A methodology of adaptive time series analysis based on Empirical Mode Decomposition (EMD) has been applied to investigate Be7 activity concentration variability, along with temperature. Analysed data were sampled daily at ground level by 28 different stations of the CTBTO-IMS network. The adopted methodology allows to characterise trend component, yearly cycles and outlier occurrence of Be7. Trend component is first estimated via simple EMD and removed. The recent time-varying filter EMD (tvf-EMD) technique is instead employed to extract local narrowband oscillatory modes from the detrended data. Denoising is carried out using a threshold on the Hurst exponent of extracted oscillatory modes (IMFs). It is found that the phase of the yearly cycles is shifted at high latitudes, possibly due to the Hadley cell dynamics. Furthermore, high values of Be7 yearly cycle are found for some stations in 2009–2010. Due to their location, this is possibly due to the El Nino event occurring that year. Though, further studies are needed in this regard.