A study of the temporal dynamics of ambient particulate matter using stochastic and chaotic techniques

Temporal dynamics of particulate matter (PM) concentration are affected by a variety of complex physical and chemical interactions among ambient pollutants and various exogenous factors (e.g. meteorological variables). Consequently, the dynamics of PM concentration can be considered either as a stochastic process or as a deterministic process. Many studies have applied stochastic and chaotic approaches independently to study the dynamics of PM concentration. However, none of them has compared these two complementary approaches for verification and possible confirmation of the outcomes. The present study makes an attempt to address this issue, through application of the dynamic factor analysis (DFA) (a stochastic method) and the correlation dimension (CD) method (a chaotic method) to study the temporal dynamics of ambient pollutants. More specifically, these two methods are employed to identify the number of variables dominantly governing the dynamics of PM concentration, with analysis of PM10, PM2.5, and ten other variables observed at the Hsing-Chuang station in Taipei (Taiwan). The results from the two methods are found to be consistent, with the DFA method suggesting eight common trends among the observed time series and the CD method suggesting eight variables dominantly governing the dynamics of both PM10 and PM2.5. This study provides an excellent example for the utility of both stochastic and chaotic approaches in modeling atmospheric and environmental systems, as these approaches not only shed light in their own ways but also complement each other in capturing the salient characteristics of such systems, especially from the perspective of simplified modeling.

► PM10, PM2.5 and their associated ambient pollutants are investigated by both stochastic and chaotic methods.
► Both stochastic and chaotic methods reveal eight dominant underlying temporal dynamics of PM10 and PM2.5.
► Multivariate dynamic factor analysis identifies the common trends among the ambient pollutants.
► Correlation dimension method identifies the deterministic dynamics without linear assumption.