Cluster Analysis of Hourly Wind Measurements to Reveal Synoptic Regimes Affecting Air Quality

Abstract

Abstract A clustering algorithm is developed to study hourly, ground-level wind measurements obtained from a network of monitoring stations positioned throughout the San Francisco Bay Area of California. A statistical model based on principal components analysis (or empirical orthogonal functions) is used to cluster these autocorrelated and cross-correlated observations. Patterns at the synoptic time scale are isolated by using windowing and scaling operations to treat the data. Four dominant wind patterns that affect air quality are identified for the study region, and summer days from 8 yr of historical data are assigned to these modes. One cluster captures a high pressure system over the western United States, the anticyclonic winds of which block the typical marine flow through the study region. Differential heating convects a polluted air mass to a nearby valley in which severe episodes of higher-than-average ozone composition occur. A second pattern represents a seasonal, offshore ridge of high pressure that reduces marine flow and produces a shallow boundary layer. These two clusters capture distinct meteorological regimes that are favorable to ozone buildup and account for nearly all “exceedances” of ozone air-quality standards. Two other clusters have strong marine flow that inhibits ozone buildup.