A Systemic Analysis of Multiscale Deep Convective Variability over the Tropical Pacific

The multiscale tropical deep convective variability over the Pacific Ocean is examined with the 4-month highresolution deep convection index ( ) derived from satellite imagery. With a systemic view, the complex ITBB phenomenon is described with succinct parameters known as generalized dimensions associated with the correlation structures embedded in the observed time series, with higher-order dimensions emphasizing extreme convective events. It is suggested that convective activities of lifetimes ranging from 1 h to ;21 days have interdependence across scales that can be described by a series of power laws; hence, a spectrum of generalized dimensions, that is, the time series is multifractal. The spatiotemporal features of the time series is I I T T BB BB preliminarily examined by changing the spatial domain from 0.18 3 0.18 to 258 3 258. The multifractal features are weakened with increasing strength of spatial averaging but cannot be eliminated. Furthermore, the data ITBB has the property of long-range dependency, implying that its autocorrelation function decays with a power law in contrast to the zero or exponentially decaying autocorrelation functions for white and commonly used red noise processes generated from autoregressive models. Physically, this means that intensified convection tends to be followed by another intensified event, and vice versa for weakened events or droughts. Such tendency is stronger with larger domain averaging, probably due to more complete inclusion of larger-scale variability that has more definite trends, such as the supercloud clusters associated with the Madden–Julian oscillation (MJO). The evolution of cloud clusters within an MJO event is studied by following the MJO system across the analysis domain for ;21 days. Convective activities along the front, center, and rear parts of the MJO event continuously intensify while approaching the date line, indicating multifractal features in the range of 1 h to about 5–10 days. Convective activity along the front and rear edges of the MJO event are more intermittent than in the center. The multifractal features of the time series can be approximated by the random multiplicative cascade ITBB processes, suggesting likely mechanisms for the multiscale behavior and casting concern on the predictability time scale of the observed phenomena.