Structurally Flexible Bayesian Microphysics with BOSS
First Author
Marcus van Lier-Walqui
mv2525@columbia.edu
All Authors
Hugh Morrison, Sean P. Santos, Kaitlyn Loftus
Topic
Atmosphere
Project
ecosystem project (ESMD)
Abstract
We present work on developing structurally flexible hybrid statistical-physical microphysics schemes suitable for use within both E3SM as well as high-resolution models, called the Bayesian Observationally-constrained Statistical-physical Scheme (BOSS, Morrison et al. 2020, van Lier-Walqui et al. 2020). We have expanded BOSS (initially rain-only) to include cloud processes including condensational growth and rain formation. BOSS has the advantage of systematically adjustable structure (e.g. any combination of prognostic moments describing the size distribution), a bare minimum of a priori assumptions (e.g. no assumed size distribution shape), and the ability to quantify uncertainty in parameters that control its rate equations. Fitting of BOSS parameters to reference bin and bulk schemes has been tested in offline mode (directly fitting process rates) as well as within online 1D kinematic simulations, and most recently, 3D Large Eddy Simulations of drizzling stratocumulus. The latter takes advantage of methodologies for parameter estimation using perturbed-parameter ensembles, machine learning emulators, and Bayesian Markov Chain Monte Carlo samplers. We also compare versions of BOSS with varying structure, including versions of BOSS that eschew the traditional distinction between cloud and precipitation particles, and instead evolve prognostic moments of the full range of liquid drop sizes (single liquid category). Prospects for implementation in E3SM are discussed, including preliminary tests within the MG2 microphysics scheme in SCM E3SM simulations.
In-person
no
Poster
Add a poster or a presentations here
Discussion Link
Add a link to your chosen place for discussion: either slack link, public google doc link or email address
