Extended Abstract (296K)2.6
Representativeness of two ozone episodes selected for modeling in the San Francisco Bay Area
David Fairley, Bay Area Air Quality Management District, San Francisco, CA; and P. T. Martien, S. T. Soong, and S. Tanrikulu
This study analyzes the representativeness of two ozone episodes selected for photochemical modeling in the San Francisco Bay Area (SFBA). For this purpose, a cluster analysis of recent SFBA ozone episodes was conducted. Cluster analysis revealed two major distinct types of ozone episodes in the region. One of the modeled periods (July 30-August 2, 2000) was captured during the comprehensive field program of the Central California Ozone Study (CCOS). But in part because all CCOS episodes fell within the same cluster, a second episode (July 11-12, 1999) was chosen from the other cluster during a period of routine measurement only. This study then examined the representativeness of these episodes in terms of meteorology, severity, extent, and timing.
The San Francisco Bay Area is home to approximately seven million people, five million automobiles, and several large industrial emissions sources, including electric power generating facilities and refineries. Its topography is complex. The SFBA is bordered by the Pacific Ocean in the west and Central Valley in the east. The Coast Range extending in the north south-direction has gaps in the SFBA allowing air flow between the Pacific Ocean and Central Valley. The remaining portions of the Coast Range have a number of hills and valleys of various sizes. A large temperature gradient develops in most afternoons of the summer between the Ocean and Central Valley.
Ozone concentrations in the SFBA occasionally exceed the federal 1-hour standard. However, ozone has declined significantly in much of the SFBA since the 1980s. A shift in the locations of ozone exceedances is also evident. In the 1980s, peak ozone typically occurred near San Jose, in the south of the Bay Area. However, in recent years, this area has had few exceedances; high ozone has concentrated almost exclusively in the Livermore Valley, in the eastern Bay Area. Forecasting of high ozone in the SFBA is challenging, both because it occurs infrequently, and because it depends on the timing and extent of marine intrusion, factors that are not easily predicted.
This study utilized air quality data from CCOS and also the Bay Area Air Quality Management District (District) monitoring sites. Meteorological data were obtained from the Oakland radiosonde (RAOB) soundings in the central Bay Area as well as the District's network of meteorological sites and several airport sites.
Analysis of high ozone days shows that high temperatures and light winds are associated with high ozone. Analysis of wind direction showed that high ozone was associated with previous-day afternoon 850 mb Oakland RAOB winds having an easterly component. High ozone at Livermore is always associated with northerly morning surface winds in the Concord area (to the north of Livermore) suggesting possible transport from that area.
To investigate whether multiple types of episodes existed, a cluster analysis was performed utilizing daily maximum 1-hour ozone from District monitoring sites. Any day with an exceedance of the 1-hour ozone standard (ozone at least 125 ppb) at any site was included, yielding 36 days from 1995 through 2002. The cluster analysis resulted in 3 clusters. The third cluster contained only 2 days, both from 1995, and was considered rare enough to be ignored. The July 30-August 2, 2000 episode fell into Cluster 1 as did the other CCOS episode days. A comparison of clusters showed that the mean of the District-maximum ozone for Cluster 2 was significantly higher than that for Cluster 1, and the mean ozone from all sites was much higher. Cluster 2 episodes tended to be widespread, with exceedances at several sites, whereas Cluster 1 episodes had exceedances only at one site, usually Livermore.
Cluster 2 had higher average maximum temperatures, especially at San Jose ? 98oF vs. 92oF. Cluster 2 winds were somewhat lighter, and the 16:00 PST 850 mb wind directions were southerly, whereas Cluster 1 winds were westerly. In terms of timing of the episodes, there was not a significant difference ? with most episode days occurring in July and August in both clusters. Both weekends and weekdays were represented in both clusters. An analysis of temperatures and winds at several locations shows that the modeled episode days were not exceptional for their clusters.
The Weibull regression was used to assess the severity of episodes, following the U.S. EPA guidance. This method uses regression to predict ozone from various meteorological and temporal variables. These variables were limited to those with long time series because of the need to establish an index of meteorological ozone potential. Here, daily basin-wide maximum ozone for 1990-2002 was regressed against Livermore and nearby Travis AFB surface meteorology, Oakland RAOB temperature and wind variables, one-day lags of these variables, day-of-year, and an indicator for weekend/weekday.
The Weibull regression analysis found the following variables to be useful predictors: Livermore maximum temperature, Travis AFB mid-day average windspeed, previous-day afternoon 850 mb winds with an easterly component, and whether it was a weekend or weekday. The resulting formula was then applied to a longer time period ? the 24 years from 1979 through 2002.
The Weibull regression produced predictions of ozone potential for each day, which were then ranked from highest potential to lowest. The rankings of the modeling episode days that exceeded the federal 1-hour ozone standard were as follows: 7/11/99: 7, 7/12/99: 25, and 7/31/00: 67. A day with rank 24 also corresponds to the ozone design value, in other words, an ozone potential that occurs once per year on average. Thus, the 7/12/99 episode represents a day that on average occurs nearly once per year. The 7/11/99 and 7/31/00 episodes represent days with ozone potential that occur less than once per year and more than once per year on average, respectively.
Session 2, Air Quality Forecasting - Case Studies
Monday, 23 August 2004, 10:30 AM-12:00 PM
Previous paper
Recorded presentation