The Bimodal Distribution of Tropical Cyclone Intensity and Rapid Intensification

In tropical cyclone (TC)-climate research, the distribution of the storms' lifetime maximum intensity (LMI) is an important metric to determine whether a given model can capture the storm intensity climatology. The distribution of LMI is bimodal; the present study aims to understand why. More specifically, we examine to what degree the bimodal distribution can be explained from a large-scale perspective. Monthly-mean ERA-Interim fields and historical global TC best-track data from 1981 to 2012 are used. Storms are first classified according to storm-scale parameters (e.g., genesis location, length of storm's lifetime, intensification rate, etc). Results show that the second mode consists of storms that go through rapid intensification (RI, intensify more than 30kt per 24 hours) at least once in their lifetime, i.e., RI storms. We then use linear discriminant analysis (LDA) to separate two classes of storms with according to linear combinations of environmental variables. The synoptic scale parameters considered include potential intensity, shear, atmospheric stability, moisture, and others thought relevant to storm intensity generally and RI in particular. With data from the North Western Pacific, we found that in LDA, using potential intensity accumulated over the period from the storm formation to the time when it reaches its LMI, the mean upper ocean temperature, and deep-layer mean shear can best classify storms into two groups. In other words, the critical environmental conditions controlling the likelihood of RI in the WPC are potential intensity, shear, and upper ocean structure. With data from the North Atlantic, however, results suggest that it is the TC-trough interaction and atmospheric instability that matter the most for separating RI and non-RI storms.