“Temporal Memory and Three-Dimensional Structure in Cumulus Convection: Toward Improved MJO Prediction in the Unified Forecast System”
September 25, 2025
Lisa Bengtsson
Hosted by Eric Maloney
Abstract
With increasing resolution and shorter time steps in global numerical weather prediction, traditional assumptions in cumulus parameterizations, such as steady state and statistical equilibrium, become invalid. Convection can evolve over multiple timesteps and cover a substantial fraction of the grid box, particularly when it becomes organized. In this talk, I will present recent advances in the Unified Forecast System (UFS) global applications (GFS/GEFS and SFS) aimed at representing convective memory, three-dimensional organization, and stochasticity. These include a prognostic closure that evolves convective area fraction and updraft velocity, along with cellular automata to represent subgrid (and cross-grid) organization.Case studies show that the new prognostic closure improves Madden-Julian Oscillation (MJO) prediction, with better amplitude, phase, and propagation. Ongoing work suggests these improvements arise from stronger coupling between moisture flux convergence and MJO activity, including preconditioning of the western and central Pacific that supports eastward propagation across the Maritime Continent.
While physics is critical, I will also present recent results demonstrating large sensitivity in circulation-based MJO metrics to differences in the initial-state static stability, where reduced stability enhances vertical motion and upper level divergence for a given heating under the weak temperature gradient approximation, leading to a slow adjustment and enhanced MJO amplitude lasting several weeks.