Analysis of environmental time series in the frequency domain

摘要

Analysis in the frequency domain is a common technique usually performed to reveal periodic components of a physical phenomenon that are not manifest after a preliminary study of the related time series. It can also be performed to study correlation between two variables. This work illustrates procedures and results of spectral analysis performed over air pollutant time series. The focus has been set on photochemical pollutants traced by ozone concentrations, considering potential correlations with chemical precursors and atmospheric parameters such as solar radiation, wind velocity and air temperature. Primary pollutants have also been considered. The selected data set consists of concentration hourly averages and elementary values, provided by the monitoring network of Ancona Province, Italy. We have analysed hourly averages to investigate long-term and medium-term behaviour of pollutants, while elementary values have been used to study scattering phenomena (high frequency fluctuations in concentration levels) registered in some ozone time series. We employed the Lomb-Scargle periodogram to estimate the power density spectrum, since such a method is able to deal with unevenly sampled data that are quite common in pollutant time series in which instrumental dropouts or validation problems may lead to incomplete data set with several gaps. Results have yielded to possible explanations regarding the origins of oscillations.