Assessment of PM transport patterns using advanced clustering methods and simulations around the San Francisco Bay Area, CA

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner
Monday, 18 January 2010: 1:45 PM
B309 (GWCC)
Scott Beaver, Bay Area Air Quality Management District, San Francisco, CA; and A. Palazoglu, A. Singh, and S. Tanrikulu

Presentation PDF (1.0 MB)

The San Francisco Bay Area (SFBA) has recently been designated in nonattainment of the federal 24-hr fine particulate matter (PM2.5) standard. SFBA 24-hr average PM2.5 levels have exceeded the National Ambient Air Quality Standard (NAAQS) of 35 μg/m3 on around 5-35 days per year since ambient measurements started in 1999. These exceedances occur during the winter months, typically when the atmosphere is stable and low level winds enter the SFBA from the east. Under such episodic conditions, the heavily polluted, neighboring air basins in the inland Sacramento and San Joaquin Valleys are upwind of SFBA.

The goal of this study is to identify representative PM transport patterns using advanced clustering methods and demonstrate their impacts on SFBA using a photochemical model. Depending upon prevailing wind direction, SFBA may be either upwind or downwind of the Sacramento and San Joaquin Valleys. For the purpose of this study, however, we focus only on conditions in which transport is likely to occur into SFBA.

The cluster analysis was applied to extended records (around 10 winter seasons) of routine surface wind measurements. These observations were taken from monitors spanning the three basins between which PM transport is suspected to occur. In this manner, inter-regional air flow patterns directly impacting transport are identified by the clustering.

A total of six air flow patterns were identified by the clustering. Only two of these six clusters are associated with significant proportions of PM2.5 exceedance days for SFBA. These episodic air flow patterns were named I-R2 and I-R4. The leading “I” in the cluster names refers to the inter-regional nature of the detected air flow patterns. The “R” in each cluster's name refers to an aloft ridge of high pressure. This large-scale ridging effect provides enhanced stability and subsidence which limit vertical dispersion of PM and its precursors.

The two SFBA episodic clusters share similar vertical dispersion characteristics; however, they are distinguished by horizontal surface winds. Patterns I-R2 and I-R4 have easterly flows through the SFBA extending into the Delta region and beyond. For I-R2, winds enter the Delta from the Sacramento Valley to the north. For I-R4, winds enter the Delta from the San Joaquin Valley to the south. This pair of clusters is believed to represent two distinct transport scenarios in which the SFBA is impacted by different upwind air basins. I-R2 occurs on around 60% of the SFBA exceedance days, and the highest SFBA PM levels are observed in its southern portion (around San Jose). I-R4 occurs on around 20% of the SFBA exceedance days, and the highest SFBA PM levels are observed in its eastern portion (near Livermore Valley and Carquinez Strait). These unbiased estimates for the proportions of exceedance days for each cluster were determined by pooling only days having PM measurements.

Next, to quantify transport impact, we simulated the two-month “core” winter PM season of December-January for both 2000-01 and 2006-07. During this simulation period, each of the above two SFBA episodic clusters occurred. Meteorology was simulated using MM5 with 4 km horizontal grid resolution and 30 vertical layers. Model validation techniques confirm that the simulated meteorological fields indeed represent flow features known to impact SFBA PM levels. The MM5 output is used to drive air quality simulations for PM using CMAQ.

Transport impacts were gauged using a brute force sensitivity analysis approach. A base case emissions inventory was prepared using the most recently available data for the Central California modeling domain. A base case CMAQ simulation was performed using this inventory.

A first transport simulation was performed with zero anthropogenic emissions for the SFBA. The simulated SFBA PM levels largely represent transported PM. This simulation confirms significant impacts from upwind basins when clusters I-R2 and I-R4 occur. This simulation cannot, however, directly resolve the source basin for the transported PM. Also, differences in PM levels between this simulation and the base case simulation reflect contributions from local SFBA sources.

A second set of transport simulations were performed to distinguish between impacts from the Sacramento and San Joaquin Valleys. Simulations were performed using two additional emissions inventories: with Sacramento Valley anthropogenic emissions zeroed out, and with San Joaquin Valley anthropogenic emissions zeroed out. Reductions of simulated SFBA PM levels for these cases relative to the base case reflect transport impacts from the respective basin for which emissions were eliminated. These simulations evidence variability in the transport impacts. Simulated SFBA PM levels are most strongly impacted by the Sacramento Valley emissions for I-R2 days, and by San Joaquin Valley emissions for I-R4 days. Primary PM transport is dominant for I-R2 days, while secondary PM transport is dominant for I-R4 days.

This study of PM transport in the SFBA explicitly combined results from ambient data analysis and simulation. The data analysis was used to guide the simulations by identifying representative transport patterns. Extensive simulations (over 120 days, using 4 emissions inventories each) were performed to robustly estimate transport impacts. This framework was able to distinguish between and quantify transport impacts under different scenarios having different upwind source areas.