A Moist Physics Parameterization Based on Deep Learning

• Published in: Journal of advances in modeling earth systems Vol. 12; no. 9
• Main Authors: Han, Yilun, Zhang, Guang J., Huang, Xiaomeng, Wang, Yong
• Format: Journal Article
• Language: English
• Published: Washington John Wiley & Sons, Inc 01-09-2020; American Geophysical Union (AGU)
• Subjects: Atmospheric circulation; Climate; Clouds; Convection; Deep learning; Energy conservation; ENVIRONMENTAL SCIENCES; General circulation models; Hydrologic data; Moist static energy; Moisture; Monsoon convection; Neural networks; Parameterization; Physics; Precipitation; Radiation; Simulation; Tropical climate; Tropical convection; Warm seasons; Weather forecasting
• ISSN: 1942-2466; 1942-2466
• DOI: 10.1029/2020MS002076

Current moist physics parameterization schemes in general circulation models (GCMs) are the main source of biases in simulated precipitation and atmospheric circulation. Recent advances in machine learning make it possible to explore data‐driven approaches to developing parameterization for moist physics processes such as convection and clouds. This study aims to develop a new moist physics parameterization scheme based on deep learning. We use a residual convolutional neural network (ResNet) for this purpose. It is trained with 1‐year simulation from a superparameterized GCM, SPCAM. An independent year of SPCAM simulation is used for evaluation. In the design of the neural network, referred to as ResCu, the moist static energy conservation during moist processes is considered. In addition, the past history of the atmospheric states, convection, and clouds is also considered. The predicted variables from the neural network are GCM grid‐scale heating and drying rates by convection and clouds, and cloud liquid and ice water contents. Precipitation is derived from predicted moisture tendency. In the independent data test, ResCu can accurately reproduce the SPCAM simulation in both time mean and temporal variance. Comparison with other neural networks demonstrates the superior performance of ResNet architecture. ResCu is further tested in a single‐column model for both continental midlatitude warm season convection and tropical monsoonal convection. In both cases, it simulates the timing and intensity of convective events well. In the prognostic test of tropical convection case, the simulated temperature and moisture biases with ResCu are smaller than those using conventional convection and cloud parameterizations.
Plain Language Summary
Moist physics parameterization, an algorithm for clouds and convection based on empirical relationships and limited observations, is the main source of biases in rainfall and atmosphere circulation in global climate model simulations. A sophisticated deep learning algorithm with refined inherent architecture is introduced as a new parameterization in this study. It is trained with a year‐long simulation from a global climate model that has a two‐dimensional high‐resolution cloud‐scale model embedded in each climate model grid box, where clouds and convection are explicitly simulated. In addition, we consider energy conservation and past history of atmospheric states, clouds, and convection. It is designed to predict heating and moistening rates from convection and clouds, as well as cloud water and ice amount. This new parameterization accurately reproduces target simulations in 1‐year independent testing and also performs well in predicting convective events in both midlatitude summer continental land convection and tropical monsoon convection.
Key Points
An advanced deep residual convolutional neural network is used for moist physics parameterization
The neural network is trained with a 1‐year‐long simulation from superparameterized CAM5 with actual global land‐ocean distribution
It reproduces accurately the target simulation in independent test data evaluation and in single‐column model prognostic validations