Characteristics of the Variability and Predictability of Idealized Hurricanes in Steady State

The prediction of hurricane intensity poses a major challenge to the forecasting community today, and improvements in operational tropical cyclone (TC) intensity forecasts lag behind those of TC track forecasts. While TC track is mainly determined by synoptic-scale steering influences, intensity is determined by a host of multi-scale phenomena tied to the internal dynamics of the storm, the ambient environment that it is embedded within and convective-scale stochastic forcing. We focus here on the internal, or intrinsic, dynamics associated with the structure of the cyclone itself. We propose that the intrinsic dynamics, associated with convective heating organized by balanced circulations, may provide a baseline for studying the variability and predictability of TCs. We choose an idealized framework, which allows for control of the environment, and avoids complications inherent to observed storms, such as data scarcity, inhomogeneous sampling and data assimilation. Furthermore, we select long-running three-dimensional simulations in statistically steady state in order to exclude transients due to initial conditions and to capture the natural variability of the storm in a quiescent environment; this approach also allows for robust sampling and statistical analysis in a manner not possible with real storms.

The primary model used for the idealized simulation is the CM1 cloud-resolving model. A three dimensional hurricane is simulated for one hundred days, yielding an extended period of statistically steady intensity. To avoid environmental influences, the domain is initialized with no ambient flow and a constant and uniform sea surface temperature. The variability of the axisymmetric component of the circulation is studied using principal component/empirical orthogonal function (PC/EOF) analysis and the relationship of the leading modes of axisymmetric variability with the asymmetric component is investigated. Predictability is analyzed using the autocorrelation e-folding time. When compared to two-dimensional results, marked similarities are found in spite of the fact that the three-dimensional storm has a weaker intensity.