COOKIES: By using this website you agree that we can place Google Analytics Cookies on your device for performance monitoring. |
University of Cambridge > Talks.cam > Bullard Laboratories Wednesday Seminars > Radical uncertainty in subsurface science: How do we make decisions?
Radical uncertainty in subsurface science: How do we make decisions?Add to your list(s) Download to your calendar using vCal
If you have a question about this talk, please contact Yihe Xu. The impact of subsurface science can often be assessed by decisions that it affects. Scientists decide whether to adopt newly proposed models, scientific evaluation of the likelihood of hazards impacts on the decision between mitigating actions, and estimates of subsurface properties and dynamics affect economically important decisions. In all cases an assessment of uncertainty in the scientific results allows decisions to be better informed, and more robust to possible variations. Seismic Tomography is a method to image the Earth’s subsurface, and is used to inform all three types of decisions outlined above. In order to better interpret the resulting images it is important to assess imaging uncertainties, but this is hard to achieve. Monte Carlo random sampling methods are often applied for this purpose but the ‘curse of dimensionality’ makes them computationally intractable for high-dimensional parameter spaces. This talk extends uncertainty analysis to larger systems by introducing new variational inference methods to seismic tomography. In contrast to random sampling, variational methods solve an optimization problem yet provide probabilistic results. Even the most demanding uncertainty assessments, in three-dimensional full seismic waveform tomography, are achievable to some extent. A remaining difficulty is that geoscientific models are always wrong, and are usually biased in unknown ways. Real world uncertainties beyond the modelled domain are therefore unknown and potentially are unknowable. In a world of so-called radical uncertainty, in which true probabilities are unable to be assessed, a new approach to decision making will be required, as will be discussed. This talk is part of the Bullard Laboratories Wednesday Seminars series. This talk is included in these lists:
Note that ex-directory lists are not shown. |
Other listsother talks Genomics and Evolutionary Genetics Club Cambridge University Biological SocietyOther talksGrand Rounds Simple and multiple linear regression Constraints on DM and the cusp/core problem with MUSE Two and three dimensional diffusion processes with helical persistence Machine Learning Integrability in 1D and 2D Models of Gravity Fast Tagging of Pollinator Field Videos with Convolutional Tsetlin Machines |