development of an ozone screening tool for the midwest alexander cohan 1, scott leopold 2, greg...
TRANSCRIPT
![Page 1: Development of an Ozone Screening Tool for the Midwest Alexander Cohan 1, Scott Leopold 2, Greg Yarwood 3, Kirk Baker 4 1 LADCO 2 Illinois EPA 3 ENVIRON](https://reader036.vdocuments.net/reader036/viewer/2022081515/56649cca5503460f94991e66/html5/thumbnails/1.jpg)
Development of an Ozone Screening Tool for the Midwest
Alexander Cohan1, Scott Leopold2, Greg Yarwood3, Kirk Baker4
1LADCO2Illinois EPA3ENVIRON4U.S. EPA
10/30/13
![Page 2: Development of an Ozone Screening Tool for the Midwest Alexander Cohan 1, Scott Leopold 2, Greg Yarwood 3, Kirk Baker 4 1 LADCO 2 Illinois EPA 3 ENVIRON](https://reader036.vdocuments.net/reader036/viewer/2022081515/56649cca5503460f94991e66/html5/thumbnails/2.jpg)
Motivation• July 2010 - Sierra Club petitions EPA to designate
air quality models for PSD permitting• January 2012 – EPA grants Sierra Club’s petition
and commits to updating the Guideline on Air Quality Models (Appendix W)
• Regional photochemical modeling is best science for addressing ozone impacts, but computational intensive and impractical for routine permitting
• Regulators would like an easy to use screening tool to assess the ozone impact of stationary sources for PSD permitting applications
![Page 3: Development of an Ozone Screening Tool for the Midwest Alexander Cohan 1, Scott Leopold 2, Greg Yarwood 3, Kirk Baker 4 1 LADCO 2 Illinois EPA 3 ENVIRON](https://reader036.vdocuments.net/reader036/viewer/2022081515/56649cca5503460f94991e66/html5/thumbnails/3.jpg)
Reduced Form Models• Use regional photochemical model results to
develop a simplified localized framework – Equivalency Ratio (Margaret McCourtney, MPCA)– Interpollutant Trading Ratios (James Boylan,
Georgia EPD)– Response Surface Model (Carey Jang, EPA)– Parametric Model (Greg Yarwood, ENVIRON)
![Page 4: Development of an Ozone Screening Tool for the Midwest Alexander Cohan 1, Scott Leopold 2, Greg Yarwood 3, Kirk Baker 4 1 LADCO 2 Illinois EPA 3 ENVIRON](https://reader036.vdocuments.net/reader036/viewer/2022081515/56649cca5503460f94991e66/html5/thumbnails/4.jpg)
Background• Parametric Model (Yarwood, 2011)– Screening tool developed for Sydney• 3 Km CAMx higher-order direct decoupled method
(HDDM) simulations of the summer • Assumptions:
– Ground source– Located at center of emissions by mass
Yarwood, G., Scorgie, Y., Agapides, N., Tai, E., Karamchandani, P., Bawden, K., Spencer, J., Trieu, T, 2011. A screening method for ozone impacts of new sources based on high-order sensitivity analysis of CAMx simulations for Sydney. Proceedings, 10th Annual CMAS Conference, Chapel Hill, NC.
![Page 5: Development of an Ozone Screening Tool for the Midwest Alexander Cohan 1, Scott Leopold 2, Greg Yarwood 3, Kirk Baker 4 1 LADCO 2 Illinois EPA 3 ENVIRON](https://reader036.vdocuments.net/reader036/viewer/2022081515/56649cca5503460f94991e66/html5/thumbnails/5.jpg)
Background
Base Emission Inventory
Single Point Emissions
CAMx
CAMx
Base Met
Base Ozone
Ozone W/ Point Source
HDDM Ozone Sensitivities
–BF Ozone
Impact
HDDM Ozone Impact+
2nd order Taylor series
€
∇ O3[ ] = ∇E
NOX
∂ O3[ ]
∂ENOX
+∇EVOC
∂ O3[ ]
∂EVOC
+12∇E
NOX
2∂ 2 O
3[ ]∂E
NOX
2+∇E
VOC∇E
NOX
∂ 2 O3[ ]
∂EVOC∂E
NOX
+∇EVOC
2∂ 2 O
3[ ]∂E
VOC
2
⎛
⎝ ⎜
⎞
⎠ ⎟
2nd Order Taylor Series
![Page 6: Development of an Ozone Screening Tool for the Midwest Alexander Cohan 1, Scott Leopold 2, Greg Yarwood 3, Kirk Baker 4 1 LADCO 2 Illinois EPA 3 ENVIRON](https://reader036.vdocuments.net/reader036/viewer/2022081515/56649cca5503460f94991e66/html5/thumbnails/6.jpg)
Problem Statement• Question: How do ozone sensitivities to
emission rates vary with emission rate and stack characteristics?
• Approach: Use multiple CAMx HDDM simulations of individual point sources to train a statistical model to empirically relate
€
S = f ENOX
,EVOC
,SH ,∇x,∇y( )
![Page 7: Development of an Ozone Screening Tool for the Midwest Alexander Cohan 1, Scott Leopold 2, Greg Yarwood 3, Kirk Baker 4 1 LADCO 2 Illinois EPA 3 ENVIRON](https://reader036.vdocuments.net/reader036/viewer/2022081515/56649cca5503460f94991e66/html5/thumbnails/7.jpg)
Statistical Models• Potential statistical models– CART– Neural network– Kriging– Fixed-Effects– Response Surface Model
• Fixed-Effects (FE) – Represents observed quantities in terms of
explanatory variables that are treated as if the quantities were non-random
– Transparent and easy to implement
![Page 8: Development of an Ozone Screening Tool for the Midwest Alexander Cohan 1, Scott Leopold 2, Greg Yarwood 3, Kirk Baker 4 1 LADCO 2 Illinois EPA 3 ENVIRON](https://reader036.vdocuments.net/reader036/viewer/2022081515/56649cca5503460f94991e66/html5/thumbnails/8.jpg)
Methodology• A linear parametric fit will not work!• Sensitivities behave like Gaussian plume
Seinfield & Pandis (1998)
€
S = f ENOX
,EVOC
,SH ,∇x,∇y( ) = α ENOX
,EVOC( )β SH ,∇x,∇y( )
€
β = 1eP1 +P2∇x
2 +P3∇y2 +P4SH
2
![Page 9: Development of an Ozone Screening Tool for the Midwest Alexander Cohan 1, Scott Leopold 2, Greg Yarwood 3, Kirk Baker 4 1 LADCO 2 Illinois EPA 3 ENVIRON](https://reader036.vdocuments.net/reader036/viewer/2022081515/56649cca5503460f94991e66/html5/thumbnails/9.jpg)
Methodology– Sources to the left or right might behave
differently…
– Use binomial expansion for α
– Fit parametric equation at each grid cell
– Solve for P1…P11 using R to minimize mean square error
€
S = f ENOX
,EVOC
,SH ,∇x + ,∇y+ ,∇x − ,∇y−
( )
€
∇x + = ∇x×H ∇x( )
€
∇x − = ∇x×H −∇x( )
€
∇y+ = ∇y×H ∇y( )
€
∇y− = ∇y×H −∇y( )
€
S =P
1E
NOX+P
2E
VOC+P
3E
NOX
2 +P4E
NOXE
VOC+P
5E
VOC
2
eP6 +P7 ∇x +
( )2
+P8 ∇x −( )
2+P9 ∇y +
( )2
+P10 ∇y −( )
2+P11SH
2
€
α ENOX
,EVOC( ) = P
1E
NOX+P
2E
VOC+P
3E
NOX
2 +P4E
NOXE
VOC+P
5E
VOC
2
![Page 10: Development of an Ozone Screening Tool for the Midwest Alexander Cohan 1, Scott Leopold 2, Greg Yarwood 3, Kirk Baker 4 1 LADCO 2 Illinois EPA 3 ENVIRON](https://reader036.vdocuments.net/reader036/viewer/2022081515/56649cca5503460f94991e66/html5/thumbnails/10.jpg)
Methodology• Proof of concept conducted for test case in
Illinois• Based on LADCO 2007 Modeling platform• 4 km CAMx HDDM modeling of summer 2007
![Page 11: Development of an Ozone Screening Tool for the Midwest Alexander Cohan 1, Scott Leopold 2, Greg Yarwood 3, Kirk Baker 4 1 LADCO 2 Illinois EPA 3 ENVIRON](https://reader036.vdocuments.net/reader036/viewer/2022081515/56649cca5503460f94991e66/html5/thumbnails/11.jpg)
• 20 hypothetical point sources modeled with HDDM– Stack diameter is linearly related to stack height
(SH)– Exit velocity of 44 ft/s– Exit temperature 70 °f
Methodology
![Page 12: Development of an Ozone Screening Tool for the Midwest Alexander Cohan 1, Scott Leopold 2, Greg Yarwood 3, Kirk Baker 4 1 LADCO 2 Illinois EPA 3 ENVIRON](https://reader036.vdocuments.net/reader036/viewer/2022081515/56649cca5503460f94991e66/html5/thumbnails/12.jpg)
Methodology
FE Model
R Model FitP1 … P11
HDDM Ozone Sensitivities PT01
ENOx , EVOC , x , y, SH
HDDM Ozone Sensitivities PT20
HDDM Ozone Sensitivities PTXX
ENOx , EVOC
FE Ozone Impact
2nd Order Taylor Series
2nd Order Taylor Series
HDDM Ozone Impact
FE Approach
HDDM Approach
FE Ozone Sensitivities
![Page 13: Development of an Ozone Screening Tool for the Midwest Alexander Cohan 1, Scott Leopold 2, Greg Yarwood 3, Kirk Baker 4 1 LADCO 2 Illinois EPA 3 ENVIRON](https://reader036.vdocuments.net/reader036/viewer/2022081515/56649cca5503460f94991e66/html5/thumbnails/13.jpg)
Methodology
FE Model
R Model FitP1 … P11
HDDM Ozone Sensitivities PT01
ENOx , EVOC , x , y, SH
HDDM Ozone Sensitivities PT20
HDDM Ozone Sensitivities PTXX
ENOx , EVOC
FE Ozone Impact
2nd Order Taylor Series
2nd Order Taylor Series
HDDM Ozone Impact
FE Approach
HDDM Approach
FE Ozone Sensitivities
Screening Tool
![Page 14: Development of an Ozone Screening Tool for the Midwest Alexander Cohan 1, Scott Leopold 2, Greg Yarwood 3, Kirk Baker 4 1 LADCO 2 Illinois EPA 3 ENVIRON](https://reader036.vdocuments.net/reader036/viewer/2022081515/56649cca5503460f94991e66/html5/thumbnails/14.jpg)
Sensitivities from CAMx HDDM vs. FE
CAMx HDDM
FE M
odel
€
∂ O3[ ]
∂ENOX
€
∂ O3[ ]
∂EVOC
€
∂2 O3[ ]
∂ENOX
2
€
∂2 O3[ ]
∂EVOC
2
€
∂2 O3[ ]
∂EVOC∂E
NOX
![Page 15: Development of an Ozone Screening Tool for the Midwest Alexander Cohan 1, Scott Leopold 2, Greg Yarwood 3, Kirk Baker 4 1 LADCO 2 Illinois EPA 3 ENVIRON](https://reader036.vdocuments.net/reader036/viewer/2022081515/56649cca5503460f94991e66/html5/thumbnails/15.jpg)
ENOx = 718 tpy, EVOC = 61 tpy, SH = 120 ft
Training Data Point 2
![Page 16: Development of an Ozone Screening Tool for the Midwest Alexander Cohan 1, Scott Leopold 2, Greg Yarwood 3, Kirk Baker 4 1 LADCO 2 Illinois EPA 3 ENVIRON](https://reader036.vdocuments.net/reader036/viewer/2022081515/56649cca5503460f94991e66/html5/thumbnails/16.jpg)
Training Data Point 5ENOx = 80 tpy, EVOC = 197 tpy, SH = 48 ft
![Page 17: Development of an Ozone Screening Tool for the Midwest Alexander Cohan 1, Scott Leopold 2, Greg Yarwood 3, Kirk Baker 4 1 LADCO 2 Illinois EPA 3 ENVIRON](https://reader036.vdocuments.net/reader036/viewer/2022081515/56649cca5503460f94991e66/html5/thumbnails/17.jpg)
Training Data Point 6ENOx = 943 tpy, EVOC = 70 tpy, SH = 454 ft
![Page 18: Development of an Ozone Screening Tool for the Midwest Alexander Cohan 1, Scott Leopold 2, Greg Yarwood 3, Kirk Baker 4 1 LADCO 2 Illinois EPA 3 ENVIRON](https://reader036.vdocuments.net/reader036/viewer/2022081515/56649cca5503460f94991e66/html5/thumbnails/18.jpg)
Training Data Point 9ENOx = 610 tpy, EVOC = 25 tpy, SH = 194 ft
![Page 19: Development of an Ozone Screening Tool for the Midwest Alexander Cohan 1, Scott Leopold 2, Greg Yarwood 3, Kirk Baker 4 1 LADCO 2 Illinois EPA 3 ENVIRON](https://reader036.vdocuments.net/reader036/viewer/2022081515/56649cca5503460f94991e66/html5/thumbnails/19.jpg)
Training Data Point 18ENOx = 380 tpy, EVOC = 51 tpy, SH = 268 ft
![Page 20: Development of an Ozone Screening Tool for the Midwest Alexander Cohan 1, Scott Leopold 2, Greg Yarwood 3, Kirk Baker 4 1 LADCO 2 Illinois EPA 3 ENVIRON](https://reader036.vdocuments.net/reader036/viewer/2022081515/56649cca5503460f94991e66/html5/thumbnails/20.jpg)
Testing Point 1ENOx = 143 tpy, EVOC = 143 tpy, SH = 71 ft
![Page 21: Development of an Ozone Screening Tool for the Midwest Alexander Cohan 1, Scott Leopold 2, Greg Yarwood 3, Kirk Baker 4 1 LADCO 2 Illinois EPA 3 ENVIRON](https://reader036.vdocuments.net/reader036/viewer/2022081515/56649cca5503460f94991e66/html5/thumbnails/21.jpg)
Testing Point 2ENOx = 6 tpy, EVOC = 292 tpy, SH = 177 ft
![Page 22: Development of an Ozone Screening Tool for the Midwest Alexander Cohan 1, Scott Leopold 2, Greg Yarwood 3, Kirk Baker 4 1 LADCO 2 Illinois EPA 3 ENVIRON](https://reader036.vdocuments.net/reader036/viewer/2022081515/56649cca5503460f94991e66/html5/thumbnails/22.jpg)
Testing Point 3ENOx = 165 tpy, EVOC = 192 tpy, SH = 83 ft
![Page 23: Development of an Ozone Screening Tool for the Midwest Alexander Cohan 1, Scott Leopold 2, Greg Yarwood 3, Kirk Baker 4 1 LADCO 2 Illinois EPA 3 ENVIRON](https://reader036.vdocuments.net/reader036/viewer/2022081515/56649cca5503460f94991e66/html5/thumbnails/23.jpg)
Summary• FE matches HDDM training data well• FE/HDDM matches BF in magnitude and
extent of impact, however BF produces a higher peak impact
• FE Model would benefit from additional training data
![Page 24: Development of an Ozone Screening Tool for the Midwest Alexander Cohan 1, Scott Leopold 2, Greg Yarwood 3, Kirk Baker 4 1 LADCO 2 Illinois EPA 3 ENVIRON](https://reader036.vdocuments.net/reader036/viewer/2022081515/56649cca5503460f94991e66/html5/thumbnails/24.jpg)
Future Work• Investigate differences in peak impact
predicted from BF and HDDM• Examine different VOC profiles• Explore other statistical models• Apply methodology to develop PM screening
tool
![Page 25: Development of an Ozone Screening Tool for the Midwest Alexander Cohan 1, Scott Leopold 2, Greg Yarwood 3, Kirk Baker 4 1 LADCO 2 Illinois EPA 3 ENVIRON](https://reader036.vdocuments.net/reader036/viewer/2022081515/56649cca5503460f94991e66/html5/thumbnails/25.jpg)
Acknowledgments
• Scott Leopold – IL EPA• Greg Yarwood – ENVIRON• Kirk Baker – OAQPS• Margaret McCourtney – MN PCA• Randall Robinson – EPA Region 5• Mark Derf – IN DEM
![Page 26: Development of an Ozone Screening Tool for the Midwest Alexander Cohan 1, Scott Leopold 2, Greg Yarwood 3, Kirk Baker 4 1 LADCO 2 Illinois EPA 3 ENVIRON](https://reader036.vdocuments.net/reader036/viewer/2022081515/56649cca5503460f94991e66/html5/thumbnails/26.jpg)
20 HDDM Modeled Point Sources
![Page 27: Development of an Ozone Screening Tool for the Midwest Alexander Cohan 1, Scott Leopold 2, Greg Yarwood 3, Kirk Baker 4 1 LADCO 2 Illinois EPA 3 ENVIRON](https://reader036.vdocuments.net/reader036/viewer/2022081515/56649cca5503460f94991e66/html5/thumbnails/27.jpg)
LADCO 2007 Modeling Platform• Ozone Performance