A137
Use of Machine Learning and Causal Discovery to Advance Knowledge in the Atmospheric Sciences – Methods, Limitations and Trade-offs

Session ID#: 76723

Session Description:
This session focuses on the use of machine learning (ML) algorithms to advance understanding and knowledge in the atmospheric sciences. (This is in contrast to ML methods used for pure prediction tasks.) We are hoping for a 'skeptical' ML session with a focus on means, limitations and trade-offs for gaining physical insights via ML methods, including strategies to overcome their “black-box” tendencies. We focus on five primary topics within the atmospheric sciences: (1) Use of causal discovery methods to identify potential cause-effect relationships; (2) Use of ML methods for identification of patterns and relationships in large data sets, and for identification of the most important parameters within complex systems; (3) Limitations of the above algorithms in the advancement of understanding; (4) How to navigate trade-offs between algorithm capabilities and limitations; and (5) Tools that increase transparency, visualization, interpretability and usability.

Session co-organized by IS-GEO.

Co-Organized:
  • IN - Earth and Space Science Informatics
Index Terms:

0365 Troposphere: composition and chemistry [ATMOSPHERIC COMPOSITION AND STRUCTURE]
0520 Data analysis: algorithms and implementation [COMPUTATIONAL GEOPHYSICS]
0555 Neural networks, fuzzy logic, machine learning [COMPUTATIONAL GEOPHYSICS]
3336 Numerical approximations and analyses [ATMOSPHERIC PROCESSES]

Primary Convener:  Benjamin Brown-Steiner, Atmospheric and Environmental Research, Lexington, MA, United States
Conveners:  Imme Ebert-Uphoff, Colorado State University, Fort Collins, CO, United States and Christiane Jablonowski, University of Michigan, Ann Arbor, MI, United States
Primary Liaison:  Benjamin Brown-Steiner, Atmospheric and Environmental Research, Lexington, MA, United States
 
Experiences of Machine Learning for Atmospheric Chemistry or Domain Knowledge is not a Bad Thing. (Invited) (490850)
Mathew J Evans1, Christoph Andrea Keller2, Tomás Sherwen3 and Peter Ivatt1, (1)Wolfson Atmospheric Chemistry Laboratories, Department of Chemistry, University of York, York, United Kingdom, (2)NASA Goddard Space Flight Center,, Global Modeling and Assimilation Office, Greenbelt, MD, United States, (3)Wolfson Atmospheric Chemistry Laboratories,, Department of Chemistry, University of York, York, United Kingdom
 
Cloud Characterization With Deep Learning II (492604)
Takuya Kurihana1, Ian Foster1, Elisabeth J Moyer2, Rebecca Willett1, Michael Maire1, Sydney Jenkins3, Kathryn Koenig4 and Ruby Werman5, (1)University of Chicago, Computer Science, Chicago, IL, United States, (2)University of Chicago, Geophysical Sciences, Chicago, IL, United States, (3)University of Chicago, Chicago, IL, United States, (4)University of Chicago, Harris school of Public Policy, Chicago, IL, United States, (5)University of California Berkeley, Berkeley, CA, United States
 
Lessons Learned Using ML for Knowledge Discovery in the Atmospheric Sciences (Invited) (499007)
Amy McGovern, University of Oklahoma, Computer Science, Norman, United States
 
Opening the “Black Box”: Tools to Improve Understanding of Neural Network Reasoning for Geoscience Applications (502473)
Imme Ebert-Uphoff1, Kyle A Hilburn1, Benjamin A Toms1 and Elizabeth A Barnes2, (1)Colorado State University, Fort Collins, CO, United States, (2)Colorado State University Fort Collins, Fort Collins, CO, UNITED STATES
 
Emulating critical cloud processes Using Machine Learning (506720)
Chih-Chieh Chen1, Andrew Gettelman2 and David John Gagne II1, (1)National Center for Atmospheric Research, Boulder, CO, United States, (2)NCAR, Boulder, CO, United States
 
Clustering Algorithms Provide Answers, but How Do You Know When You’ve Found the Answer? (506847)
Benjamin Brown-Steiner, Matthew James Alvarado, Karen Elena Cady-Pereira and Xiong Zhou, Atmospheric and Environmental Research, Lexington, MA, United States
 
Deep Learning Emulation and Compression of an Atmospheric Chemical System using a Chained Training Regime (508096)
Makoto Michael Kelp, Harvard University, Earth and Planetary Sciences, Cambridge, MA, United States, Nathan Kutz, University of Washington Seattle Campus, Applied Mathematics, Seattle, WA, United States, Julian D Marshall, University of Washington Seattle Campus, Civil and Environmental Engineering, Seattle, WA, United States and Christopher W Tessum, University of Illinois at Urbana Champaign, Civil and Environmental Engineering, Urbana, IL, United States
 
The Data Archive as the Dataset: Machine Learning the Metadata (508753)
Don Stott1, Omar Kaheel2 and Summer Stafford1, (1)National Center for Atmospheric Research, Boulder, CO, United States, (2)National Center for Atmospheric Research, Boulder, United States
 
Recurrent Neural Networks and the study of Internal Node Activations for the understanding of Greater California Drought and El Nino Interactions (519382)
Justin Le1, Hesham Mohamed El-Askary2, Mohamed Allai1, Thomas C. Piechota1 and Daniele Struppa3, (1)Chapman University, Orange, CA, United States, (2)Chapman Univ, Orange, CA, United States, (3)Chapman University, Schmid College of Science and Technology, Orange, CA, United States
 
Recurrent U-net: Deep learning to predict summertime ozone over the United States (532836)
Tai-Long He1, Binxuan Huang2, Yuyang Liu3, Dylan B. A. Jones1, Kazuyuki Miyazaki4, Zhe Jiang5, Edward Charles White1, Helen Marie Worden6 and John Worden7, (1)University of Toronto, Department of Physics, Toronto, ON, Canada, (2)Carnegie Mellon University, School of Computer Science, Pittsburgh, PA, United States, (3)University of Toronto, Department of Computer Science, Toronto, ON, Canada, (4)NASA Jet Propulsion Laboratory, California Institute of Technology, Pasadena, CA, United States, (5)University of Science and Technology of China, Hefei, China, (6)National Center for Atmospheric Research, Atmospheric Chemistry Observations & Modeling Laboratory, Boulder, CO, United States, (7)Jet Propulsion Laboratory, California Institute of Technology, Pasadena, CA, United States
 
Sensitivity analyses of exploiting spatial information via deep-learning from radiometer images to identify aerosols and distinct cloud types (539754)
Willem Marais, University of Wisconsin Madison, Madison, WI, United States, Robert Holz, UW SSEC, Madison, WI, United States, Jeffrey S. Reid, Marine Meteorology Division, Monterey, CA, United States and Rebecca Willett, University of Chicago, Computer Science, Chicago, IL, United States
 
Simple Machine Learning Applied to Landfalling Atmospheric Rivers at Various Lead Times (550310)
Sol Kim, University of California Berkeley, Berkeley, CA, United States and John Chiang, UC Berkeley, Geography, Berkeley, CA, United States
 
Revisiting impacts of MJO on soil moisture: a causality perspective (553391)
Diego Bueso, Universitat de València, image processing laboratory (IPL), Castelló de la Plana, Spain, Maria Piles, Universitat de València, Image Processing Laboratory (IPL), València, Spain, Gustau Camps-Valls, Universitat de València, Image Processing Laboratory (IPL), Burjassot, Spain and Eniko Székely, Swiss Data Science Center, EPFL/ETHZ, Lausanne, Switzerland
 
Using Causal Discovery Methods to Explore Subseasonal Teleconnections in a Changing Climate. (553725)
Savini Manthila Samarasinghe1, Elizabeth A Barnes2, Imme Ebert-Uphoff1 and Lantao Sun3, (1)Colorado State University, Fort Collins, CO, United States, (2)Colorado State University Fort Collins, Fort Collins, CO, UNITED STATES, (3)Colorado State University, Department of Atmospheric Science, Fort Collins, CO, US
 
Deep Neural Networks for Detection of Hydrometeors from ARM Radars (555098)
Jitendra Kumar, Oak Ridge National Laboratory, Climate Change Science Institute, Oak Ridge, TN, United States and Yuping Lu, University of Tennessee, Knoxville, United States
 
Improvement of OMI ozone profile retrieval in the lower troposphere and boundary layer by using machine learning and artificial intelligence techniques (563378)
Guanyu Huang1, Xiong Liu2, Christopher Chan Miller3 and Kelly Chance3, (1)Department of Environmental & Health Sciences, Spelman College, Atlanta, GA, United States, (2)Harvard-Smithsonian CFA, Cambridge, MA, United States, (3)Harvard-Smithsonian Center for Astrophysics, Cambridge, MA, United States
 
Understanding What the Machine Sees When Using Deep Learning for Image Segmentation to Detect Convection Initiation (582765)
Jebb Stewart, Earth System Research Laboratory, Boulder, CO, United States, Mark Govett, NOAA Earth System Research Laboratory, Boulder, CO, United States, David Hall, NVIDIA, Lafayette, CO, United States and Christina Kumler, NOAA Earth System Research Laboratory and Cooperative Institute for Research in Environmental Sciences, Boulder, CO, United States
 
The Success and Struggles of Using Deep Learning to Detect Extratropical and Tropical Cyclone Regions of Interest (ROI) in Seconds from Satellite Data (590557)
Christina Kumler, NOAA Earth System Research Laboratory and Cooperative Institute for Research in Environmental Sciences, Boulder, CO, United States, Jebb Stewart, NOAA Earth System Research Laboratory and Cooperative Institute for Research in the Atmosphere, Boulder, CO, United States, David Hall, NVIDIA, Lafayette, CO, United States and Mark Govett, NOAA Earth System Research Laboratory, Boulder, CO, United States
 
End-to-end Machine Learning Applications Framework for Earth Science (590682)
Brian Freitag1, Ashish aa0132@uaH.edu1, Muthukumaran Ramasubramanian1, Andrew Bollinger2, Iksha Gurung1, Aaron S Kaulfus1, Manil Maskey3 and Rahul Ramachandran3, (1)University of Alabama in Huntsville, Huntsville, AL, United States, (2)Development Seed, Washington, DC, United States, (3)NASA Marshall Space Flight Center, Huntsville, AL, United States
 
Towards Unsupervised Segmentation of Extreme Weather Events (596999)
Adam Rupe1, Karthik Kashinath2, Nalini Kumar3, Victor Lee3, Mr. Prabhat2 and James P Crutchfield1, (1)University of California Davis, Physics, Davis, CA, United States, (2)Lawrence Berkeley National Laboratory, Berkeley, CA, United States, (3)Intel, Santa Clara, United States
 
The Potential of Integrated Machine Learning Algorithms for Tropical Cyclone Detection in Advanced Climate Modeling (606295)
Warren Leon Davis IV1, Kevin A Reed2, Tim Shead1, Hemanth Kolla1, Gabriel Popoola3, Philip Kegelmeyer4, Aditya Konduri5 and Julia Ling6, (1)Sandia National Laboratories, Albuquerque, NM, United States, (2)National Center for Atmospheric Research, Boulder, CO, United States, (3)North Carolina A & T State University, Greensboro, United States, (4)Sandia National Laboratories, Albuquerque, United States, (5)Indian Institute of Science, Bangalore, Bangalore, India, (6)Citrine Informatics, Redwood City, CA, United States
 
Pixel Based Model For High Latitude Dust Detection (611761)
George Priftis1, Brian Freitag1, Muthukumaran Ramasubramanian1, Iksha Gurung1, Ashish aa0132@uaH.edu1, Santiago Gasso2, Rahul Ramachandran3 and Manil Maskey3, (1)University of Alabama in Huntsville, Huntsville, AL, United States, (2)GESTAR/NASA, Silver Spring, MD, United States, (3)NASA Marshall Space Flight Center, Huntsville, AL, United States
 
Data-driven Identification of Microphysical Processes by QVP-based Hierarchical Clustering of Polarised X-band Doppler Radar Observations (612252)
Maryna Lukach1,2, David Dufton3, Joshua Hampton2, Lindsay Bennett4 and Ryan R Neely III3,4, (1)University of Leeds, Leeds, LS2, United Kingdom, (2)National Centre for Atmospheric Science, Leeds, United Kingdom, (3)University of Leeds, Institute for Climate and Atmospheric Science, Leeds, United Kingdom, (4)National Centre for Atmospheric Science, Atmospheric Physics Programme, Leeds, United Kingdom
 
Developing Unsupervised Learning Models for Cloud Classification (612555)
Sydney Jenkins1, Elisabeth J Moyer2, Ian Foster3, Takuya Kurihana3, Rebecca Willett3, Michael Maire3, Kathryn Koenig4 and Ruby Werman5, (1)University of Chicago, Chicago, IL, United States, (2)University of Chicago, Geophysical Sciences, Chicago, IL, United States, (3)University of Chicago, Computer Science, Chicago, IL, United States, (4)University of Chicago, Harris school of Public Policy, Chicago, IL, United States, (5)University of California Berkeley, Berkeley, CA, United States
 
Climate-model Ensemble Consistency Testing (612799)
Dorit Hammerling, Colorado School of Mines, Golden, CO, United States, Allison Baker, National Center for Atmospheric Research, Boulder, CO, United States and Daniel Milroy
 
Machine Learning and Wildfire Burned Area: Examining the Influence of Weather and Socioeconomic Factors (613814)
Aditya Tadimeti, The Harker School, San Jose, United States and Erwan Monier, University of California Davis, Davis, CA, United States
 
What can we learn from machine learning models developed for short-term forecasting of PM2.5? (622962)
Meenakshi Rao1,2, Aaron Fellows3, Philip Orlando4 and Linda Acha George1, (1)Portland State University, Portland, OR, United States, (2)Department of Environmental Quality Oregon, Portland, OR, United States, (3)Department of Environmental Quality Oregon, Portland, United States, (4)Portland State University, Environmental Science and Management, Portland, OR, United States
 
Exploration of solving downscaling problem using deep learning-based image super-resolution techniques (633996)
Xingying Huang, University of California Los Angeles, Los Angeles, CA, United States
See more of: Atmospheric Sciences
<< Previous Session Proposal | Next Session Proposal >>