The advent of the short term National Ambient Air Quality Standards (NAAQS) prompted modelers to reassess the common practices in dispersion modeling analyses. The probabilistic nature of the new short term standards also opens the door to alternative modeling techniques that are based on probability. One of these is the Monte Carlo technique that can be used to account for emission variability in permit modeling. Currently, it is assumed that a given emission unit is in operation at its maximum capacity every hour of the year. This assumption may be appropriate for facilities that operate at full capacity most of the time. However, in most cases, emission units operate at variable loads that produce variable emissions. Thus, assuming constant maximum emissions is overly conservative for facilities such as power plants that are not in operation all the time and which exhibit high concentrations during very short periods of time. Another element of conservatism in NAAQS demonstrations relates to combining predicted concentrations from the AMS/EPA Regulatory Model (AERMOD) with observed (monitored) background concentrations. Normally, some of the highest monitored observations are added to the AERMOD results yielding a very conservative combined concentration. A case study is presented to evaluate the use of alternative probabilistic methods to complement the shortcomings of current dispersion modeling practices. This case study includes the use of the Monte Carlo technique and the use of a reasonable background concentration to combine with the AERMOD predicted concentrations. The use of these methods is in harmony with the probabilistic nature of the NAAQS and can help demonstrate compliance through dispersion modeling analyses, while still being protective of the NAAQS.

1. Use of Probabilistic Statistical Techniques
in AERMOD Modeling Evaluations
A&WMA’s 108th Annual Conference & Exhibition –
Raleigh, NC
June 24, 2015
Sergio A. Guerra, Ph.D. - CPP, Inc.
Jesse Thé, Ph.D., P.Eng. - Lakes Environmental Software

2. Outline
• AERMOD’s Probabilistic Performance Evaluation
• Monte Carlo Statistical Technique
• Combining Modeled Results and Background
Concentrations
• Case Study Example

3. Model’s Accuracy
Appendix W: 9.1.2 Studies of Model Accuracy
a. A number of studies have been conducted to examine model accuracy,
particularly with respect to the reliability of short-term concentrations
required for ambient standard and increment evaluations. The results of
these studies are not surprising. Basically, they confirm what expert
atmospheric scientists have said for some time: (1) Models are more
reliable for estimating longer time-averaged concentrations than for
estimating short-term concentrations at specific locations; and (2)
the models are reasonably reliable in estimating the magnitude of
highest concentrations occurring sometime, somewhere within an
area. For example, errors in highest estimated concentrations of ± 10 to 40
percent are found to be typical, i.e., certainly well within the often quoted
factor-of-two accuracy that has long been recognized for these models.
However, estimates of concentrations that occur at a specific time and site,
are poorly correlated with actually observed concentrations and are much
less reliable.
• Bowne, N.E. and R.J. Londergan, 1983. Overview, Results, and Conclusions for the EPRI Plume Model Validation and Development
Project: Plains Site. EPRI EA–3074. Electric Power Research Institute, Palo Alto, CA.
• Moore, G.E., T.E. Stoeckenius and D.A. Stewart, 1982. A Survey of Statistical Measures of Model Performance and Accuracy for Several
Air Quality Models. Publication No. EPA–450/4–83–001. Office of Air Quality Planning & Standards, Research Triangle Park, NC.

4. Perfect Model
MONITORED CONCENTRATIONS
AERMODCONCENTRATIONS
100
1000
-
-

5. Monitored vs Modeled Data:
Paired in Time and Space
AERMOD performance evaluation of three coal-fired electrical generating units in Southwest Indiana
Kali D. Frost
Journal of the Air & Waste Management Association
Vol. 64, Iss. 3, 2014

6. SO2 Concentrations Paired in Time & Space
Probability analyses of combining background concentrations with model-predicted concentrations
Douglas R. Murray, Michael B. Newman
Journal of the Air & Waste Management Association
Vol. 64, Iss. 3, 2014

7. SO2 Concentrations Paired in Time Only
Probability analyses of combining background concentrations with model-predicted concentrations
Douglas R. Murray, Michael B. Newman
Journal of the Air & Waste Management Association
Vol. 64, Iss. 3, 2014

8. AERMOD’s Evaluation

9. Are We Using the Model Correctly?
Temporal matching is not justifiable
Perfect model AERMOD

10. Solutions to AERMOD’s Limitations
Advanced Modeling
Techniques
Traditional Modeling Technique
Variable emissions Use EMVAP to account for
variability
Assume continuous maximum
emissions
Background
Concentrations
Combine AERMOD’s
concentration with the 50th %
observed
Tier 1: Combine AERMOD’s
concentration with max. or design
value (e.g., 98th % observed for
SO2)
Tier 2: Combine predicted and
observed values based on
temporal matching (e.g., by
season or hour of day).

11. Monte Carlo Approach
• Pioneered by the Manhattan Project scientists in 1940’s
• Technique is widely used in science and industry
• EPA has approved this technique for risk assessments
• Used by EPA in the Guidance for 1-hour SO2
Nonattainment Area SIP Submissions (2014)

12. Emission Variability Processor
• Assuming fixed peak 1‐hour emissions on a continuous basis
will result in unrealistic modeled results
• Better approach is to assume a prescribed distribution of
emission rates
• EMVAP assigns emission rates at random over numerous
iterations
• The resulting distribution from EMVAP yields a more
representative approximation of actual impacts
• Incorporate transient and variable emissions in modeling
analysis
• EMVAP uses this information to develop alternative ways to
indicate modeled compliance using a range of emission rates
instead of just one value

13. Background Concentrations

14. Siting of Ambient Monitors
According to the Ambient Monitoring Guidelines for Prevention of Significant
Deterioration (PSD):
The existing monitoring data should be representative of three types of area:
1) The location(s) of maximum concentration increase from the proposed
source or modification;
2) The location(s) of the maximum air pollutant concentration from existing
sources; and
3) The location(s) of the maximum impact area, i.e., where the maximum
pollutant concentration would hypothetically occur based on the combined
effect of existing sources and the proposed source or modification. (EPA, 1987)
U.S. EPA. (1987). “Ambient Monitoring Guidelines for Prevention of Significant
Deterioration (PSD).”EPA‐450/4‐87‐007, Research Triangle Park, NC.

15. Exceptional Events
http://blogs.mprnews.org/updraft/2012/06/co_smoke_plume_now_visible_abo/

17. Exceptional Events

18. 24-hr PM2.5 Santa Fe, NM Airport
Background Concentration and Methods to Establish Background Concentrations in Modeling.
Presented at the Guideline on Air Quality Models: The Path Forward. Raleigh, NC, 2013.
Bruce Nicholson

19. Probability of Two Unusual Events
Happening at the Same Time

20. Combining 99th Percentile Pre and Bkg
(1-hr SO2)
99th percentile is 1st rank out of 100 days = 0.01
P(Pre ∩ Bkg) = P(Pre) * P(Bkg)
= (1-0.99) * (1-0.99)
= (0.01) * (0.01)
= 0.0001 = 1 / 10,000 days
Equivalent to one exceedance every 27 years!
= 99.99th percentile of the combined distribution

21. Proposed Approach to Combine Modeled
and Monitored Concentrations
• Combining the 99th (for 1-hr SO2) % monitored
concentration with the 99th % predicted
concentration is too conservative.
• A more reasonable approach is to use a
monitored value closer to the main distribution
(i.e., the median).
Evaluation of the SO2 and NOX offset ratio method to account for secondary PM2.5 formation
Sergio A. Guerra, Shannon R. Olsen, Jared J. Anderson
Journal of the Air & Waste Management Association
Vol. 64, Iss. 3, 2014

22. Combining 99th Pre and 50th Bkg
50th Percentile is 50th rank out of 100 days = 0.50
P(Pre ∩ Bkg) = P(Pre) * P(Bkg)
= (1-0.99) * (1-0.50)
= (0.01) * (0.50)
= 0.005 = 1 / 200 days
Equivalent to 1.8 exceedances every year
= 99.5th percentile of the combined distribution
Evaluation of the SO2 and NOX offset ratio method to account for secondary PM2.5 formation
Sergio A. Guerra, Shannon R. Olsen, Jared J. Anderson
Journal of the Air & Waste Management Association
Vol. 64, Iss. 3, 2014

23. Case Study: Three Cases Evaluated
1. Using AERMOD by assuming a constant
maximum emission rate (current modeling
practice)
2. Using AERMOD by assuming a variable
emission rate
3. Using EMVAP to account for emission
variability

25. Three Cases Used to Model the Power Plant
Input parameter Case 1 Case 2 Case 3
Description of
Dispersion
Modeling
Current
Modeling
Practices
AERMOD with
hourly emission
EMVAP
(500 iterations)
SO2 Emission rate
(g/s)
478.7
Actual hourly
emission rates
from CEMS
data
Bin1: 478.7
(5.0% time)
Bin 2: 228.7
(95% time)
Stack height (m) 122
Exit temperature
(degrees K)
416
Diameter (m) 5.2
Exit velocity (m/s) 23

26. Results of 1-hour SO2 Concentrations
Case 1
(µg/m3)
Case 2
(µg/m3)
Case 3
(µg/m3)
Description of
Dispersion
Modeling
Current
Modeling
Practices
AERMOD
with hourly
emission
EMVAP
(500
iterations)
H4H 229.9 78.6 179.3
Percent of
NAAQS
117% 40% 92%

27. St. Paul Park 436 Ambient Monitor Location

28. Positively Skewed Distribution
http://www.agilegeoscience.com

29. Histogram of 1-hr SO2 Observations
Innovative Dispersion Modeling Practices to Achieve a Reasonable Level of Conservatism in AERMOD Modeling Demonstrations.
Sergio A. Guerra
EM Magazine, December 2014.

30. Concentrations at Different Percentiles for the St. Paul
Park 436 monitor (2011-2013)
Percentile µg/m3
50th 2.6
60th 3.5
70th 5.2
80th 6.1
90th 9.6
95th 12.9
98th 20.1
99th 25.6
99.9th 69.5
99.99th 84.7
Max. 86.4

31. Case 3 with Three Different Background Values
Case 3 with
99th % Bkg
(µg/m3)
Case 3 with
50th % Bkg
(µg/m3)
179.3 179.3 179.3
Background 86.4 25.6 2.6
Total 265.7 204.9 181.9
Percent of NAAQS 135.6% 104.5% 92.8%

32. Conclusion
• Probabilistic standards provide a stringent level of protection
based on the likelihood of complying with the NAAQS
• AERMOD’s evaluations are based on the probability of a
maximum occurrence happening sometime and somewhere in
the modeling domain
• Probabilistic methods can be used to achieve more reasonable
results
• Use of EMVAP can help achieve more realistic concentrations
• Use of 50th % monitored concentration is statistically conservative
when pairing it with the 99th % predicted concentration
• Methods are :
• protective of the NAAQS,
• provide a reasonable level of conservatism,
• are in harmony with probabilistic nature of 1-hr standards
32

33. Advanced Model Input Analysis Solutions
• Emission Variability
Processor (EMVAP)
• Evaluation of
background
concentrations
EM Magazine, December 2014
Guerra, S.A. “Innovative Dispersion Modeling
Practices to Achieve a Reasonable Level
of Conservatism in AERMOD Modeling
Demonstrations.” EM Magazine, December 2014.

34. Sergio Guerra, PhD
sguerra@cppwind.com
Direct: + 970 360 6020
www.SergioAGuerra.com
www.cppwind.com @CPPWindExperts
Thank You!