We investigate the amplification and predictability of tropical cyclones in the context of a minimal, three-dimensional numerical model. In the prototype problem for intensification, starting with a tropical storm strength vortex in a quiescent environment on an f-plane, the emergent flow in the inner region of the vortex becomes highly asymmetric and dominated by deep convective vortex structures, even though the problem as posed is essentially axisymmetric. The details of the intensification process, including the asymmetric structures that develop, are highly sensitive to small perturbations in the low-level moisture field at the initial time. This sensitivity is manifest in a significant spread in the intensity of vortices from an ensemble of calculations in which random moisture perturbations are added in the lowest model level. Similar experiments are carried out on a beta-plane and in the case where there is an anticyclonic shear flow at upper levels. The former set shows no significant difference from the f-plane calculations in the evolution of intensity, but the latter set shows a significantly weaker vortex, contrary to a broadly held hypothesis that upper-level outflow channels are favourable to intensification. Copyright (C) 2008 Royal Meteorological Society
