AGU Session - A028: Combining Physical Simulation and Machine Learning across Geophysical Sciences
We would like to draw your attention to AGU session A028 - "Combining Physical Simulation and Machine Learning across Geophysical Sciences." This is a broad session intended to cross disciplinary boundaries and therefore listed under Atmospheric Sciences and cross-listed with Biogeosciences, Hydrology, and Seismology. Our intention is to survey the breath of the Earth Sciences on their approaches toward informing physical simulation from machine learning.
Simulation of physical processes through solution of differential equations, or mechanistic models, serves as a core tool of geophysical studies. In contrast to this cause and effect-driven Physical Simulation (PS), Machine Learning (ML) models instantiate pattern recognition techniques and often operate as a black box. A key question is how to combine PS and ML solutions to advance understanding of the physical system and improve the PS. This session is seeking cross-disciplinary presentations that demonstrate applied combinations of PS and ML. Examples may include data assimilation (where ML may drive the design of PS), using ML to filter PS outputs, use of PS to develop training data sets for ML and ML approaches to classification of large scale PS outputs. Demonstrations of practical applications are strongly encouraged.
Sean McKenna, IBM Ireland
Ronni Grapenthin, New Mexico Tech
Anna Michalak, Carnegie Institution for Science, Global Ecology
Markus Reichstein, Max Planck Institute for Biogeochemistry
Ronni Grapenthin Assistant Professor of Geophysics
Dept. of Earth & Environmental Science New Mexico Tech 801 Leroy Pl. Socorro, NM 87801