NG004
Data Assimilation meets Machine Learning

Session ID#: 77631

Session Description:
Data Assimilation is a method to combine information from observations and from dynamical models and is firmly rooted in Bayesian Inference. Machine Learning has developed very powerful methods to analyze large data-sets, leading to unprecedented results and computational techniques that are of high interest to the geosciences. Recently machine learning has started to incorporate physical constraints which helps stabilize the algorithms and allows them to extrapolate outside the bounds of the training data. This brings machine learning very close to data assimilation, suggesting powerful mergers are possible. This session calls for research that combines the benefits of both. Examples are efforts to bring uncertainty quantification into machine learning, methods to implement physical prior knowledge into machine learning algorithms, exploring machine learning techniques in existing data-assimilation algorithms and to enhance empirically formulated dynamical models as well as forward operators, and completely new ideas on how to combine the two.
Co-Organized:
  • A - Atmospheric Sciences
  • IN - Earth and Space Science Informatics
  • OS - Ocean Sciences
Index Terms:

0520 Data analysis: algorithms and implementation [COMPUTATIONAL GEOPHYSICS]
0555 Neural networks, fuzzy logic, machine learning [COMPUTATIONAL GEOPHYSICS]
3260 Inverse theory [MATHEMATICAL GEOPHYSICS]
3315 Data assimilation [ATMOSPHERIC PROCESSES]

Primary Convener:  Peter Jan van Leeuwen, Colorado State University, Atmospheric Sciences, Fort Collins, CO, United States
Conveners:  Vipin Kumar, University of Minnesota Twin Cities, Minneapolis, MN, United States, Kayo Ide, University of Maryland, College Park, MD, United States and Alan Geer, ECMWF, Reading, United Kingdom
Primary Liaison:  Peter Jan van Leeuwen, Colorado State University, Atmospheric Sciences, Fort Collins, CO, United States
 
Employing Ensemble Data for the Tangent Linear and Adjoint Model Development (491654)
Max Yaremchuk, US Naval Research Laboratory, Stennis Space Center, MS, United States, Dmitri Nechaev, Univ Southern Mississippi, Stennis Space Center, MS, United States and Sergey Frolov, US Naval Research Laboratory, Monterey, CA, United States
 
Data Analytics for Clustering and Statistical Modeling of Oceanography Data (500412)
Rebecca Hicks, University of Central Oklahoma, Edmond, OK, United States and Sam Abuomar, Coastal Carolina University, Conway, SC, United States
 
Machine learning-based prediction of error characteristics in global wind velocity reanalyses (501280)
Christopher Irrgang, Jan Saynisch and Maik Thomas, Helmholtz Centre Potsdam GFZ German Research Centre for Geosciences, Potsdam, Germany
 
PICOSS: Python Interface for the Classification of Seismic Signals (511046)
Angel Bueno Rodriguez1, Luciano Zuccarello2, Alejandro Diaz Moreno3, Jack Woolam4, Manuel Titos Luzón1, Carmen Benitez1, Isaac Álvarez Sr.5, Janire Prudencio6, Silvio De Angelis7 and Jesús M Ibáñez8, (1)University of Granada, Signal Theory, Telematic and Communications, Granada, Spain, (2)Istituto Nazionale di Geofisica e Vulcanologia, Catania, Italy, (3)University of Liverpool, Liverpool, L69, United Kingdom, (4)University of Liverppol, Liverpool, United Kingdom, (5)University of Granada, Signal Theory, Telematics and Communications, Granada, Spain, (6)Universidad de Granada, Granada, Spain, (7)University of Liverpool, Liverpool, United Kingdom, (8)University of Granada, Instituto Andaluz de Geofísica, Granada, Spain
 
Learning Argo Profiles by using the Signature Method (513192)
Nozomi Sugiura and Shigeki Hosoda, JAMSTEC Japan Agency for Marine-Earth Science and Technology, Kanagawa, Japan
 
Machine learning accelerates parameter optimization and uncertainty quantification of a land surface model (525891)
Yohei Sawada, University of Tokyo, Bunkyo-ku, Tokyo, Japan
 
A Reduced Order Deep Data Assimilation Model (537270)
Rossella Arcucci Arcucci, Imperial College London, London, SW7, United Kingdom, César Andrés Quilodrán, Imperial College London, Space and Atmospheric Physics, London, United Kingdom and Yi-Ke Guo, Imperial College London, London, United Kingdom
 
The Learnability of Feedforward Neural Networks in Dissipative Chaotic Dynamical Systems (550326)
Ziwei Li, Massachusetts Institute of Technology, Earth, Atmospheric, and Planetary Sciences, Cambridge, MA, United States and Sai Ravela, Massachusetts Institute of Technology, Cambridge, MA, United States
 
Big Data Assimilation Incorporating Deep Learning with Phased Array Radar Data and Numerical Weather Prediction (575900)
Takemasa Miyoshi1, Marimo Ohhigashi2, Shigenori Otsuka3, Takumi Honda3, Arata Amemiya2, Yasumitsu Maejima3, Shunji Kotsuki3, Yoshihiro Ishikawa4, Hiromu Seko5, Yoshito Yoshizaki6, Naonori Ueda7, Hirofumi Tomita8, Yutaka Ishikawa2, Shinsuke Satoh9, Tomoo Ushio10, Kana Koike11 and Yasuhiko Nakada12, (1)RIKEN, Wako, Japan, (2)RIKEN, Kobe, Japan, (3)RIKEN Center for Computational Science, Kobe, Japan, (4)Japan Meteorological Agency, Tokyo, Japan, (5)Meteorological Research Institute, Tsukuba, Japan, (6)Japan Meteorological Agency, Kiyose, Japan, (7)RIKEN, Tokyo, Japan, (8)RIKEN Advanced Institute for Computational Sciences, Kobe, Japan, (9)National Institute of Information and Communications Technology, Koganei, Japan, (10)Tokyo Metropolitan University, Tokyo, Japan, (11)MTI Co. Ltd., Tokyo, Japan, (12)Tokyo Electric Power Company Holdings, Tokyo, Japan
 
Assimilation of satellite fire detections in coupled atmosphere-fire model WRF-SFIRE by machine learning (594425)
Jan Mandel1, Angel Farguell1, James Haley1, Lauren Hearn1, Adam Kochanski2 and Kyle A Hilburn3, (1)University of Colorado Denver, Mathematical and Statistical Sciences, Denver, CO, United States, (2)University of Utah, Atmospheric Sciences, Salt Lake City, UT, United States, (3)Colorado State University, Fort Collins, CO, United States
 
Informative Learning: A Nonlinear Dynamical Systems Approach to Deep Learning (601299)
Sai Ravela1, Sonia Reilly2, Ziwei Li3 and Aarti K Dwivedi1, (1)Massachusetts Institute of Technology, Cambridge, MA, United States, (2)Massachusetts Institute of Technology, Mathematics, Cambridge, United States, (3)Massachusetts Institute of Technology, Earth, Atmospheric, and Planetary Sciences, Cambridge, MA, United States
 
A Particle Flow data-assimilation method for high-dimensional systems (620032)
Peter Jan van Leeuwen, Colorado State University, Atmospheric Sciences, Fort Collins, CO, United States and Manuel A Pulido, University of Reading, Meteorology, Reading, United Kingdom
 
Implementation of Replica Exchange Monte Carlo on 4DVar for Global Optimization (624799)
Hiromichi Nagao1, Shin-ichi Ito1, Kei Hasegawa1, Masayuki Kano2, Masato Okada3, Hideitsu Hino4, Kenji Nagata5 and Naoshi Hirata1, (1)Earthquake Research Institute, University of Tokyo, Tokyo, Japan, (2)Graduate School of Science, Tohoku University, Sendai, Miyagi, Japan, (3)The University of Tokyo, Graduate School of Frontier Sciences, Chiba, Japan, (4)ISM Institute of Statistical Mathematics, Tokyo, Japan, (5)National Institute for Materials Science, Tsukuba, Japan
See more of: Nonlinear Geophysics
<< Previous Session Proposal | Next Session Proposal >>