INCORPORATING UNCERTAINTY INTO AIR QUALITY MODELING

1
INCORPORATING UNCERTAINTY INTO AIR QUALITY
MODELING PLANNING A CASE STUDY FOR GEORGIA

• 7th Annual CMAS Conference
• 6-8th October, 2008

Antara Digar, Daniel S. Cohan, Dennis Cox, Wei
Zhou Rice University Maudood Khan, James
Boylan Georgia Environmental Protection Division
2
Introducing the Project
JAMES BOYLAN MICHELLE S. BERGIN
DANIEL S. COHAN (PI) DENNIS COX ANTARA DIGAR
MICHELLE BELL
ROBYN WILSON
3
Why is this project important ? (to be removed)
4
Background Objective
Measure Control Emission
NOx
VOC
SOx
NH3
PM
PM2.5
O3
Non-attainment In U.S.
Controlling Multiple Pollutants How Much to
Control ? Which Measure is Effective?
Scientists Air Quality Modelers have come up
with techniques to estimate Sensitivity of O3 and
PM2.5 to their precursor emissions
But in reality the model inputs are sometimes
uncertain
GOAL Estimate this Uncertainty
Uncertainty in Model Input causes Uncertainty
in O3 PM2.5 Sensitivities
5
How to achieve this goal ? (to be removed)
6
Model Used
Achieving the Goal
CMAQ - High-order Decoupled Direct Method

• H- High-order sensitivity analysis
• N- Nonlinear relationship between secondary
  pollutants and its precursor emission
• N- Non-liner sensitivity model can be used to
  determine the impact of uncertain Emission
  inventory, Photochemical rate constants,
  Deposition velocities on O3 and PM2.5 sensitivity
  to their precursor emission control

C
A
CA
CB
B
E
-?E
E denotes precursor emission C denotes
secondary pollutant concentration

• HDDM determines slope at any point by
  calculating the local derivative at that point

Source Hakami et. al. 2003 Cohan et. al. 2005
7
Introducing Uncertainty
Effect of Uncertain Input Parameters
Ozone
Modeled value
Actual value
Actual value
Modeled value
E
E
-D? EA
Effect of Control Strategy (Emission Reduction)
High-or Self Sensitivity
Sensitivity to parameter j if j is uncertain
Cross Sensitivity
Sensitivity to parameter j if k ? j is uncertain
Source Cohan et. al., 2005
8
HDDM in Selection of Control Strategy
Control measures

• reduction in regional emission (NOx, VOC, NH3,
  etc.)
• Specific amount of reduction at power plant
  (NOx, SOx)

Pollutant Levels Exposure Metrics

• O3 at worst monitor
• O3 population exposure
• PM2.5 at worst monitor
• PM2.5 population exposure

• Uncertainty in emission inventory
• Uncertainty in reaction rate constants
• Uncertainty in deposition velocities

Uncertainties
9
Example Case
Control measures
Pollutant Levels Exposure Metrics

• reduction in regional NOx emission
• Specific amount of reduction at power plant

• O3 at worst monitor
• O3 at Atlanta
• PM2.5 at worst monitor
• PM2.5 population exposure

• Uncertainty in emission self/cross (NOx, VOC,
  etc.)
• Uncertainty in reaction rate constants
• Uncertainty in deposition velocities

Uncertainties
10
Our Approach
Sensitivity of O3 to precursor emission f(Ei,
Rj, Vdk, )
11
Methodology
Sensitivity of secondary pollutant to any
parameter j given both j and any other input
parameter k ? j is also uncertain
SURROGATE MODEL
CMAQ-HDDM
MONTE CARLO
Input Parameter

• Sensitivity estimated by CMAQ-HDDM

• PDFs for input parameters from literature

• Monte Carlo Sampling

• Develop output PDFs using Surrogate Model
• Characterize uncertainty in output sensitivity,
  S

Output Sensitivity
12
How ACCURATE IS OUR MODEL ? (to be removed)
13
Applying to Georgia A Case Study(May 30
June 06, 2009)
ALGA 12km domain
14
Accuracy of CMAQ-HDDM
Sensitivity of Ozone to NOx Emission
Impact of Uncertainty in ENOx
R2 gt 0.99 Limitation CMAQ-HDDM is
not yet capable of handling high-order PM
sensitivities, hence BF will be used for such
cases
(Self Sens)
Impact of Uncertainty in R(NO2 OH)
(Cross Sens)
HDDM
Brute Force
15
Uncertain Emission Inventory
EVOC
ENOX
ESOX

• First Scenario

ENH3
EPM
16
Case 1A Self sensitivity
Control measures
Pollutant Levels Exposure Metrics
Reduction in NOx emission

• Atlanta O3
• Scherer O3

NOx emission uncertain by 30
Uncertainties
17
If NOx emission is larger than expected, O3 _ENOx
generally increases but some locations have NOx
disbenefit
Impact of Uncertainty in ENOx
Sensitivity of O3 to Atlanta NOx
Sensitivity of O3 to Scherer NOx
18
Case 1B Cross Sensitivity
Control measures
Pollutant Levels Exposure Metrics
Reduction in VOC emission

• Atlanta O3
• Scherer O3

NOx emission uncertain by 30
Uncertainties
19
If ENOx is larger than expected, sensitivity of
O3 to EVOC is slightly increased
Impact of Uncertainty in ENOx
Sensitivity of O3 to Atlanta VOC
Sensitivity of O3 to Scherer VOC
20
Uncertain reaction Rate
HRVOCsNO3?products
HRVOCsO3?products
O3NO?NO2
NO2h??NOO
NO2NO3?N2O5
NO2OH?HNO3

• Second Scenario

HRVOCsOH?products
21
Case 2 Cross Sensitivity
Control measures
Pollutant Levels Exposure Metrics
Reduction in NOx emission

• Atlanta O3
• Scherer O3

R(NO2OH) uncertain by 30
Uncertainties
22
If R(NO2OH ? HNO3) is larger than expected,
sensitivity of O3 to ENOx decreases
Impact of Uncertainty in R(NO2OH)
Sensitivity of O3 to Atlanta NOx
Sensitivity of O3 to Scherer NOx
23
Preliminary Findings

• Uncertain NOx emissions inventory
• A larger NOx inventory generally increases the
  sensitivity of Ozone to ENOx, however some
  locations show NOx disbenefit
• A larger NOx inventory increases the sensitivity
  of Ozone to EVOC
• Uncertain Reaction Rate of HNO3 formation
• A larger rate than expected greatly decreases the
  Ozone sensitivity to ENOx

24
Take Home Message (to be removed)

• CMAQ-HDDM would be able to address issues
  like
• How uncertain is O3/PM2.5 sensitivity to
  precursor emissions when model inputs (primary
  pollutant emissions , photochemical reaction
  rates, deposition velocities, etc.) are uncertain
  ?
• What would be the benefits of a hypothetical
  reduction in primary pollutant concentration like
  NOx, SOx, VOC at worst monitor given the emission
  inventory is uncertain ?
• What would be the benefits of a percentage
  reduction in anthropogenic or biogenic emission
  like NOx, SOx, VOC at a given region when the
  rate constants for photochemical reactions are
  uncertain ?
• What would be the benefits of a percentage
  reduction in emission when the deposition
  velocities and/or meteorological conditions are
  uncertain ?

25
Overall Project Goal
An Optimum Control Strategy
ANALYSIS
OUTCOME

• Control Strategy that satisfies the 3 criteria
• Reduces multiple pollutants (air quality)
• Cost Effective (economic)
• Maximum health benefit (health)

Response of pollutant sensitivity to
uncertainty (CMAQ-HDDM)
Impact on pollutant level at worst monitor
air quality
Cost of Emission Control (Lit / AirControlNET /
CoST)
economic
Impact on Population Exposure Human Health
Impact on Population Exposure
Health Impacts Benefits of Emission
Control (BENMAP)
health
26
Future Plan of Action

• Estimate cost of control strategies
• Calculate health benefits for a given population
  exposure
• Interlink CMAQ-HDDM sensitivity output with
  health and cost assessment
• Select control strategy that reduces multiple
  pollutants (O3 and PM2.5) based on maximum health
  benefit and minimum cost of implementation

27
Acknowledgement

• U.S. EPA
• For funding our project
• GA EPD
• For providing emission data
• Byeong Kim for technical assistance
• CMAS

28
For further information updates of our project

• Contact antara_at_rice.edu
• Log on to http//uncertainty.rice.edu/

29
Brute Force vs. HDDM
Source Cohan et. al., 2005
Ozone
A

CA
?C
a1
B

CB
EVOC
-DeEA
EB
EA