Scale-specific stochastic modeling of tropical cyclone intensity changes

Tropical Cyclone (TC) intensity forecasting has come a long way since the first, axisymmetric, two-layered model developed by Ooyama. Despite significant improvements in the prediction of rapid intensity changes, multi-scale interactions, and the predictability across-scales, we still find it very difficult to predict the scenarios where the same magnitude of external forcing such as wind-shear or landfall can result in vastly different trajectories. For such cases, a cascade of internal cross-scale stochastic interactions may influence the TC intensity evolution under the absence of any abrupt external forcing. We identify the modeling of multi-scale stochastic shocks and associated impacts as the future challenge for TC intensity predictions.

As a step forward, we present a new conceptual framework to characterize the spectrum of possible intensity pathways for a TC vortex at any given time and quantify the non-stationary probability distributions of stochastic intensity transitions. Each of the TC's intensity pathways is viewed as a distinct attractor basin and a combination of several external and internal factors across multiple scales dictates which of the many pathways the TC vortex takes. In our framework, a stochastic shock may arise from any of the various scales within a TC vortex and the subsequent cross-scale energy transactions may rapidly increase the probability of the vortex intensifying or weakening. The novel aspect of this work is the modeling of scale-specific stochasticity and cross-scale feedbacks within the TC vortex. The stochastic term is modeled in a realistic manner such that its amplitude and memory parameter are consistent with our understanding of low and high-wavenumber asymmetries in real TCs. Two Bay of Bengal TCs- Phailin (intensifying) and Lehar (weakening), serve as case studies and we simulate them using the Hurricane Weather Research Forecast model. An ensemble of intensity pathways is generated, and the non-stationary probability distributions of the intensity transitions at each time are examined. Our approach is another step toward an improved understanding of the stochastic dynamics of multi-scale transitions of a TC vortex.