Towards a new tropical cyclone risk assessment system – The stochastic intensity and track models

We are developing a system to estimate the risk of a rare, high-impact tropical cyclone landfall by generating ensembles of synthetic tropical cyclones whose properties depend on the climates in which they occur. In this talk, we will present results from two components of the system: an auto-regressive (AR) TC intensity model and a beta-advection track model. The AR TC intensity model includes a deterministic component, derived empirically via regression, which advances the TC intensity in time and accounts for the surrounding large-scale environment; and a stochastic forcing, which represents the component of TC intensification that is not linearly related to the storms' ambient conditions. Potential intensity, deep layer mean vertical shear, and mid-level relative humidity are the only environmental variables considered in the deterministic component. The track model advects storms with the synthetic deep-layer steering flow plus beta drift.

Models are developed and verified with the environmental information from ERA-Interim monthly reanalysis data, and the historical best-track tropical cyclone data. The figure of merit is the statistical distribution of TC intensity and track density, i.e., the TC climatology. A good simulation of the observed intensity climatology requires the auto-correlated stochastic forcing, and the inclusion of nonlinear potential intensity terms in the deterministic forcing. The preliminary results from the track model show a track density similar to that of the historical records, except that the synthetic tracks are slightly smoother than the observed ones. In additional to describing the development and verification of the models, we will discuss also the dependence of the TC climatology on the environment using the synthetic storms data.