Development of ensemble neural network convection parameterizations for climate models using ARM data

Poster PDF

Description

The neural network (NN) approach is formulated and used for development of a NN ensemble stochastic convection parameterization for climate models. This fast parameterization is built based on data from Cloud Resolving Model (CRM) simulations initialized with and driven/forced by observational (TOGA-COARE and ARM) data. The observational data are also used for validation of model simulations. The SAM (System for Atmospheric Modeling), developed by D. Randall, M. Khairoutdinov, and their collaborators, has been provided by M. Khairoutdinov and used for CRM simulations. CRM-emulated data have been averaged and projected onto the GCM space of atmospheric states to implicitly define a stochastic convection parameterization. This parameterization is emulated using an ensemble of neural networks (NN). An ensemble of NNs with different architecture, i.e., with different inputs and outputs, has been trained and tested. The inherent uncertainties of the stochastic convection parameterization are described and estimated. Due to these inherent uncertainties, their ensemble is used to constitute a stochastic NN convection parameterization. The major challenges of development of stochastic NN convection parameterizations based on our initial results are discussed. At the next step of the project, the stochastic NN convection parameterizations will be included into the NCAR SCM CAM and/or CAM in diagnostic and prognostic modes and tested in climate simulations using data from the SGP (Southern Great Plains) and TWP (Tropical Western Pacific) ARM sites and TOGA-COARE data. Acknowledgments: The investigators would like to thank Prof. Marat Khairoutdinov (SUNY) for providing SAM and consultations and Dr. Peter Blossey (UWA) for consultations on SAM.