Numerical experiments on the predictability of tropical-cyclone intensification

Numerical model experiments are carried out to investigate the predictability of vortex growth in the prototype problem for tropical-cyclone intensification, which considers the evolution of a prescribed, initially cloud-free, axisymmetric vortex on an f-plane in an environment at rest. The calculations are carried out in a large square domain with impervious boundaries with the vortex axis initially at the centre. They are performed on a square horizontal grid. The model used is the Pennsylvania State University/National Center for Atmospheric Research Mesoscale Model (MM5).

As in many previous calculations, the vortex evolution begins with a gestation period during which the vortex slowly decays due to surface friction, but moistens due to evaporation from the underlying sea surface. Subsequently, moist convection begins near the radius of maximum tangential wind speed and there ensues a period during which the vortex rapidly intensifies. At the end of this period, which is typically 48 h in our calculations, the vortex attains a quasi-steady state in which the vortex exhibits many realistic features of a mature tropical cyclone, with spiral bands of convection surrounding an approximately symmetric eyewall and a central convection-free eye. Our interest here is focused on the structure of the asymmetries and their evolution.

During the gestation period the flow remains close to axisymmetric with a weak azimuthal wavenumber-4 asymmetry that necessarily arises from the representation of a circular flow on a square grid. For a relatively coarse horizontal resolution of 15 km, saturation occurs first with a wavenumber-4 pattern, but other wavenumbers quickly emerge, first wavenumber-2 and then other wavenumbers including wavenumber-1. As the mature stage is approached, the flow consolidates into a monopole vortex once again, much as in numerical experiments using the shallow-water approximation described by Guinn and Schubert. As the horizontal resolution is increased while keeping the horizontal diffusivity as low as possible, the initial pattern of convection has increasing azimuthal wavenumber: for example with a 5 km grid, the pattern has wavenumber-12, but again other wavenumbers rapidly emerge. We have found that the asymmetries that develop are highly sensitive to the surface moisture distribution. If a random moisture perturbation is added in the boundary layer at the initial time with a magnitude that is below the accuracy with which moisture can be measured, the pattern of evolution of the flow asymmetries is dramatically changed. We conclude that the flow is not deterministic and only those features that survive in an ensemble average of many realizations can be regarded as robust features. There are clear implications for the possibility of deterministic forecasts of the mesoscale structure of hurricanes, which may have a large impact on the intensity and on rapid intensity changes.