Inverse problems in meteorology and seismology
Aimé Fournier
Mesoscale & Microscale Meteorology Laboratory (MMM), National Center for Atmospheric Research (NCAR)
Date and time:
Friday, April 10, 2015 - 3:00pm
Location:
ECCR 265
Abstract:
In this talk I will review recent work on gigabyte-scale inverse problems, namely: data assimilation for regional meteorology; and anisotropic exploration seismology. The solution to these and other problems partly unifies the approaches of regularization, Bayesian inference, linear-model preconditioning and truncated SVD. Selected results and delivered products include: compressed representation of meteorological measurement influence; uncertainty quantification for faults, traps and wells in oil & gas exploration; and formal decision analysis and value of information. I will conclude by discussing potential new research directions: compression of data assimilation for high-order non-uniform grids e.g., spectral elements; and value of information for wind-turbine farm simulation and management.