Border Air Quality Strategy Project

1
Border Air Quality Strategy Project 4 Progress
Report
Jason Su Michael Buzzelli
Department of Geography
University of British Columbia
June 26-27, 2006
2
Contents of Update

• Mapping Ammonia
• Street Canyon Study
• Seattle Land Use Regression
• Source Area Analysis

3
1. Mapping Ammonia

• Objective Build a surface model of ammonia
  concentrations in the Georgia Basin/Puget Sound
  Airshed.
• Emission source Primarily from agricultural land
  use
• Data source for analysis GVRD emission
  inventory, Statistics Canadas census of
  livestock counts, and Environment Canadas NH3
  monitoring project.

4
(No Transcript)
5
(No Transcript)
6
2. Street Canyon Study

• Definition Street canyons are micro-environments
  created by the existence of multi-storey
  buildings flanking both sides of a street.
• Objective Identify air quality. Air quality
  deteriorates in street canyons due to a trapping
  effect caused by buildings lining the road and
  micro-meteorological conditions which operate to
  concentrate pollutants, in particular vehicle
  emissions (Vardoulakis, Fisher, Pericleous,
  Gonzalez-Flesca 2003).

7
Street Canyon Measurements
Aspect Ratio (H /W) Average height of canyon
divided by the average spacing or width of a
street (Vardoulis, 2003 Grimmond Oke, 1999).
Percent Permeability Average proportion of
ground plan covered by built features (buildings,
roads, paved and other impervious areas). The
rest of the area is occupied by pervious cover
(green space, water and other natural surfaces).
Roughness (z0) A spatial distribution (in
metres) of the roughness elements in the street
canyon. From Grimmond Oke (1999), roughness can
be calculated as a function of the average height
of features in a canyon (H) and the empirical
coefficient derived from observation (f0). z0
(f0)(H) Where f0 (urban land surfaces) 0.10
f0 (field crops) 0.13 f0 (forests) 0.06
8
Street CanyonMeasurements
9
Street Canyon Sampling

• 13 samplers in the GVRD
• Measured
• Land use type
• Street length
• Aspect ratio
• Permeability
• Urban climate zone

10
GVRD Street Canyon Sites
11
Downtown Streets with street location
Downtown Footprints with building elevation
Downtown Streets and Footprints
What Can We Get From This Map?
12

• A series of segments with street width, building
  heights, street length
• We can estimate in GIS the Aspect Ratio and
  Roughness of a street canyon.

13
What Can We Do?

• Improve the current land use regression model?
• OR??? Some other method

14
3. Seattle Land Use Regression
15
MESA NO2 Dataset

• Pilot study for community-scale NO2 sampling
• Passive badge, two-week measurements

26 sampling locations
16
The SW regression found ADT2000, DENS_RD123,
RES750 to be the predominant variables for a
multivariate linear regression model
Predicted NO2 13.974 0.000002756ADT2000
17.594DENS_RD123 - 0.02045RES_LC_750 R2 0.685
17
(No Transcript)
18
Predicted vs Measured NO2
19
Three Models Used to Extend Analysis to SMSA

• Variables
• DENS_RD123 Density of all roads within 300m
  buffer (km/ha)
• ATD_2000 Density of traffic (PSRC EMME model)
  within a 2000m search distance (vehicles/day/ha)
• RES750 Size of residential land cover within a
  specified buffer radius (ha)

• Models
• Used the above three variables but RES750 used
  land cover data
• Used the above three variables but RES750 used
  property assessment data
• Only use the previous two variables, no
  residential land use data being used

20

• Seattle
• NO2
• Predicted

21
241,123 bronchiolitis locations
22
4. Source area analysis

• We want to be able to
• Compare native SA analysis within our GIS to
  ensure were getting the same results
• Compare land use regression (based on simple
  buffering) with a more established meteorological
  appraoch
• For example, with the sampled traffic-NO2 data.
• Possibly/hopefully refine SA analysis with
  topography
• Ultimately, build time AND space resolved
  exposures with a SA model for health effects
  modelling (any pollutant, all sources)
• ie. Spatiotemporal cohort exposures and health
  effects models, rather than cross-sectional
  analysis

23
Source Area Analysis Mechanism
Source area model
Meso-scale model

• 1). Estimate hourly concentration hourly
  emission / (areamix hgt)
• 2). Calculate daily concentration sum hourly
  conc in a day
• 3). Calculate seasonally concentration average
  daily conc.

24
Comparing with Monit. Stn Data daily
averaged CO
R20.237 and 0.802, respectively, with and
without an intercept for the regression.
25
Comparing with Monit. Stn Data daily
averaged NO2
R20.427 and 0.897, respectively, with and
without an intercept for the regression.
26
Comparing with Monit. Stn Data daily
averaged NOX
R20.202 and 0.493, respectively, with and
without an intercept for the regression.
27
Comparing with Monit. Stn Data daily
averaged CO at stn 2
R20.68 and 0.91, respectively, with and without
an intercept for the regression.
28
Comparing with Monit. Stn Data daily
averaged NO2 at stn 2
R20.56 and 0.96, respectively, with and without
an intercept for the regression.
29
Comparing with Monit. Stn Data daily
averaged NOX at stn 2
R20.38 and 0.73, respectively, with and without
an intercept for the regression.
30
Interpolation Atmospheric Parameters to the 116
Samplers

• Including wind speed, wind direction, cloud,
  stability class and mixing height
• Minimum search distance 5, 10, 15 or 20km
• The weights are a decreasing function of distance

31
Hourly Wedge Shapes of the 116 Samplers

• 5-day hourly wedges at the 116 samplers

32
Comparing with Sampler Measurements
Which one did not work Meteorological
interpolation? Emission grid? Sampling (5
days)?
33
Wedge Radius Histogram interpolation issue
34
Comparing with Major Road Buffers part 1
interpolation issue
Buffer distance 1000m interval to 10,000m
35
Comparing with Major Road Buffers part 1
(cont.) interpolation issue
36
Comparing with Major Road Buffers part 2
interpolation issue
37
Conclusion 1

• Radius of a wedge from interpolation might need
  to be adjusted.

38
Comparing with Hourly Circular Buffers of Radius
3000m - emission surface issue
Table Correlation matrix of the 5 day average
1 Extracted from an emission surface hourly
pollution without considering the volume data
(emission). 2 Extracted from an emission surface
hourly pollution with volume info being
considered (emission/volume). Correlation is
significant at the 0.01 level (2-tailed).
39
Comparing with Seasonal Circular Buffers of
Radius 3000m - emission surface issue
Table Correlation coefficients between sampler
measured and seasonally estimated pollution.
1 Single buffer of radius 3000m for each sampler
reflects the seasonal average of pollution.
Correlation is significant at the 0.01 level
(2-tailed).
40
Conclusion 2

• Emission surface adequate?

41
Issues 3

• Testing period used (Feb 24-28) not
  representative of the whole measurement period
  (Feb24-March14)?

42
Other issues

• When comparing with monitoring data, we were
  aggregating both hourly monitoring data and
  hourly pollution estimations however, when
  comparing with the 116 samplers, the samplers
  readings and the hourly emissions did not follow
  the same aggregation procedure.
• If we do use the variables as Henderson and
  Brauer did, should we still use some atmospheric
  parameters such as mixing height and volume?