1/f noise and pulse‐like events in the tropical atmospheric surface variabilities

This study discusses the 1 /f ‐noise behaviour and its connection with pulse‐like events in tropical surface meteorological data. The 1 /f noise, with f being frequency, refers to the unique spectrum shape of signals that represents the same degree of variability for a whole spectrum range with very long lag‐correlations. The pulse‐like events are characterized by their high localizations in time and a lack of a distinct spectrum peak. Approximate 1 /f ‐noise behaviour is found in tropical surface temperature, moisture, and wind speed for periods of 1 hour to beyond 10 days and in oceanic wind stress up to 2‐year periods. The origin of the 1 /f noise can be closely tied to pulse‐like events with a highly intermittent nature in the time series. Examples for such pulse‐like events include the cumulus convection events leading to convective downdraughts and intraseasonal westerly wind events associated with the Madden–Julian oscillations (MJO). A method based on the wavelet is introduced to extract individual pulse‐like events. These events are characterized by a 1 /f spectrum extending up to 1–2 decades of frequency centred around their characteristic durations. Their successive occurrences are mutually uncorrelated and the individual pulse‐like events themselves constitute a 1 /f spectrum for a frequency range around their durations. The superposition of these events of various durations yields the 1 /f ‐noise spectrum for much wider ranges. Physical implications of these findings to convective quasi‐equilibrium and MJO‐El Niño Southern Oscillation relationship are discussed.