Non-local convergence coupling in a simple stochastic convection model

- A stochastic convection model with non-local convergence coupling is developed.
- Linearized analysis shows non-locality is needed to stabilize high wavenumbers.
- Idealized nonlinear stochastic simulations show desirable low frequency variability.

Observational studies show a strong correlation between large-scale wind convergence and precipitation. However, using this as a convective closure assumption to determine the total precipitation in a numerical model typically leads to deleterious wave-CISK behavior such as grid-scale noise. The quasi-equilibrium (QE) schemes ameliorate this issue and smooth the precipitation field, but still inadequately represent the intermittent and organized nature of tropical convection. However, recent observational evidence highlights that the large-scale convergence field primarily affects precipitation by increasing the overall convective cloud fraction rather than the energetics of individual convective elements. In this article, the dynamical consequences of this diagnostic observation are studied using a simple one baroclinic mode stochastic model for convectively coupled waves. A version of this model is implemented which couples the stochastic formation of convective elements to the wind convergence. Linearized analysis shows that using the local convergence results in a classic wave-CISK standing instability where the growth rate increases with the wavenumber. However, using a large-scale averaged convergence restricts the instability to physically plausible scales. Convergence coupling is interpreted as a surrogate for the non-local effects of gregarious convection. In nonlinear stochastic simulations with a non-uniform imposed sea surface temperature (SST) field, the non-local convergence coupling introduces desirable intermittent variability on intraseasonal time scales. Convergence coupling leads to a circulation with a similar mean but higher variability than the equivalent parameterization without convergence coupling. Finally, the model is shown to retain these features on fine and coarse mesh sizes.