Direct Bayesian model reduction of smaller scale convective activity conditioned on large-scale dynamics

Abstract

We pursue a simplified stochastic representation of smaller scale convective activity conditioned on large-scale dynamics in the atmosphere. For identifying a Bayesian model describing the relation of different scales we use a probabilistic approach by Gerber and Horenko (2017) called Direct Bayesian Model Reduction (DBMR). This is a Bayesian relation model between categorical processes (discrete states), formulated via the conditional probabilities. The convective available potential energy (CAPE) is applied as a large-scale flow variable combined with a subgrid smaller scale time series for the vertical velocity. We found a probabilistic relation of CAPE and vertical up- and downdraft for day and night. This strategy is part of a development process for parametrizations in models of atmospheric dynamics representing the effective influence of unresolved vertical motion on the large-scale flows. The direct probabilistic approach provides a basis for further research on smaller scale convective activity conditioned on other possible large-scale drivers.