranked values without explicit regard to transport distances, and may significantly understate
differences found across the full modeling domain.
The results for Test No. 1, showing the impact of bug fixes without modifications to input files
to override new default parameters, are summarized in Table 2 as the range of absolute
maximum percent differences, computed as [100*(Beta-Base)/Base], for high ranked values
across all sources and averaging periods by scenario.
Table 2: Results for Test No. 1 : Range of
Maximum Absolute Percent Differences by
Scenario
Scenario 1
Scenario 2
Scenario 3
Scenario 4
Scenario 5
Scenario 6
Scenario 7
Scenario 8
Scenario 9
Scenario 10
-22.0 to +24.5
-21. 7 to +81. 8
-17.9 to +60.3
-3. 3 to +30.5
-41. 8 to +19.5
-9.7 to +6.3
-1.7 to +1.0
-9.3 to +5.6
-1.2 to +1.0
No differences
These results clearly did not support the assertion that "the VISTAS code is equivalent to the
EPA-approved code once the coding errors in the EPA code are corrected." In fact, the range of
differences was quite similar to the range of differences found from the initial comparison of the
then-current EPA-approved version vs. the current VISTAS version. This did not necessarily
imply that the impact of bug fixes on model results was insignificant, as will be shown later.
At the other end of the spectrum of initial tests, Test No. 4 was designed to eliminate all of the
documented causes of potential differences relative to the VISTAS version, including Bug Fixes,
Non-optional Technical Enhancements, and New Default Parameters for Optional Technical
Enhancements. Table 3 summarizes the range of maximum percent differences of high ranked
values by scenario for Test No. 4, which shows better overall agreement than initial test results.
-------
Table 3: Results for Test No. 4: Range of
Maximum Absolute Percent Differences by
Scenario
Scenario 1
Scenario 2
Scenario 3
Scenario 4
Scenario 5
Scenario 6
Scenario 7
Scenario 8
Scenario 9
Scenario 10
-17.7 to +0.6
-20.9 to +10.3
-0.4 to +5.1
No differences
No differences
No differences
No differences
No differences
No differences
No differences
While no differences were found for several scenarios, some significant differences remained for
Scenarios 1, 2 and 3, indicating that other undocumented factors were contributing to differences
in concentrations. Preliminary analysis of the spatial pattern of differences across the domain
showed much larger differences in high ranked values for distance ranges typical of long range
transport/Class 1 area impact assessments, as shown in Figure 1 for the Jordan Valley 65m
buoyant source within Scenario 2, confirming the caution stated above that differences based
solely on high ranked values may significantly understate the magnitude of differences found
across the full modeling domain.
The causes of differences found for Test No. 4 were initially unresolved based on the
documentation available at the time. However, given the results summarized above for Tests 1
and 4, additional tests were performed to further isolate the differences that could be attributed to
each of the three known causes identified above: 1) bug fixes; 2) non-optional technical
enhancements; and 3) new default parameters for optional technical enhancements. Results from
the following additional tests are presented below:
Test 5. CALPUFF (5.71 la) and CALMET (5.53a) vs. CALPUFF (5.711c) and
CALMET (5.53c) - Differences Due to Bug Fixes;
Test 6. CALPUFF (5.71 Ic) and CALMET (5.53c) vs. CALPUFF (5.711c) and
CALMET (5.53c2) - Differences Due to Non-optional Technical Enhancements;
and
Test 7. CALPUFF (5.576) and CALMET (5.526) without modified inputs for new default
parameters vs. CALPUFF (5. 576) and CALMET (5. 526) with modified inputs
for new default parameters - Differences Due to New Default Parameters
The range of maximum absolute percent differences of high ranked values by scenario for Test 5
(Bug Fixes), Test 6 (Non-optional Technical Enhancements), and Test 7 (New Default
Parameters are presented in Table 4.
-------
Figure 1: Contour Plot of Percent Differences for Scenario 2
Jordan Valley 65m Point Source; 4th-Highest 24-Hour Averages
Test 4 - CALPUFF (5.711c) and CALMET (5.53c2) vs. VISTAS (5.756/5.726)
(with modified inputs to override new default parameter)
200
100
100
200
300
th
Figure 1. Contour plot of percent differences in 4 -highest 24-hour averages for
Scenario 2 from Test 4.
-------
Table 4: Range of Maximum Absolute Percent Differences by Scenario
Scenario 1
Scenario 2
Scenario 3
Scenario 4
Scenario 5
Scenario 6
Scenario 7
Scenario 8
Scenario 9
Scenario 10
Test 5 - Bug Fixes
-1.0 to +22.6
-15.4 to +27.8
-5.1 to +6.0
-18.4 to +8.4
-32.4 to +9.5
-1.4 to +0.6
Nodifferences>|0.01|
-1.4 to +0.6
No differences > |0.01|
No differences
Test 6 - Non-Optional
Technical Enhancements
-18. 6 to +1.9
-30.5 to +0.7
-0.8 to +1.7
-0.7 to +4.2
-9.2 to +0.6
-2.0 to +6.3
-1.7 to +1.0
-2.0 to +3.1
-1.2 to +1.0
No differences
Test 7 - New Default
Parameters
-22.0 to +24.4
-21. 8 to +118.3
-19.3 to +60.8
-3.2 to +30.5
-4 1.8 to +20.9
-12.6 to +5.5
0.0 to +0.4
-12.1 to +5.5
No differences > |0.01|
No differences
3.3 SUMMARY OF SIGNIFICANT FINDINGS
While the summaries of comparison results presented above are limited in scope, based only on
the comparison of high ranked values, a few significant findings are worth noting:
1. Significant differences could be attributed to each of the three known factors that had
been identified (Bug Fixes, Non-optional Technical Enhancements, and New Default
Parameters for Optional Technical Enhancements);
2. Of the three known factors, the New Default Parameters for Optional Technical
Enhancements appeared to cause the largest differences overall;
3. Differences varied significantly across different scenarios and source types, with no
significant overall bias evident, although there appeared to be some tendency for the
Non-optional Technical Enhancements to decrease concentrations, while the New
Default Parameters for Optional Technical Enhancements showed an opposite trend,
to some degree.
The significant contribution of New Default Parameters for Optional Technical Enhancements to
differences between the then-current EPA-approved version and the current VISTAS version
was somewhat unexpected, based on the limited documentation of differences between the two
versions. These additional tests confirmed previous results indicating that most of the
differences were due to CALMET rather than CALPUFF. The only significant differences
attributable to CALPUFF were for the valley sources in Scenario 5, which is a near-field
complex flow scenario along the Columbia River gorge. These additional tests indicated that all
of the differences attributable to CALPUFF were due to the New Default Parameter for
minimum sigma-v over water, which was reduced from 0.5m to 0.37m in the VISTAS version.
The maximum percent differences for Scenario 5 due to this new default in parameter in
CALPUFF ranged from -0.2 to +7.0.
-------
Further evaluation of results related to the New Default Parameters in CALMET indicated that
the new default threshold buoyancy energy flux for convective mixing heights over land
(THRESHL) and over water (THRESHW) were the major factors contributing to these
differences. This finding supported earlier indications that differences in mixing heights
between the then-current EPA-approved version and the VISTAS version could be a major factor
contributing to the differences in concentrations. Technical concerns regarding the
implementation of these new default parameters for convective mixing heights were shared with
TRC, but these concerns remain unresolved. The importance of convective mixing to both near
field and long range dispersion estimates for many sources, the lack of adequate technical
justification or validation for these new default parameters, and the significance of their impact
on results based on these tests, presented an insurmountable obstacle to approval of the VISTAS
version of the CALPUFF modeling system. These technical concerns regarding the new default
parameters for convective mixing heights are described in more detail in the next section.
10
-------
4.0 DESCRIPTION OF TECHNICAL ISSUES
4. 1 ISSUE WITH NEW DEFAULT MIXING HEIGHT PARAMETERS
In 2006, significant modifications to the CALPUFF modeling system were introduced by Earth
Tech under contract to the U.S. Department of the Interior Minerals Management Service
(MMS) for overwater and coastal applications of the modeling system. These include the option
to use similarity profiles and bulk flux algorithms based upon observations from the Coupled
Ocean- Atmosphere Response Experiment (CO ARE) and use of the convective boundary layer
algorithms in marine environments. These efforts were motivated in part by a recognition that
CALMET formulations for overwater dispersion based on the Offshore and Coastal Dispersion
(OCD) model lacked a convective mixing component, and could significantly underestimate
turbulence levels and boundary layer depths over the Gulf of Mexico.
With the addition of the option to compute convective boundary layer heights over water, the
model developer noted the potential for unrealistic growth of the convective boundary layer
heights over areas with significant and sustained values of surface energy flux. According to the
report entitled "Development of the Next Generation Air Quality Models for Outer Continental
Shelf (OCS) Applications, Final Report: Volume I",6 it was noted that observed positive values
of surface buoyancy flux over warm bodies of water such as the Gulf of Mexico were persistent
for days during situations with cold air advection from continental air masses. Observed marine
boundary layer heights were at equilibrium even in the presence of sustained surface buoyancy
fluxes, supporting the notion that sustained convective mixing growth must be offset by some
form of energy dissipation. Without some mechanism of buoyant energy dissipation in the
model, convective mixing heights would grow continuously without interruption in these cases.
In order to prevent the continual growth of the convective boundary layer, the model developer
introduced a threshold surface buoyancy energy flux necessary to sustain growth of the
convective boundary layer. It was conjectured that as the marine convective boundary would
grow, so would the energy requirements to sustain that growth. A similar threshold (THRESHL)
was encoded for overland convective mixing heights for consistency. The model developers
conjectured that the overland threshold would be much less important because of diurnal
variations as well as the much higher values of surface heat flux overland.
The threshold buoyancy energy flux is characterized as the quantity of surface heat flux required
to sustain the convective mixing height growth, expressed in W/m2 per meter of mixing height.
Separate threshold terms were added for land (THRESHL) and water (THRESHW), and were
incorporated into both the Maul-Carson (current recommended) and new Gryning-Batchvarova
mixing height algorithms. Default values of 0.05 W/m2/m were set for both THRESHL and
THRESHW. The magnitude of the threshold kinematic heat flux at the surface is computed by
the following equation:
11
-------
where
(w'6')th = threshold kinematic heat flux (K m/s)
THRESH = threshold buoyancy energy flux parameter = 0.05 W/m2/m
ZiCh-\ = convective mixing height from previous hour (m)
p = density of air (kg/m3)
cp = specific heat at constant pressure (J/kg/K)
At each time step, CALMET checks to see whether sufficient buoyant energy is available to
sustain the growth of the convective boundary layer by simply subtracting the threshold quantity
necessary to sustain the growth based on Equation 1 from the hourly buoyant energy flux. While
the documentation provided describes this threshold as a threshold to "sustain the growth" of the
boundary layer, a review of its implementation in the CALMET code indicated that once the
difference falls below zero, CALMET resets the convective boundary layer height for the current
time step to zero. The effect of setting the convective mixing height to zero would be partially
masked in most applications because the hourly mixing height output from CALMET for these
cases would be higher of the default minimum mixing height (Zmin=50m) and the daytime
mechanical mixing height (Zmech), and also due to the default option for upwind averaging of
mixing heights. However, the convective velocity scale (w%), which is used by CALPUFF to
scale vertical turbulence under convective conditions, is based only on the convective mixing
height.
As noted previously, EPA's assessment found that significant differences between the previous
EPA-approved version and the VISTAS version were attributable to the implementation of these
new mixing height parameters. Visualization of the CALMET mixing heights from Scenario 4,
a typical LRT Class I area scenario including the Shenandoah National Park and other nearby
Class I areas, showed multiple periods when the convective boundary layer would collapse over
large areas over land, contrary to the conjecture that the overland threshold would be much less
important. For example, at the location identified in Figure 2, the overland convective boundary
layer collapsed during the middle of the day on three out of four consecutive simulation days
(Figure 3). The boundary layer heights on these three days (10/17, 10/18, and 10/20) changed
from 957m, 598m, and 1088m to Om and then back to 599m, 437m, and 267m within a span of
three hours on each of the three respective days. A very small difference between the buoyant
energy flux and the threshold is all that is necessary to cause the complete collapse of the
convective boundary layer. Note that the convective mixing heights presented in these figures
were generated using no minimum mixing height and no upwind averaging of mixing heights in
CALMET.
4.2 POTENTIAL RAMIFICATIONS OF TECHNICAL ENHANCEMENT
When this anomalous behavior was confirmed in the CALMET meteorological fields, some
important implementation issues were raised by EPA. First, the sudden collapse and
regeneration of the convective boundary layer creates physically unrealistic mixing height
patterns with extremely tight spatial gradients that would not normally be observed in the
12
-------
atmosphere. When using CALMET with meteorological observations only (no prognostic
meteorological datasets used for the first guess windfield), certain derived meteorological values
rely upon data from the nearest surface or upper air meteorological site in the modeling domain.
These create physical discontinuities in two-dimensional meteorological fields when relying
upon nearest station data. In CALMET, the surface buoyant energy flux calculation found in
both the Maul-Carson and the Gryning-Batchvarova convective mixing height schemes relies
upon the conversion of sensible heat flux which is in turn derived from the nearest station with
non-missing cloud cover data. When the buoyant energy flux is derived in such a way and the
energy is insufficient to sustain the convective boundary layer growth, the convective boundary
layer will collapse at every grid point which derives sensible heat flux from cloud cover
observations at a particular surface station. This anomalous convective mixing height behavior
is illustrated in Figure 5 as a series of contour plots of convective mixing height for a 3-hour
period on Oct. 19, 1992 based on Scenario 4. The visualization shows that the collapse of
convective mixing height occurs over distances of more than 100 km. Derived
micrometeorological values such as convective velocity scale (vc«), which contains the
convective boundary layer height in its equation, will exhibit the same spatial patterns as the
convective boundary layer.
Another important issue is the impact that unrealistic changes in micrometeorological parameters
have on dispersion. When employing dispersion coefficients from internally computed
micrometeorological variables (MDISP=2), these unrealistic changes to the micrometeorological
variables will have a significant impact upon puff growth rates. The collapse of the convective
boundary layer and associated changes in dispersion will have varying impacts on surface
concentrations depending upon source type. For low-level sources, this sudden collapse of the
daytime boundary layer is likely to significantly increase ground level concentrations by
essentially trapping the puffs within an artificially compressed daytime boundary layer. For
elevated point sources, the opposite effect may manifest itself, as these sources will be emitting
puffs well above the mixing height. Comparison results from Test 7 show evidence of these
patterns of impact where the largest positive percent difference of+188 percent (VISTAS
version > EPA-approved version) was for a low-level area source in Scenario 2, and the largest
negative difference of-41.8 percent was for a 65m buoyant source in Scenario 5.
13
-------
MixH.
1431.318
1227.273
1022.727
818,1818
613.6364
409.0909
1204.5455
0
575
625
650 675
UTM East (km)
700 725
750
Figure 2. CALMET convective mixing heights for one time step, Test 4.
14
-------
1,500
1,450
1,400
1,350
1,300
1,250
1,200
1,150
1,100
1,050
1,000
950
900
550
500
350
300
250
15/10^99200:00 1 5.-1 Oil 992 > 2 Or 1010.0092 00 00 I F.'l o.n 092 I 2 00 ! 70 0.0 092 00 00 1 0 0 " 99 J < 2 ::. I:.'.'! 0." 00 J .10 :: I ti.n 0 - 90 J 1 2 10 I 9.0 O.'l 992 00 10 -.9.0109921200 20.i'!01 092 or. 00 20.il I'll 092 1200
Figure 3. Time series of convective mixing height from point identified in Figure 1.
1 j£VU
1000 -
800 -
i,
*-^
D)
"O 600 -
X
m
400 -
200 -
n -
f:
']
-!
* i I
. T
?Jj/si
I * '
9 4
!
|«
to
s
/ «
3 ' \ i
:| 3 / i|
3 /A iJt
i /-jit
1 t n
\ \ I ||
200
150
100
E
I
u
JD
'
- 50
10
20
30
Hours
40
50
60
Sensible Heat Flux
Convective Boundary Layer
Figure 4. Time series of sensible heat flux and convective mixing height for 2 days from Figure
15
-------
4.3 RESOLUTION OF ISSUES
Based on the results of the comparisons summarized above, EPA determined that insufficient
documentation and justification was available to approve the VISTAS version of the CALPUFF
modeling system for regulatory applications. In order to resolve this situation, TRC agreed to
implement the following changes in the CALPUFF modeling system code:
1. Incorporate the non-optional technical enhancements under the optional technical
enhancements, removing non-optional technical enhancements as a potential source
of differences;
2. Incorporate a new regulatory default switch (MREG=1) in CALMET to allow the
optional technical enhancements to be included in the model, but to require the user
to override the default options to exercise these optional technical enhancements; and
3. Modify the CALPUFF model code to include the minimum sigma-v of 0.5 m/s over
water as part of the regulatory default setting.
These modifications were incorporated into CALPUFF (v5.8) and CALMET (v5.8), which were
then approved by EPA for regulatory modeling. These actions effectively isolated bug fixes as
the only source of differences between the previous EPA-approved version and the new versions
of CALMET and CALPUFF (v5.8), when applied in the regulatory default mode. Table 5
includes a summary of the final comparisons between the previous EPA-approved version
(CALPUFF, v5.71 la) and the new EPA-approved version (CALPUFF, v5.8), identified as Test
8. Comparison results for Tests 5, 6, and 7 are also presented for reference. Table 5 confirms
that comparison results for Test 8 are nearly identical to results for Test 5, which was intended to
isolate impacts due solely to bug fixes. The minor differences noted between Tests 5 and 8 for
Scenarios 2 and 3, both of which include the use of MM5 data, were attributable to additional
bug fixes incorporated into version 5.8 that were identified by TRC as part of this assessment. A
more detailed summary of maximum absolute differences in high ranked values by source and
scenario are presented in Table 6 for Scenarios 1 through 5 and in Table 7 for Scenarios 6
through 10.
16
-------
5.0 REFERENCES
1. EPA, 2005. Guideline on Air Quality Models. 40 CFR Part 51 Appendix W.
2. EPA, 2008. Protocol for Updating the CALPUFF Modeling System.
3. EPA, 2005. EPA 8th Modeling Conference CALPUFF presentation,
(http://www.epa.gov/ttn/scram/8thmodconf/presentations/day2morning/calpuffanalysisto
olrevised.ppt)
4. EPA, 2006. CALPUFF Modeling System Update #1 (version 5.71 la).
5. TRC, 2007. Summary of Model Tests Conducted with the VISTAS-Approved CALMET
and CALPUFF Model Codes vs. EPA-Approved Versions of the Models, prepared for
VISTAS.
6. Earth Tech, 2006. Development of the Next Generation Air Quality Models for Outer
Continental Shelf (OCS) Applications, Final Report: Volume 1, MMS 2006-006,
prepared for U.S. Department of the Interior, Minerals Management Service.
17
-------
Table 5: Range of Maximum Absolute Percent Differences by Scenario
Scenario 1
Scenario 2
Scenario 3
Scenario 4
Scenario 5
Scenario 6
Scenario 7
Scenario 8
Scenario 9
Scenario 10
Test 5 - Bug Fixes
-1.0 to +22.6
-15.4 to +27.8
-5.1 to +6.0
-18.4 to +8.4
-32.4 to +9.5
-1.4 to +0.6
No differences > |0.01|
-1.4 to +0.6
No differences >|0. 01 1
No differences
Test 6 - Non-Optional
Technical Enhancements
-18.6 to +1.9
-30.5 to +0.7
-0.8 to +1.7
-0.7 to +4.2
-9.2 to +0.6
-2.0 to +6.3
-1.7 to +1.0
-2.0 to +3.1
-1.2 to +1.0
No differences
Test 7 - New Default
Parameters
-22.0 to +24.4
-21. 8 to +118.3
-19.3 to +60.8
-3.2 to +30.5
-4 1.8 to +20.9
-12.6 to +5.5
0.0 to +0.4
-12.1 to +5.5
No differences > |0.01|
No differences
Test 8 - Base 5.71 la vs.
Beta5.8(FinalTest)a
-1.1 to +22.6
-2 1.8 to +29.2
-4.0 to +7.3
-18.4 to +8.4
-32.4 to +9.5
-1.4 to +0.6
No differences > |0.01|
-1.4 to +0.6
No differences > |0.01|
No differences
Test 8 differences are due to bug fixes, but include some additional bug fixes compared to Test 5.
18
-------
Table 6. CALPUFF Comparison Results for Scenarios 1 - 5 for Base v5.71 la vs. Beta v5.8
Scenario
1
2
3
4
5
Source Type
Area Source
X 0.12
X 0.41
X -21.78
X 0.68
X -0.02
X 8.40
X 0.02
X 0.32
Volume Source
X 0.20
X 1.07
X -6.71
X 2.35
X 0.00
X -4.55
X -0.30
X 0.12
30 m Point Source
X 0.48
X 0.28
X -2.84
X 0.58
X -0.05
X .4.95
X -0.88
X -5.10
65 m Point Source
X 7.59
X 22.63
X 29.24
X 25.24
X 7.26
X -18.37
X -32.43
X -7.36
99 m Point Source
X -3.98
Buoyant Area Sourc*
X -13.02
No differences in highest ranked (design) values
Differences in one or more design values
19
-------
Figure 7. CALPUFF Comparison Results for Scenarios 6 -10 for Base v5.711a vs. Beta v5.8
Stable
Neutral
Unstable
Similarity
PG
Area
0.002
0.002
-0.003
-0.003
0.137
0.005
NX-
0.002
-0.003
-0.003
0.137
0.005
V-"
v^-
Volume
V-"
^^
0.005
vx*
0.565
\x--
vx-
vx-
0.005
V-"
0.565
vx-
xx-
vx-
30m stack
\^
NX*
0.006
NX*
0.626
NX*
NX*
NX*
0.006
NX*
0.626
NX-
NX*
NX-
Key to results for Scenarios 6-10
Simulated Hill Flat Terrain ?mulated Hill
(with wind shear)
6
7
8
9
10
65 m stack
NX*
NX*
0.018
NX-
-1.419
NX*
NX-
NX*
0.018
NX-
-1.419
NX-
NX*
NX-
NX*
X
Downwash from
35m capped stack
NX*
NX-
-0.001
NX*
0.541
NX*
Downwash from
35m non-capped stack
-0.824
NX-
0.003
NX*
-0.068
NX*
Downwash from
50 m non-capped stack
NX-
NX*
0.022
NX-
-1.282
NX*
No differences found
Differences found
20
-------
Corrective Mixing Height (10/19/1992) at 15:00 LSI
MixH.
1431.818
1227.273
1022.727
818,1818
613.6364
409.0909
204,5455
0
575 600 625 650 675
UTM East (km)
700
725
750
Figure 5a. Contour plot of convective mixing heights (m) with default THRESHL for Scenario 4, 10/19/92, 15:00 LSI
21
-------
Convective Mixing Height (1 Ofi 9/1992) at 16:00 LSI
MixH.
1431.818
1227,273
1022.727
818.1818
613,6364
409.0909
204.5455
0
575
600 625
650 675
UTM East (km)
700
725
750
Figure 5b. Contour plot of convective mixing heights (m) with default THRESHL for Scenario 4, 10/19/92, 16:00 LSI
22
-------
Convective Mixing Height (10/19/1992) at 17:00 LSI
4316.15
575
600
625
650 675 700
UTM East (km)
750
Figure 5c. Contour plot of convective mixing heights (m) with default THRESHL for Scenario 4, 10/19/92, 17:00 LSI
23
-------
United States Office of Air Quality Planning and Standards Publication No. EPA-454/R-08-007
Environmental Protection Air Quality Assessment Division August 2008
Agency Research Triangle Park, NC
------- |