Transient Environmental Sensitivities of Explicitly Simulated Tropical Convection

Abstract

A three-dimensional cloud-resolving model, maintained in a statistically steady convecting state by tropics-like forcing, is subjected to sudden (10 min) stimuli consisting of horizontally homogeneous temperature and/or moisture sources with various profiles. Ensembles of simulations are used to increase the statistical robustness of the results and to assess the deterministic nature of the model response for domain sizes near contemporary global model resolution. The response to middle- and upper-tropospheric perturbations is predominantly local in the vertical: convection damps the imposed stimulus over a few hours. Low-level perturbations are similarly damped, but also produce a vertically nonlocal response: enhancement or suppression of new deep convective clouds extending above the perturbed level. Experiments show that the ?effective inhibition layer? for deep convection is about 4 km deep, far deeper than traditional convective inhibition defined for undilute lifted parcels. Both the local and nonlocal responses are remarkably linear but can be highly stochastic, especially if deep convection is only intermittently present (small domains, weak forcing). Quantitatively, temperature-versus-moisture perturbations in a ratio corresponding to adiabatic vertical displacements produce responses of roughly equal magnitude. However, moisture perturbations seem to provoke the nonlocal (upward spreading) type of response more effectively. This nonlocal part of the response is also more effective when background forcing intensity is weak. Only at very high intensity does the response approach the limits of purely local damping and pure determinism that would be most convenient for theory and parameterization.