Multiscale Processes of Hurricane Sandy (2012) as Revealed by the Parallel Ensemble Empirical Mode Decomposition and Advanced Visualization Technology

Abstract

Hurricane Sandy in October 2012 is currently the second costliest tropical cyclone (TC) in the U.S. history, surpassed only by Hurricane Katrina (2005). This paper uses advanced data analysis methods and visualization technology to examine the role of multiscale processes in the initial formation and movement of Hurricane Sandy. To efficiently analyze high-resolution, global, and multiple-dimensional datasets, a parallel ensemble empirical mode decomposition (PEEMD) method is developed by implementing a multi-level parallelism into an ensemble EMD (EEMD). The augmentation resulted in a parallel speedup of 720 using 200 eight-core processors. Here, we discuss performance for the PEEMD in decomposing multiscale signals from data sets that represent: (i) idealized tropical waves and (ii) large-scale environmental flows associated with Hurricane Sandy (2012). Our results indicate that the PEEMD can efficiently reveal major wave characteristics such as wavelengths and periods within the data by sifting out the dominant (wave) components. Visualization tools have been developed to make four-dimensional (4D) visualizations of Sandy. The 4D visualizations help elucidate the following factors which led to the sinuous track of Sandy: (i) the initial steering impact of an upper-level trough (appearing over the Northwestern Caribbean Sea and the Gulf of Mexico); (ii) the blocking impact of systems to the Northeast of Sandy; and (iii) interaction with a mid-latitude, upper-level trough that appeared at 130 degrees West longitude on October 23, moved to the east coast, and intensified from October 29–30 prior to Sandy’s landfall. Both the PEEMD method and advanced visualization technology have been integrated with other modules in order to examine the statistical relationship between tropical waves and TC formation in a hurricane climate study.