Sensitivity Analysis of WeekdayWeekend Differences in Photochemical Air Pollution

1
Sensitivity Analysis of Weekday/Weekend
Differences in Photochemical Air Pollution

• Laurent Vuilleumier,
• Lawrence Berkeley National Laboratory

2
Collaborators

• Robert A. Harley
• University of California, Berkeley
• Nancy J. Brown
• Lawrence Berkeley National Laboratory

3
Presentation Outline

• Motivations
• The weekend effect What is it? Why choosing it?
• What can sensitivity analysis bring?
• Model description
• Emissions.
• Results
• Is it possible to model the weekend effect?
• Using sensitivity analysis to understand the
  weekend effect.
• Accuracy of sensitivity analysis prediction.

4
Motivations

• Despite lower emissions and concentration of
  NOx, VOC (volatile organic compound) and CO on
  weekend, average ozone concentrations are higher
  on weekend than on weekdays in many California
  urban areas.
• Provide the detailed understanding of weekend
  effect needed for designing strategies to attain
  National Ambient Air Quality Standard for ozone.
• Demonstrate the usefulness of sensitivity
  analysis for obtaining detailed understanding of
  model mechanism.
• Develop and evaluate sensitivity analysis
  techniques to be used in model assessment and
  uncertainty analysis.

5
The Model

• Lagrangian box model
• Equations for both the pollutant concentrations
  and their sensitivities to the model input
  parameters are solved in parallel.
• Model governing equation

6
Sensitivities

• Analytical derivation of sensitivities
• The model governing equation is derived with
  respect to input parameters under the assumption
  that they are constant within a simulation.

7
Solution with a Greens function method

• The solution of the sensitivity equation is given
  by
• The Greens function is computed as follows

8
Emissions

• Vehicle and stationary source emissions are
• treated separately.
• The pollutant concentration changes due to
• emissions are computed with
• Emphasis was put on vehicle emissions when
• analyzing weekday/weekend differences.

Emission time profile
Mixing height
Emission fluxes (g m-2 day-1)
9
Emission Fluxes
A published emission inventory for California is
used.

• Alameda County, CA, emission fluxes are used.
  Vehicle emissions fluxes for weekdays, Saturday
  and Sunday are different. For stationary sources,
  the same value is used for weekdays, Saturday and
  Sunday.

10
Vehicle Emission Time Profile

• Time profiles are derived from traffic intensity
  measurements at 3 sites.
• The sum of two gaussians is fitted to the data.
• The average site-to-site variability is used as
  a measure of the error.

11
Site to site variability

• Data from the 3 sites were averaged. The
  residuals between the individual site
  measurements and the averages were used to
  estimate the error.

12
Chemical mechanism

• Used two published mechanisms
• The Carbon Bond IV mechanism (CB4) and the
  condensed version of the surrogate species
  mechanism of Lurmann, Carter and Coyner (LCC).
• The chemical mechanism is the only part that
  allows influence of one species on another

Concentration vector
Parameter vector
13
Tropospheric ozone production
14
Dilution and Mixing Height

• Due to surface air heating, the height of the
  well-mixed layer above ground varies during the
  day.
• Resulting in dilution, entrainment of pollutants
  aloft, and changing the volume of air in which
  the pollutants are emitted.
• A mixing height profile was produced with data
  from acoustic sounding in Claremont, CA.

15
Results

• Simulations conducted with weekend conditions
  resulted in ozone concentrations about 20 higher
  than simulations with weekday conditions.

A
16
Ozone concentration (ppm) vs. time (hour) for
weekdays, Saturday and Sunday
Environmental Energy Technologies Division
16
17
Results

• Ozone sensitivities to emission fluxes and
  initial conditions are typical of a VOC-limited
  system.

B
18
Ozone sensitivities (ppm) to emissions and
initial concentrations vs. time (hour)
Environmental Energy Technologies Division
18
19
Results

• Sensitivity analysis estimation of the
  differences in peak ozone concentrations between
  weekdays and weekend are of correct sign and
  correct order-of-magnitude. Sensitivity analysis
  shows that the decrease in NOx emissions on
  weekends is the main cause for the ozone
  concentration increase.

C
20
Sensitivity analysis predictions for
weekday/Saturday differences in O3
Environmental Energy Technologies Division
20
21
Sensitivity analysis predictions for
weekday/Sunday differences in O3
Environmental Energy Technologies Division
21
22
Validity of sensitivity analysis approximations

• Why is the sensitivity analysis approximation
  for the Sunday ozone concentration change ?
  60higher than the real prediction?
• Is it possible to resolve the discrepancy by
  computing higher orders?

23
First- and second-order sensitivity analysis
approximations

• First-order
• Second-order

24
Test of sensitivity approximations for O3 when
ESNO value is changed
Environmental Energy Technologies Division
24
25
To be or not to be VOC-limited?

• A change in model mechanism is often the reason
  for sensitivity analysis inaccuracies.
• The VOC to NOx ratio suggest that the model is
  less likely to be in VOC-limited mode on weekend
  days than on weekdays.
• Greens function analysis can reveal the
  mechanistic information.

26
Greens function relating O3(t) to NO (t)
for weekdays and Sunday
Environmental Energy Technologies Division
26
27
Conclusions

• It is possible to reproduce the weekend effect
  with a simulation.
• Sensitivity analysis showed the reason for these
  differences.
• Greens function analysis is a powerful tool to
  obtain a detailed picture of the model mechanism.