Environmental Engineering Reference
In-Depth Information
performing meteorologically adjusted trend analyses; and evaluating emissions
control strategy efficacy.
2. Method
Extensive use was made of cluster analysis techniques. These algorithms identify
distinct groups of days sharing similar, recurring boundary layer characteristics. A
main goal was to identify conditions allowing 8-h ozone to build to the NAAQS
exceedance threshold (85 ppb). First, clustering of wind field measurements
identified distinct air flow patterns associated with different ozone source-receptor
relationships. Second, clustering of ozone level measurements identified distinct
ozone spatial distributions potentially associated with different meteorological
conditions. These complimentary clustering methods are commonly referred to as
circulation-to-environment
and
environment-to-circulation
approaches, respec-
tively (Ainslie and Steyn, 2007). We considered data from 9 recent years after
reformulated gasoline (RFG) was introduced to California. It is assumed that
emissions characteristics during the period were relatively constant, so that
patterns isolated by the clusterings largely reflect meteorological variability and
day-of-week emissions differences. Clustering a total of 1,656 days since RFG
introduction ensures the identified meteorological patterns are representative for
use in current and future air quality studies and policy development efforts.
The large size and meteorological complexity of Central California preclude a
direct clustering of measurements spanning the entire domain. Instead, independent
clusterings were performed for seven subdomains to characterize the regional
variability of meteorology and air quality within each. SFBA and SV are each
considered as individual subdomains. SJV was subdivided into northern, central,
and southern subdomains (N-SJV, C-SJV, and S-SJV, respectively). The remaining
subdomains are the Mountain Counties (MC) and Sequoia National Park area
(Sequoia). Direct clusterings of MC and Sequoia measurements were not instruc-
tive due to highly localized meteorology in these mountainous regions. Thus, their
air quality is evaluated in response to the clusterings for SV and SJV, respectively,
with which these elevated subdomains are contiguous and meteorologically coupled.
3. Results and Discussion
The wind field clustering identified four to seven weather patterns for each
subdomain. Such relatively small numbers of clusters are typical of air quality
meteorology studies. Forcing the clustering algorithm to yield larger numbers of
patterns tends to produce clusters that are too meteorologically similar to capture
distinct ozone source-receptor relationships and dispersion patterns. The more
