Temporal evolution characteristics of PM2.5 concentration based on continuous wavelet transform

Abstract

Fine particulate matter (PM2.5) is an important haze index, and the researches on the evolutionary characteristics of the PM2.5 concentration will provide a fundamental and guiding prerequisite for the haze prediction. However, the past researchers were usually based on the overall time-domain evolution information of PM2.5. Since the temporal evolution of PM2.5 concentration is nonstationary, previous studies might neglect some important localization features that the evolution has various predominant periods at different scales. Therefore, we applied the wavelet transform to study the localized intermittent oscillations of PM2.5. First, we analyze the daily average PM2.5 concentration collected from the automatic monitoring stations. The result reveals that the predominant oscillation period does vary with time. There exist multiple oscillation periods on the scale of 14–32 d, 62–104 d, 105–178 d and 216–389 d and the 298d is the first dominant period in the entire evolutionary process. Moreover, we want to figure out whether the temporal characteristics of PM2.5 in the days with heavy haze also have localized intermittent periodicities. We select the hourly average PM2.5 concentration in 120 h when the haze pollution is serious. We find that the principal period has experienced two abrupt shifts and the energy at the 63-hour scale is the most powerful. The results in these two independent analyses come into the same conclusion that the multiscale features shown in the temporal evolution of PM2.5 cannot be ignored and may play an important role in the further haze prediction.