Transient Response of Axisymmetric Hurricanes to Sea Surface Temperature Perturbations

While the relationship between tropical cyclone maximum intensity and sea surface temperature (SST) is well understood theoretically, and well documented empirically, the transient response of storm intensity to SST changes is less well understood. In particular, it has long been recognized that SST changes alone are not a good predictor of intensity change due to contributions from numerous other factors in observed storms. Eliminating these other factors in order to isolate and better understand the transient response to SST changes is the subject of this paper. Moreover, I explore the contributions to intensity predictability from intrinsic variability (due to initial condition error) and to SST forcing.

The need for large samples to derive statistically significant results, combined with the need to control environmental factors, motivates an idealized axisymmetric numerical modeling approach. Here we use version-15 of Bryan's CM1 model, and sample from a 1000-day simulation of a hurricane in statistical equilibrium. Results reveal an asymmetric response between negative and positive SST perturbations, where negative SST perturbations more consistently affect storm intensity over a wide range of initial intensities. The relative contribution of initial condition error (here, "roundoff error") and forcing (SST) to intensity change predictability is measured using relative entropy (RE). For the control, RE shows that the predictability limit associated with initial condition error is about 2 days, consistent with previous estimates. Relative to the control, the loss of RE due to dispersion in the initial conditions is approximately the same for negative SST perturbations, and significantly faster for positive SST perturbations. The timescale for the contribution to RE associated with SST forcing to exceed that from initial condition dispersion is approximately 24 hours for both positive and negative perturbations. At 99% confidence, the predictability limit associated with -3K SST forcing exceeds 5 days, compared to only 2.5 days for 3K SST forcing. The sensitivity of these results to the temporal evolution of SST will also be discussed.