A wavelet-based approach applied to suspended particulate matter time series in Portugal

Air Quality, Atmosphere & Health, 9 (2016), 847–859
http://link.springer.com/article/10.1007/s11869-016-0393-4

This study intends to analyse the particulate matter (PM) levels in Portugal (mainland and islands) throughout a 3-year period. Although a decreasing trend has been observed, the WHO guidelines for the PM10 and PM2.5 annual mean concentrations have been exceeded in all monitoring stations. Most inland urban, rural and suburban sites follow a pronounced seasonal variation with much higher values in winter than in summer. Lower levels and a weak seasonal variability were registered in the two urban background stations of Madeira Island, which are permanently under the influence of clean air masses over the Atlantic. Receiving long-range transported pollution, rural stations located in mountain sites presented an opposite seasonal pattern, with higher levels in summer. Diurnal profiles were also analysed and compared between stations. A mining process was also carried out, consisting in the application of multi-scale wavelet transforms, data pattern identification using cluster analysis and examination of the contribution to the total variance/covariance of the time series per wavelet scale for all stations. Groups of stations exhibiting similar variance/covariance profiles were identified. One group contains urban and rural stations with diurnal and daily time scales. Urban background stations located in the island of Madeira constitute another cluster, corresponding to higher wavelet scales (lower periodicity phenomena). One traffic station in the Oporto metropolitan area was grouped with a suburban/industrial station of central Portugal, suggesting the need for reclassification in what concerns the type of environmental influence.