A Weak Temperature Gradient Framework to Quantify the Causes of Potential Intensity Variability in the Tropics

Abstract

<jats:title>Abstract</jats:title><jats:p>Potential intensity (PI) has been shown to have a linear sensitivity to sea surface temperature (SST) of about 8 m s<jats:sup>−1</jats:sup> K<jats:sup>−1</jats:sup>, which is close to the sensitivity of PI in simulations subject to a weak temperature gradient (WTG) approximation. This suggests that most of the PI variance is associated with local rather than global SST variations. We verify that PI perturbations are approximately linear in SST, with slopes of 1.8 ± 0.2 m s<jats:sup>−1</jats:sup> K<jats:sup>−1</jats:sup> in radiative–convective equilibrium (RCE) and 9.1 ± 0.9 m s<jats:sup>−1</jats:sup> K<jats:sup>−1</jats:sup> in WTG. To do so, we simulate the sensitivity of both RCE and WTG states in a single-column model (SCM) perturbed by changing in turn CO<jats:sub>2</jats:sub> concentration, aerosol concentrations, prescribed SST, and surface winds speeds. While PI is much more sensitive to SST in WTG than in RCE simulations, the SST itself is much less sensitive to radiative forcing in WTG than in RCE because of the absence of strong atmospheric response. Using these results, we develop a linear model, based on SST and midlevel saturation MSE perturbations, to partition SST and PI perturbations between local components occurring under a WTG constraint and global components that are representative of an RCE state. This model explains up to 95% of the variability of PI in reanalysis. The SCM-derived linear model coefficients are statistically indistinguishable from coefficients from a linear fit of reanalysis PI to SST and midlevel saturation MSE in most ocean basins. Our model shows that North Atlantic PI variations are explained almost entirely by local forcings in recent decades.</jats:p>