Abstract

Particulate matter (PM) is well known to be detrimental to health and it is crucial to apportion PM into the underlying sources to target policies. Particle number size distribution (PNSD) is the most accessible data to identify these sources, which provides information on the PM sizes. The most widely used method to analyze this data is the positive matrix factorization (PMF). However, the PMF presents several limitations. We propose a new functional factor model for PNSD, which allows to disentangle PM into sources and contributions while considering the complex dependencies of the data across different sizes and periods. The case study is the hourly data collected in London for seven years. The proposed methodology is fast, accurate, and reproducible.