NG005
Data-Driven Subgrid-Scale Parameterizations for Earth System Modeling

Session ID#: 159885

Session Description:
Subgrid-scale (SGS) parameterizations estimate effects of unresolved processes without modeling them directly. However, traditional physics-based parameterizations are often a large source of uncertainty in Earth system models (ESMs) such as general circulation models (GCMs).  The recent advances in big data, machine learning/artificial intelligence (ML/AI), and hardware (i.e. GPUs) present new methods coupled with enhanced computational capabilities, which we can utilize for scientific applications. Specifically, we are interested in developing scientific ML/AI to improve Earth system modeling by replacing or modifying existing parameterizations. 

We solicit contributions on various aspects of marrying data-driven methods to parameterizations for models, from all weather and climate modeling communities (ocean, atmosphere, chemistry, land, sea-ice, vegetation, etc.). Topics include but are not limited to:

  • Type of data used: types of observations, high-resolution model outputs, remote sensing products.
  • Types of ML/AI (or other) models and data processing used: computer vision techniques, neural networks, filtering methods, data assimilation techniques for inverse problems. 
  • Type of SGS parameterization: cloud, convection, gravity wave, other sub-grid scale processes. 
  • Does the development of data-driven models include physical knowledge to ensure stability/conservation laws? How do such models  deal with partially resolved phenomena? 
  • What metrics should be used to evaluate parameterization accuracy and how should model sensitivities and biases be treated?
  • Online coupling of data-driven parameterization in ESMs: accuracy of representation, computational challenges, generalization to new climate scenarios.
Co-Organized:
  • A - Atmospheric Sciences
  • OS - Ocean Sciences
Index Terms:

1942 Machine learning [INFORMATICS]
3337 Global climate models [ATMOSPHERIC PROCESSES]
3365 Subgrid-scale (SGS) parameterization [ATMOSPHERIC PROCESSES]
4568 Turbulence, diffusion, and mixing processes [OCEANOGRAPHY: PHYSICAL]

Primary Convener:  Laura Mansfield, Stanford, Earth System Science, Stanford, CA, United States
Conveners:  Lucia Minah Yang, New York University, Center for Atmosphere Ocean Science, New York, NY, United States, Oliver Dunbar, California Institute of Technology, Pasadena, CA, United States and Aakash Sane, Princeton University, Princeton, United States
 
Bayesian Machine Learning Method for Simultaneous Parameter Inference and Model Closure of Dynamical Systems: Application to Two-Scale Lorenz ’96 System (1115904)
Mohamed Aziz Bhouri, Columbia University, Earth and Environmental Engineering, New York, United States and Pierre Gentine, Columbia University, Department of Earth and Environmental Engineering, New York, NY, United States
 
Seasonal Surface Eddy Mixing in the Kuroshio Extension: Estimation and Machine Learning Prediction (1135991)
Ru Chen1, Wenting Guan1, Hong Zhang2, Yi Yang1 and Hao Wei1, (1)Tianjin University, School of Marine Science and Technology, Tianjin, China, (2)University of California at Los Angeles, Los Angeles, United States
 
Making Bayesian Optimization Accessible: Using Bayesian Optimization for Anything (BOA) to Minimize Canopy Resistance in an LES Model (1141834)
Madeline Scyphers, Ohio State University Main Campus, Columbus, United States, Justine E.C. Missik, Ohio State University Main Campus, Civil, Environmental & Geodetic Engineering, Columbus, United States, Theresia Yazbeck, Ohio State University Main Campus, Columbus, OH, United States, Gil Bohrer, The Ohio State University, Department of Civil, Environmental, and Geodetic Engineering, Columbus, United States and Joel Paulson, Ohio State University, Columbus, United States
 
Machine Learning-Based Causally Informed Atmospheric Parametrizations for Climate Models (1143540)
Veronika Eyring1,2, Arthur Grundner1, Fernando Iglesias-Suarez1, Tom Beucler3, Pierre Gentine4, Giorgetta A. Marco5, Mike S Pritchard6, Jakob Runge7 and Mierk Schwabe1, (1)German Aerospace Center DLR Oberpfaffenhofen, Institute of Atmospheric Physics, Oberpfaffenhofen, Germany, (2)University of Bremen, Institute of Environmental Physics, Bremen, Germany, (3)University of Lausanne, Institute of Earth Surface Dynamics, Lausanne, Switzerland, (4)Columbia University, Department of Earth and Environmental Engineering, New York, NY, United States, (5)Max Planck Institute for Meteorology, Hamburg, Germany, (6)University California Irvine, Earth System Science, Irvine, United States, (7)German Aerospace Center DLR Jena, Institute of Data Science, Jena, Germany
 
Closed-form Discovery of Subgrid-Scale Closures: Promises and Challenges (1145085)
Karan Jakhar1, Yifei Guan1, Ashesh Kumar Chattopadhyay1, Rambod Mojgani2 and Pedram Hassanzadeh3, (1)Rice University, Houston, TX, United States, (2)Rice University, Houston, United States, (3)Rice University, Department of Mechanical Engineering and Department of Earth, Environmental and Planetary Sciences, Houston, TX, United States
 
Deep Learning Parameterization of Tropical Cyclone Boundary Layer, the offline and online results (1146367)
Leyi Wang and Zhe-Min Tan, Nanjing University, Nanjing, China
 
Infrasound-driven stochastic gravity wave parameterizations (1148636)
Christophe Millet, CEA Commissariat à l'Energie Atomique DAM, Arpajon Cedex, France and Francois Lott, Laboratoire de Météorologie Dynamique-IPSL CNRS/PSL-ENS/Sorbonne Université/Ecole Polytechnique, Paris, France
 
Machine learned surrogate model for subgrid-scale turbulence effects in supercell flows using the miniWeatherML app (1154007)
Matthew R Norman, Oak Ridge National Laboratory, Oak Ridge, TN, United States and Muralikrishnan Gopalakrishnan Meena, Oak Ridge National Lab, Oak Ridge, TN, United States
 
Confronting the offline vs. online skill dilemma via prognostic testing of neural network convection parameterizations at a computationally ambitious scale (1165523)
Jerry Lin, Irvine, CA, CA, UNITED STATES, Mike S Pritchard, University California Irvine, Earth System Science, Irvine, United States, Sungduk Yu, University of California Irvine, Earth System Science, Irvine, United States, Tom Beucler, University of Lausanne, Institute of Earth Surface Dynamics, Lausanne, Switzerland and David Walling, Alaffia Health, New York, United States
 
A Deep Learning PDF Parameterization of Subgrid-Scale Clouds and Turbulence for GCMs (1166033)
Katherine Breen1, Donifan Barahona2, Nathan Arnold3 and William Putman2, (1)NASA Goddard Space Flight Center, Greenbelt, MD, United States, (2)NASA GSFC, Greenbelt, MD, United States, (3)Colorado State University, Fort Collins, CO, United States
 
Equatorial Pacific Ocean large eddy simulations enable data-driven parameterization and understanding of ocean mixing (1168825)
Daniel B Whitt, NASA Ames Research Center, Moffett Field, CA, United States and Deepak Cherian, NCAR, Boulder, United States
 
Recreating Observed Lower-Stratospheric Convective Gravity Waves from Weather Radar Observations using a Neural-Network-Forced WRF Model (1176716)
Christopher Gary Kruse1, M Joan Alexander1, Martina Bramberger2, Pedram Hassanzadeh3, Ashesh Kumar Chattopadhyay4, Brian Green5 and Alison Grimsdell6, (1)NorthWest Research Associates Boulder, Boulder, CO, United States, (2)NorthWest Research Associates Boulder, Boulder, United States, (3)Rice University, Houston, United States, (4)Rice University, Houston, TX, United States, (5)Stanford University, Department of Earth System Science, Stanford, CA, United States, (6)NorthWest Research Associates, Boulder, United States
 
Stochastic Parameterization of the Ocean Surface Boundary Layer using Generative Adversarial Networks (1177012)
Balu Nadiga, Los Alamos National Laboratory, Los Alamos, NM, United States and Luke Van Roekel, Los Alamos National Laboratory, Fluid Dynamics and Solid Mechanics Group, Los Alamos, United States
 
NVIDIA’s Earth-2: The “Hop, Skip, Tune, Leap” strategy for accelerated km-scale hybrid ML-physics climate prediction (1182896)
Mike S Pritchard1, Karthik Kashinath1, Anima Anandkumar1, Morteza Mardani1, David Hall2, Jaideep Pathak1, Thorsten Kurth1, Peter Messmer1 and Stan Posey1, (1)NVIDIA Corporation, Santa Clara, United States, (2)NVIDIA Corporation, Lafayette, United States
 
Modeling the Ocean Mesoscale Eddy Effects on Tracer Transport (1189596)
Yueyang Lu, University of Miami, Miami, FL, United States and Igor V Kamenkovich, RSMAS, Miami, FL, United States
 
A data constrained method to parameterize convective gravity waves in global circulation models (1190469)
Martina Bramberger, NorthWest Research Associates Boulder, Boulder, United States, M Joan Alexander, NorthWest Research Associates Boulder, Boulder, CO, United States, Julio T Bacmeister, AMP/CGD, Boulder, CO, United States, Albert Hertzog, Laboratoire de Météorologie Dynamique-IPSL CNRS/PSL-ENS/Sorbonne Université/Ecole Polytechnique, Palaiseau Cedex, France, Milena Corcos, Laboratoire de Météorologie Dynamique Palaiseau, Palaiseau Cedex, France and Chuntao Liu, Texas A & M University Corpus Christi, Corpus Christi, TX, United States
 
Neural network achieves high accuracy and energy conservation in the longwave radiative transfer (1192173)
Chenggong Wang, Princeton University, Atmospheric and Oceanic Sciences, Princeton, United States and Jing Feng, Princeton University, Princeton, United States
 
Diagnosing and Parameterizing Eddy Driven Transport in a Zonally Inhomogeneous Flow (1201601)
Dhruv Balwada, Courant Institute of Mathematical Sciences, NEW YORK, NY, United States, Ryan Abernathey, Lamont-Doherty Earth Observatory, Palisades, NY, United States, K. Shafer Smith, New York University, Courant Institute of Mathematical Sciences, New York, NY, United States and Takaya Uchida, Florida State University, Tallahassee, United States
 
Regression Tree Ensembles for Data-Driven Gravity Wave Parameterizations (1050518)
David S Connelly, New York University, Center for Atmosphere Ocean Science, New York, United States and Edwin P Gerber, New York University, Courant Institute of Mathematical Sciences, New York, NY, United States
 
Radiation Parameterization Emulator Development for Korean Integrated Model (1057872)
Wooyeon Park, Sojung An, Jiyeon Jang, Inchae Na, Tae-Jin Oh and Junghan Kim, KIAPS Korea Insititute of Atmospheric Prediction Systems, Seoul, South Korea
 
Stochastic-Deep Learning Parameterization of Subgrid Ocean Processes in the MOM6 Ocean Model (1059632)
Cheng Zhang1, Pavel Perezhogin2, Cem Gultekin2, Alistair Adcroft3 and Laure Zanna4, (1)Princeton University, Princeton, NJ, United States, (2)New York University, New York, United States, (3)Princeton University, Program in Atmospheric and Oceanic Sciences, Princeton, United States, (4)New York University, Courant Institute of Mathematical Sciences, New York, United States
 
Accelerated Optimal Data Acquisition When Learning Subgrid-Scale Parameters in an Idealized GCM (1059964)
Oliver Dunbar1, Michael Howland2, Tapio Schneider1 and Andrew Stuart1, (1)California Institute of Technology, Pasadena, CA, United States, (2)Massachusetts Institute of Technology, Civil and Environmental Engineering, Cambridge, United States
 
Deep Learning for Subgrid-Scale Turbulence Modeling in Large-Eddy Simulations of the Convective Atmospheric Boundary Layer (1074077)
Yu Cheng1, Marco Giovanni Giometto2, Pit Kauffmann3, Ling Lin3, Chen Cao3, Cody Zupnick3, Harold Li4, Qi Li5, Yu Huang6, Ryan Abernathey7 and Pierre Gentine8, (1)Harvard University, Department of Earth and Planetary Sciences, Cambridge, MA, United States, (2)Columbia University, Department of Civil Engineering and Engineering Mechanics, New York, United States, (3)Columbia University, Capstone Project at Data Science Institute, New York, NY, United States, (4)Department of Earth and Environmental Engineering, Columbia University, New York, NY, United States, (5)Cornell University, Civil and Environmental Engineering, Ithaca, NY, United States, (6)Columbia University, Earth and Environmental Engineering, New York, United States, (7)Lamont-Doherty Earth Observatory, Palisades, NY, United States, (8)Columbia University, Department of Earth and Environmental Engineering, New York, NY, United States
 
Exploring the use of machine learning to parameterize vertical mixing in the ocean surface boundary layer (1074483)
Junhong Liang1, Jianguo Yuan2, Xiaoliang Wan3, Jinliang Liu4, Bingqing Liu5, Mayank Tyagi3 and Hakun Jang3, (1)Louisiana State University, Department of Oceanography and Coastal Sciences, Baton Rouge, LA, United States, (2)Louisiana State University, Oceanography and Coastal Sciences, Baton Rouge, United States, (3)Louisiana State University, Baton Rouge, United States, (4)Louisiana State University, Department of Oceanography and Coastal Sciences, Baton Rouge, United States, (5)The Water Institute of the Gulf, Baton Rouge, United States
 
A Data-Driven Approach for the Submesoscale Parameterization (Invited) (1075502)
Abigail S. Bodner, New York University, Courant Institute of Mathematical Sciences, New York, NY, United States and Laure Zanna, New York University, New York, NY, United States
 
Explainable transfer learning for generalizable subgrid-scale modeling (1076121)
Pedram Hassanzadeh1, Adam Subel2, Ashesh Kumar Chattopadhyay3, Yifei Guan3, Laure Zanna4 and Andrew Ross5, (1)Rice University, Department of Mechanical Engineering and Department of Earth, Environmental and Planetary Sciences, Houston, TX, United States, (2)Rice University, Mechanical Engineering, Hosuton, TX, United States, (3)Rice University, Houston, TX, United States, (4)New York University, New York, NY, United States, (5)NYU, New York City, United States
 
A symbolic data-driven parameterization in an idealized configuration of a primitive equation model (1076794)
Laure Zanna, New York University, Courant Institute of Mathematical Sciences, New York, United States, Pavel Perezhogin, New York University, New York, United States and Alistair Adcroft, Princeton University, Program in Atmospheric and Oceanic Sciences, Princeton, United States
 
Reparameterization Required After Model Structure Changes From Carbon Only to Carbon-Nitrogen Coupling (1088546)
Song Wang, Chinese Academy of Sciences, Beijing, China, Yiqi Luo, Department of Biological Sciences, Northern Arizona University, Flagstaff, United States and Shuli Niu, Institute of Geographic Sciences and Natural Resources Research, Chinese Academy of Sciences, Beijing, China
 
Sensitivity Analysis and Machine Learning of a Sea Ice Melt Pond Parametrisation (1092507)
Simon Driscoll1, Alberto Carrassi2,3, Marc Bocquet4, Laurent Bertino2, Julien Brajard5 and Einar Olason2, (1)University of Reading, Department of Meteorology, Reading, United Kingdom, (2)Nansen Environmental and Remote Sensing Center, Bergen, Norway, (3)University of Bologna, Bologna, Italy, (4)Ecole des ponts ParisTech, Paris, France, (5)Sorbonne Université, Sorbonne, France
 
Comparison of PBL Parameterization Tendency Emulator Performance with Different Machine Learning Architectures. (1103057)
Jiyeon Jang, Tae-Jin Oh, Sojung An, Wooyeon Park, Inchae Na and Junghan Kim, KIAPS Korea Insititute of Atmospheric Prediction Systems, Seoul, South Korea
 
Data-Driven Aeolian Dust Emission Scheme for Climate Modelling (1107429)
Klaus Klingmueller1 and Johannes Lelieveld1,2, (1)Max Planck Institute for Chemistry, Mainz, Germany, (2)The Cyprus Institute, Nicosia, Cyprus
 
Physics to machine learning, and machine learning back to physics (Invited) (1107485)
Pierre Gentine1, Sara Shamekh2, Alex Connolly3, Tom Beucler4, Veronika Eyring5, Mike S Pritchard6, Rose Yu7, Kara Lamb8 and Gunnar Behrens5, (1)Columbia University, Department of Earth and Environmental Engineering, New York, NY, United States, (2)Columbia University, Earth and Environmental Engineering, New York, United States, (3)University of California Berkeley, Berkeley, CA, United States, (4)Massachusetts Institute of Technology, Cambridge, United States, (5)German Aerospace Center DLR Oberpfaffenhofen, Institute of Atmospheric Physics, Oberpfaffenhofen, Germany, (6)University California Irvine, Earth System Science, Irvine, United States, (7)UCSD, San Diego, United States, (8)Cooperative Institute for Research in Environmental Sciences, Boulder, United States
 
Sampling Strategies for Training Machine Learning Emulators of Gravity Wave Momentum Transport (1120340)
Lucia Minah Yang, New York University, Center for Atmosphere Ocean Science, New York, NY, United States and Edwin P Gerber, New York University, Courant Institute of Mathematical Sciences, New York, NY, United States
 
Parameterizing Vertical Turbulent Mixing Coefficients In The Ocean Surface Boundary Layer Using Neural Networks (1130761)
Aakash Sane1, Brandon G Reichl2, Alistair Adcroft1 and Laure Zanna3, (1)Princeton University, Program in Atmospheric and Oceanic Sciences, Princeton, United States, (2)NOAA Geophysical Fluid Dynamics Laboratory, Princeton, United States, (3)New York University, Courant Institute of Mathematical Sciences, New York, United States
See more of: Nonlinear Geophysics
<< Previous Session Proposal | Next Session Proposal >>