Abstract View
Using Dynamic Principal Components to Analyze Mobile Particulate Matter Measurements
BLAKE ACTKINSON, Robert Griffin, Katherine Ensor, Rice University
Abstract Number: 708
Working Group: Instrumentation and Methods
Abstract
Mobile monitoring is becoming an increasingly popular technique to measure particulate air pollution because of the high spatial resolution of its measurements. While the high spatial resolution of these measurements offers advances in understanding the distribution of particulate matter (PM), work is needed to characterize its sources. Past work has used Principal Component Analysis (PCA) to analyze source contributions to mobile PM measurements; however, it is problematic because these measurements are inherently non-stationary if collected over an extended period of time and space.
Here we discuss the use of Generalized Dynamic Principal Components (GDPC) to evaluate the sources of PM collected with a mobile platform in Houston, Texas. GDPCs are lagged principal component reconstructions of the data that are obtained through minimizing a mean-squared error loss function. GDPCs can be applied to non-stationary data to characterize the potential source contributions of that data. We also discuss differences between what this technique shows about source contributions compared to more traditional principal components. We believe that GDPC offer an improved way to identify source contributions of PM.