Get documents about "WRAP 2002 Visibility Modeling Overview of 2005 RMC Modeling
Document Sample
Overview of 2005 RMC Modeling Activities
WRAP 2002 Visibility Modeling:
Overview of 2005 RMC Modeling Activities
Gail Tonnesen, Zion Wang, Mohammad Omary, Chao-Jung Chien, Yingqun Wang
University of California, Riverside
Zac Adelman
University of North Carolina
Ralph Morris et al.
ENVIRON Corporation Int., Novato, CA
Overview of 2005 RMC Modeling Activities
Goals
Key RMC visibility modeling work elements include the following:
• Evaluation of the visibility model for a historical episode
– for calendar year 2002.
– Output from the model simulation is compared with ambient air quality data
for the historical episode as part of a model performance evaluation (MPE).
• Development of visibility planning scenarios
– for the regional haze baseline period of 2000-04 and
– for the initial regional haze future projection period, calendar year 2018.
• Modeling a variety of
– emissions sensitivity, emissions source apportionment, and emissions
control strategies
– to assess whether planned future regional emissions reductions will be
sufficient to demonstrate reasonable progress toward achieving visibility
goals.
Overview of 2005 RMC Modeling Activities
Modeling Activities during 2005
• CMAQ v44beta vs. v44 final release
• Preliminary 2002 Fire Sensitivity Scenarios
• 2002 Preliminary Version D
• 2002 Fugitive Dust Comparison
• CAMx versus CMAQ 4.4 Preliminary Version D Comparison
• CMAQ v4.4 36km 2002 Base Case A Model Performance Evaluation
• CMAQ v.4.4 versus v.4.5 comparison: 2002 36km Base Case A
• CMAQ v.4.5 12km vs 36km comparisons using 2002 Base Case A
• Preliminary testing and benchmarks for CAMx PSAT using 2002 Base A
• CAMx v.4.3 & CMAQ v.4.5 36-km Comparison using 2002 Base Case A
• Small Fire BaseA 2002 Sensitivity Scenarios (12-km CMAQ results)
• 2018 vs. 2002 Planning cases comparisons (CMAQ Base18a vs. Plan02a)
Overview of 2005 RMC Modeling Activities
Modeling Domain
WRAP 36-km CMAQ/CAMx Domain WRAP CMAQ domain:
within MM5 36-km domain red: 36-km blue: 12-km
Overview of 2005 RMC Modeling Activities
CMAQ Model Performance Evaluation in the
WRAP States for Calendar Year 2002
Overview of 2005 RMC Modeling Activities
CMAQ Model Performance Evaluation in the
WRAP States for Calendar Year 2002
Monthly average sulfate
MFB for CMAQ version 4.5
36-km results compared to
the CASTNet, IMPROVE,
and STN networks.
Monthly average sulfate
for CMAQ v.4.5 36-km
Overview of 2005 RMC Modeling Activities
CMAQ Model Performance Evaluation in the
WRAP States for Calendar Year 2002
• Sulfate:
– meets the performance goals for both the CASTNet and IMPROVE data
– CMAQ 4.5 had
• large under prediction for summer and
• small positive bias for winter.
• Nitrate:
– meets the performance goals for both the CASTNet and IMPROVE data,
– but fails to meet
• both the performance goals and
• the less stringent performance criteria for most months for the STN data.
• OC:
– For IMPROVE, the model met the performance goals for MFB for all months,
and it met the performance goals for MFE for all but two months.
– In contrast, the model failed to meet
• the performance goals for OC at the STN sites for all 12 months, and
• it failed to meet the less stringent performance criteria for most months.
Overview of 2005 RMC Modeling Activities
CMAQ Model Performance Evaluation in the
WRAP States for Calendar Year 2002
• EC:
– The model met the performance goals for EC for both the IMPROVE and STN
data.
– Interestingly, the model had opposite trends in performance for the two
networks.
• Soil:
– The model met the performance goals for soil for most months, and when it
failed to meet the performance goals it did meet the performance criteria.
• CM:
– The model failed to meet both the performance goals and the performance
criteria for CM for most months.
Overview of 2005 RMC Modeling Activities
CMAQ version 4.4 versus 4.5 comparison
2002 36km Base Case A
Overview of 2005 RMC Modeling Activities
CMAQ version 4.4 versus 4.5 comparison
2002 36km Base Case A
Overview of 2005 RMC Modeling Activities
CMAQ version 4.4 versus 4.5 comparison
2002 36km Base Case A
• sulfate & nitrate:
– CMAQ v4.5 had improved performance (i.e., lower positive bias) in the winter months
and
– poorer performance (i.e., increased negative bias) in the summer months.
• OC:
– CMAQ v4.5 had significantly improved performance
• EC:
– CMAQ v4.5 predicted lower EC concentrations than did CMAQ v4.4;
– larger negative bias during the winter months, but improved performance during the
summer months
• Soil:
– CMAQ v4.5 predicted lower soil concentrations than did CMAQ v4.4;
– larger negative bias during the winter months, but improved performance during the
summer months.
• CM:
– CMAQ v4.4 underpredicted CM during most months, and the underpredictions became
worse with CMAQ v4.5.
Overview of 2005 RMC Modeling Activities
CMAQ v4.5 12km versus 36km comparisons
using 2002 Base Case A
Overview of 2005 RMC Modeling Activities
CMAQ v4.5 12km versus 36km comparisons
using 2002 Base Case A
• Performed a CMAQ model sensitivity simulation using a 12-km horizontal
resolution grid for the WRAP nested subdomain
• 12-km simulation of the MM5 model and emissions processing at the 12-
km resolution were performed
• Primary adventages of running a 12-km model
– Better resolved and possibly more accurate meteorology and emissions input
data.
– Less numerical dispersion, resulting from reduced artificial dilution compared
to point sources being averaged over the coarse 36-km grid.
– Less numerical dispersion in the advection algorithms.
– Improved accuracy resulting from higher topographical resolution.
– More accurate treatment of nonlinear photochemical reactions.
– More precise location of the ambient monitoring sites relative to topographical
features and emissions sources.
Overview of 2005 RMC Modeling Activities
CMAQ v4.5 12km versus 36km comparisons
using 2002 Base Case A
• Primary disadvantages of running a 12-km model
– the additional cost of modeling the much larger number of grid cells
required to represent the nested subregion, and
– the additional cost of having to perform a coarse-grid model simulation
as well as the 12-km simulation, because the 36-km simulation is
needed for developing boundary conditions for the 12-km grid.
Overview of 2005 RMC Modeling Activities
CMAQ v4.5 12km versus 36km comparisons
using 2002 Base Case A
36 km 12 km
Overview of 2005 RMC Modeling Activities
CMAQ v4.5 12km versus 36km comparisons
using 2002 Base Case A
IMPROVE CASTNET sulfate
sulfate
20.0 10
10.0 0
Fractional Gross Error
Fraction Gross Error
0.0 -10
-20
-10.0 12km 12km
-30
-20.0 36km 36km
-40
-30.0
-50
-40.0 -60
-50.0 -70
January February June July November January February June July November
12km -0.4 -5.4 -40.8 -33.3 3.2 12km -20.5 -21.7 -56.8 -35.2 -2.2
36km 1.0 -0.8 -38.1 -23.1 10.8 36km -24.6 -19.4 -61.1 -35.1 2.3
NADP sulfate STN sulfate
40 5
30 0
Fraction Gross Error
Fraction Gross Error
-5
20
-10
10
-15
12km 12km
0 -20
36km 36km
-10 -25
-30
-20
-35
-30
-40
-40 -45
January February June July November January February June July November
12km 27.8 4.6 -4.1 -14.5 17.7 12km 0 1.9 -25.2 -37.8 -2.4
36km -8 -14.9 -15.3 -34 6.9 36km -32.7 -29.3 -42.7 -39.5 -23.5
Overview of 2005 RMC Modeling Activities
CMAQ v4.5 12km versus 36km comparisons
using 2002 Base Case A
• Spatial resolution differences:
– The spatial features or plumes in the concentrations are more detailed
in the 12-km plot than in the 36-km plot, and
– the peak model value in the domain is usually considerably greater in
the 12-km model than in the 36-km model.
– the increased spatial detail and increased peak concentration do not
significantly change the general appearance of the plots nor do they
change the regions that experience high or low PM concentrations.
– there appears to be no significant benefit in using the finer-resolution
grid for modeling the lower concentrations of PM that typically occur
at Class I areas.
Overview of 2005 RMC Modeling Activities
CMAQ v4.5 12km versus 36km comparisons
using 2002 Base Case A
• Model Performance Evaluation:
– although there are small differences between the MFBs calculated for
the 12-km and the 36-km models, their performance is quite similar.
– there would be no advantage in terms of our MPE to running the more
resource-intensive 12-km model instead of the 36-km model.
• 12-km vs 36-km responses to emissions control:
– In a separate study using the VISTAS 12-km and 36-km model
scenarios, we found that the model grid resolution had only small
effects on the model response to emissions reductions.
Overview of 2005 RMC Modeling Activities
CMAQ v4.5 12km versus 36km comparisons
using 2002 Base Case A
• In summary:
– because of the substantial increase in the time and compute resource
costs of performing high-resolution modeling,
– we do not recommend the routine application of additional 12-km
modeling as part of the WRAP planning effort.
– However, this conclusion applies only to the clean conditions
experienced at Class I areas.
– It is likely that the finer-resolution model would provide some benefit
for modeling higher ozone and PM2.5 concentrations relevant to
health-based air quality standards in urban or suburban areas.
Overview of 2005 RMC Modeling Activities
Preliminary testing and benchmarks for
CAMx PSAT using 2002 Base A
Overview of 2005 RMC Modeling Activities
Preliminary testing and benchmarks for
CAMx PSAT using 2002 Base A
1800
1584
1368
18
1152
936 12
14
720 5 8 15
9
504 4 10
288 13
72 6 16
11 3
-144 2
16
15
-360
-576 1 7
-792
-1008 15
-1224
-1440 17
-1656 16
-1872
-2088
-2736 -2304 -1872 -1440 -1008 -576 -144 288 720 1152 1584 2016 2448
Overview of 2005 RMC Modeling Activities
Preliminary testing and benchmarks for
CAMx PSAT using 2002 Base A
• Six source categories:
– Point (including stationary off-shore and anthropogenic fire)
– anthropogenic fires (prescribed fire, agricultural fire, non-federal range
fires)
– Total mobile (on-road, off-road, including planes, trains, ships in/near
port, off-shore shipping)
– natural fires (natural fire, biogenics)
– non-wrap fires
– Area sources (each WRAP state, Pacific Off-shore Marine Shipping
Region, the group of CENRAP states touching WRAP, remaining
contiguous US East, including Gulf of Mexico, Mexico, Canada)
Overview of 2005 RMC Modeling Activities
Preliminary testing and benchmarks for
CAMx PSAT using 2002 Base A
Species Number of RAM Disk Run Time Run Time
Tracers Memory Storage per with 1 CPU with 2 CPU
day no OMP OMP
SO4 2 1.6 GB 1.1 GB 4.7 hr/day 4 hr/day
NO3 7 1.7 GB 2.6 GB 13.2 hr/day Not tested
SO4 & NO3
9 1.9 GB 3.3 GB 16.8 hr/day Not tested
combined
SOA 14 3.8 GB Not tested Not tested Not tested
Primary PM
1.5 GB 3.0 GB 10.8 hr/day Not tested
species
Benchmarks for PSAT computational costs for each PM species.
Run time is for one day (01/02/2002) of the WRAP 36-km domain.
Overview of 2005 RMC Modeling Activities
CAMx v.4.3 versus CMAQ v.4.5
36-km Comparison using 2002 Base Case A
Overview of 2005 RMC Modeling Activities
CAMx v.4.3 versus CMAQ v.4.5
36-km Comparison using 2002 Base Case A
• MPE:
– performed an annual CAMx simulation on the same 36-km
grid that was used for the CMAQ modeling.
– Emissions data were consistent with the CMAQ data
• either processed using SMOKE with the same emissions input data
• or converted from CMAQ binary files to CAMx binary files using
custom software
– same horizontal (36 km) and vertical (19 layers) grid
structure was used
Overview of 2005 RMC Modeling Activities
CAMx v.4.3 versus CMAQ v.4.5
36-km Comparison using 2002 Base Case A
CAMx SO4
CASTNET IMPROVE STN
100
80
Fractional Bias(%)
60
40
20
0
-20
-40
-60
-80
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Sulfate monthly MFB for
CAMx compared to ambient
data at the CASTNet,
IMPROVE, and STN sites.
monthly average sulfate
MFB (top) and MFE (bottom)
for CAMx 36-km results
Overview of 2005 RMC Modeling Activities
CAMx v.4.3 versus CMAQ v.4.5
36-km Comparison using 2002 Base Case A
IMPROVE SO4 IMPROVE NO3
100 150
80
100
60
Fractional Bias(%)
Fractional Bias(%)
50
40
20 CAMx 0 CAMx
0 CMAQ V45 -50 CMAQ v45
-20
-100
-40
-60 -150
-80 -200
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
IMPROVE OC IMPROVE EC
30 40
20 30
10 20
Fractional Bias(%)
Fractional Bias(%)
0
10
0
-10 CAMx CAMx
-10
-20 CMAQ V45 CMAQ V45
-20
-30
-30
-40 -40
-50 -50
-60 -60
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Overview of 2005 RMC Modeling Activities
CAMx v.4.3 versus CMAQ v.4.5
36-km Comparison using 2002 Base Case A
IMPROVE SOIL IMPROVE CM
120 0
100 -20
80
Fractional Bias(%)
Fractional Bias(%)
-40
60
40 -60
CAMx CAMx
20 -80
CMAQ V45 CMAQ V45
0 -100
-20
-120
-40
-60 -140
-80 -160
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Overview of 2005 RMC Modeling Activities
CAMx v.4.3 versus CMAQ v.4.5
36-km Comparison using 2002 Base Case A
• Sulfate:
– very similar performance for the two model
– CAMx overpredicted sulfate at the IMPROVE and CASTNet sites during the
winter months, but
– good agreement with the observed sulfate, with small bias during the summer
months.
– CAMx had small negative (underprediction) bias throughout the year at the
more urban STN sites.
• Nitrate:
– CAMx had a large positive bias for nitrate at the IMPROVE and CASTNet
sites during the winter months and a large negative bias during the summer
months.
– CMAQ generally met the performance goals at the IMPROVE and CASTNet
sites
– CAMx failed to meet the performance goals and criteria during the winter
months with higher nitrate concentration
– Both models failed to meet performance criteria for some months at the STN
sites
Overview of 2005 RMC Modeling Activities
CAMx v.4.3 versus CMAQ v.4.5
36-km Comparison using 2002 Base Case A
• OC:
– CMAQ had generally better performance than CAMx for OC considering
monthly average over all IMPROVE sites.
– CMAQ met the performance goals and criteria for IMPROVE while
– CAMx failed to meet the performance goals and marginally met the criteria.
– Both models failed to meet performance goals at the STN sites, although
CMAQ met the performance for some months at the STN sites.
• EC:
– CMAQ predicted lower EC concentrations than did CAMx
– both models met the performance goals
• Soil:
– each model failed to meet the performance goals for a few months,
– with CAMx erring on the side of overprediction and CMAQ erring on the side
of underprediction.
• CM:
– both models fail to meet both the performance goals and criteria.
Overview of 2005 RMC Modeling Activities
Small Fire BaseA 2002 Sensitivity Scenarios
(12-km CMAQ results)
Overview of 2005 RMC Modeling Activities
Small Fire BaseA 2002 Sensitivity Scenarios
(12-km CMAQ results)
• removed all small fires
– less than 100 acres woodland and
– less than 300 acres grassland
– to determine whether small fires have significant visibility
impacts.
• The small fire removed case is compared CMAQ
36km BaseA 2002 (Base case minus sensitivity case)
to show the effect of small fires.
Overview of 2005 RMC Modeling Activities
Small Fire BaseA 2002 Sensitivity Scenarios
(12-km CMAQ results)
July, 2002 Nov, 2002
Overview of 2005 RMC Modeling Activities
Small Fire BaseA 2002 Sensitivity Scenarios
(12-km CMAQ results)
Overview of 2005 RMC Modeling Activities
Conclusions
Overview of 2005 RMC Modeling Activities
Conclusions
• Model performance does not appear to benefit significantly
from using the finer-resolution grid for modeling the lower
concentrations of PM2.5 that typically occur in the Class I
areas.
• We do not recommend the routine application of additional
12-km modeling as part of the WRAP regional haze
planning effort, due to the substantially higher resources
and costs associated with performing high-resolution
modeling.
• The 2002 model results are significantly improved
compared to results from the Section 309 modeling that
was performed for calendar year 1996.
Overview of 2005 RMC Modeling Activities
Conclusions
• The CMAQ and/or the CAMx 36-km modeling can be
used, in combination with the RRF approach, to evaluate
the benefits of emissions reduction strategies for all PM
species other than CM, in order to project visibility
changes at Class I areas for regional haze planning
purposes.
• Both CMAQ and CAMx are acceptable for visibility
modeling, and the choice of model should be based in part
on factors other than model performance, such as computer
run times, disk storage requirements, and source
apportionment and/or sensitivity analysis needs.
