Douw Steyn

1
Topic 5.2 Case Study International Graduate
Institute on Modelling Environmental Space-Time
ProcessesUniversity of Washington, July 9 - 13,
2007
Networks from another perspective OR Well,
thats what Jim says

• Douw Steyn
• Department of Earth and Ocean Sciences
• The University of British Columbia

2

• As air quality monitoring networks have
  proliferated over the past, so have doubts arisen
  as to where, when and what to monitor (A.C.
  Stern, 1976).
• Not much has changed in 30 years!

3
False Creek, early 1900s. Plumes from lumber
mills.
4
Regional haze over Burnaby, early 1990s.
5
The GVRD air quality monitoring network
6
T1 Robson Square
7
T2 Kitsilano
8
T2 Kitsilano
9
T9 Rocky Point Park
10
T24 Kensington Park
11
Voronoi Analysis of Fixed Monitoring Network
12

• 1) Siting criteria for individual stations.
• 2) Overall Network structure.
• For individual stations, must consider
• The area that contributes significantly to
  concentrations measured at the station
• The area that the monitoring station can be said
  to represent

13
For the entire network, must consider the
following network objectives

• Capture present conditions and spatio-temporal
  structures/trends
• Regulatory control
• Make short-term predictions
• Capture effects of land-use strategies
• Study dose-response relations
• Provide input data for process based air quality
  models
• (Munn, R.E., 1978 The design of air quality
  monitoring networks. University of Toronto, Pub.
  EE-7, 93p.)

14
Why monitor? JZ

• Lots of stuff about statistics, then address
  societal concerns politicians
• Then detect non-compliance with regulatory
  standards - regulators
• to reduce uncertainty about some aspect of the
  world - scientists

15
Why monitor? JZ (cont)

• to assess temporal trends
• are things getting worse?
• is climate changing?

16
Ozone and Emission Trends in LFV
Number of hourly ozone concentrations gt 82 ppb
(diamonds), Annual calculated VOC emissions
(solid line) Annual calculated NOx emissions
(dotted line) VOC to NOx emissions ratio (dashed
line)
17
(No Transcript)
18
(No Transcript)
19
Study of local meteorology during ozone episodes
What are mesoscale wind regimes in LFV when peak
hourly ozone exceeds NAAQO of 82 ppb? What
synoptic flow regimes occur during these
mesoscale wind regimes? Have emissions
reductions over scale of decades changed ozone
patterns for different meteorological
regimes? Has population growth in the area
significantly changed the spatial distribution of
precursor emissions and hence the resulting ozone
fields?
20
Study Methods
Episode selection criteria At least 1
station in LFV has O3gt82 ppb, AND low (lt2.5
mm) rainfall at YVR, AND temperature at YXX gt
24.7 deg C. Use statistical clustering
techniques to find common circulation
regimes. For each regime, study associated
meso-scale wind fields, MSLP patterns and ozone
footprint.
21
Study Data Sets
Hourly wind speed and direction data from YVR and
YXX (1984-2003), and BLI (1990-2003). YXX daily
maximum temperature and YVR daily total
precipitation. Hourly ozone data from GVRD
network at 34 different monitoring locations
within the LFV. 135 ozone exceedance days meet
this criterion in 1984-2003. Hourly ozone data
from US EPA for lone measuring station in Whatcom
Co. Wa.
22
Plume Centroids
To see how emissions reductions and changes in
emission patterns may have altered ozone plume on
exceedance days, calculate East-West (Xc) and
North-South (Yc) component of centre of mass of
observed ozone plume
To see shift in plume, use multiple linear
regression with mean wind speed (u or v) and
number of days since April 1, 1984 (t)
23
We analyze only ozone plumes above background
levels of 32 or 25 ppb. For all clusters and
for both background levels, the Xc linear
temporal dependence is always positive (centre of
mass has moved eastward). At a 95 confidence
level, only for clusters I, II and III is the
change significant, independent of background
concentration. For Yc, there does not appear
to be a consistent trend.
24
Temporal Variability by Cluster
A Percent of summer days which are
fair-weather days. B Percent of fair-weather
days that are exceedance days. (cluster I .
pluses, II . Crosses III . triangles, IV .
squares)
25
Conclusions
Statistical clustering used to relate
spatio-temporal ozone distributions to prevailing
meteorological conditions. Observations over
24-hour period used to capture inter-dependence
of precursor advection on ozone
production. Classification identifies 4 common
circulation patterns. Coincident with large drop
in precursor emissions, proportion of
fair-weather days which have ozone exceedances
has decreased for all regimes.
26
Conclusions (cont.)
Plume center of mass has shifted eastward over
time Likely network not capturing significant
portions of ozone plume Voronoi diagram of
network shows station bias with more stations
closer to city and fewer away from valley axis A
complete description of interplay between
emissions, ozone and local circulations regimes
will require use of Eulerian grid modeling
presently underway (see next slide).
27
Modeled Ozone Plume
28

• Then, hidden deep in the text is implying that
  the optimum design must be regularly revisited
• This is called network review. It is conducted
  about every decade, and is presently underway for
  GVRD network.
• Define Network Objectives
• Analyze data from individual stations
• Analyze network structure (add/remove/move
  stations)

29
(No Transcript)
30
(No Transcript)
31
Correlation matrix for summertime ozone at all
GVRD stations
32
Clusters of GVRD stations based on summertime
ozone correlation analysis
33
First attempts at redesign of GVRD monitoring
network, based on ozone data
34
(No Transcript)
35
(No Transcript)
36
(No Transcript)