Detecting plumes in mobile air quality monitoring time series withdensity-based spatial clustering of applications with noise

Mobile monitoring is becoming an increasingly populartechnique to assess air pollution on fine spatial scales, but methods todetermine specific source contributions to measured pollutants are sorelyneeded. One approach is to isolate plumes from mobile monitoring time seriesand analyze them separately, but methods that are suitable for large mobilemonitoring time series are lacking. Here we discuss a novel method used todetect and isolate plumes from an extensive mobile monitoring data set. Thenew method relies on density-based spatial clustering of applications withnoise (DBSCAN), an unsupervised machine learning technique. The new methodsystematically runs DBSCAN on mobile monitoring time series by day andidentifies a subset of points as anomalies for further analysis. Whenapplied to a mobile monitoring data set collected in Houston, Texas,analyzed anomalies reveal patterns associated with different types ofvehicle emission profiles. We observe spatial differences in these patternsand reveal striking disparities by census tract. These results can be usedto inform stakeholders of spatial variations in emission profiles notobvious using data from stationary monitors alone.