| description abstract | nvironmental measurements often show unsteady variability and background red noise (stochastic component) superimposed on persistent trivial daily, seasonal, and annual variability. In addition to this, environmental time series often present gaps due to a myriad of possible factors, such as malfunction of the sensors, connection loss, etc. As a result, interpreting and identifying periodicities in this type of time series by means of spectral analysis tools, like the Fourier transform, are difficult and lack precision. To overcome these difficulties, a methodology is proposed in the first part of this paper that integrates statistical tools (iterative Student?s t test), parametric reconstruction, and spectral analysis (Lomb periodogram and wavelets). In the second part of the paper, this methodology is tested (i) in the high-frequency part of the spectrum of two (well known) synthetic time series and (ii) to identify nontrivial (e.g., daily cycles) high-frequency periodicities (linked to some mesometeorological processes) in three tropospheric ozone time series recorded by the Valencia regional air quality monitoring network (on the Mediterranean side of Spain) during a 14-yr period. This methodology can determine statistically significant, seasonally dependent recurrences in the high-frequency variability (<15 days) observed in ozone time series measured in a Mediterranean region of Spain under high noise-to-signal ratios. | |
