GEOFACES

The project GEOFACES (GEOphysics-based Falsification and Corroboration in the Earth Sciences) is funded by the Swiss National Science Foundation. It aims ultimately at developing a general approach to falsify and corroborate conceptual Earth science models using multi-method geophysical data.

Open PhD positions

Sub-project A: Geophysical inversion with complex geological priors using deep learning and multiple-point statistics
Probabilistic non-linear inversions aiming at obtaining posterior model realizations with high geological realism are often exceedingly slow for large-scale 3-D Earth science applications. How can geophysical inversion with complex priors derived from training images be made more efficient and what simplifying assumptions can be made to achieve faster, albeit approximative, methods that scale well with data-, model- and space dimensions? These challenges will be addressed using recent advances in multiple-point statistics and deep learning. Project partners are Prof. Grégoire Mariethoz (University of Lausanne) and Dr. Eric Laloy (Belgian Nuclear Research Centre).

Sub-project B: Accounting for petrophysical and upscaling errors in probabilistic and deterministic geophysical inversions

The impact of petrophysical (rock physics) uncertainty and upscaling errors may dominate over those caused by observational data errors; yet these sources of uncertainty are seldom considered in geophysical inversions. How can modern statistical methods and
linearized approximations be used to obtain efficient probabilistic and deterministic inversion methods that account for petrophysical uncertainty and provide meaningful estimates of parameter uncertainty? Similarly, how can we best account for upscaling errors in non-linear inversion problems?
These challenges will initially be addressed using the correlated pseudo-marginal Markov chain Monte Carlo method and more approximate linearized approaches. Project partners are Prof. Arnaud Doucet (Oxford University) and Prof. David Ginsbourger (IDIAP, University of Bern). A one-month stay at Oxford University is planned.

Sub-project C: Geophysics-based falsification and corroboration with emphasis on frozen ground dynamics in an alpine catchment

Robust and reliable geophysics-based ranking of competing conceptual models is only possible when considering all leading sources of uncertainty. How can we develop a flexible and practical framework that enables geophysics-based falsification and corroboration of conceptual Earth science models that is applicable to a wide range of Earth science problems? We will follow a Bayesian model selection framework within which we will integrate findings from sub-projects A and B. The leading flagship example will rely on data acquired at the Niwot Ridge Long-term Ecological Research Site http://niwot.colorado.edu/) in Colorado. Geophysical data, data on soil properties and water content, as well as stream discharge and chemistry will be used to address a number of scientific hypotheses using Bayesian model selection tools, for instance: Which type of spatial and temporal patterns best describe the patchiness of frozen ground? Does hydrological flowpaths vary over time as a result of the distribution of frozen ground? The main project partner is Prof. Kamini Singha (Colorado School of Mines). It is anticipated that the PhD student will apply for a mobility grant to spend one year with Prof. Singha’s group.

How to apply
The PhD positions are funded for four years and the starting date is September 1, 2019, or a slightly earlier or later date to be decided upon. Successful candidates should ideally hold a MSc degree in geophysics, physics, statistics, mathematics or quantitative Earth Sciences. He/she should have a keen interest and preferably some experience in Bayesian statistics, scientific computing, geophysical forward modeling and inversion. Field experience and knowledge about hydrogeological, critical zone or permafrost processes are beneficial for sub-project C. We are a dynamic international research group working on a wide variety of topics in environmental and computational geophysics.

To apply, please send a cover letter clarifying your overall motivation for entering a PhD program together with your curriculum vitae and the names, telephone numbers, and e-mail addresses of two referees to Prof. Niklas Linde, Institute of Earth Sciences, Géopolis 3779, University of Lausanne, 1015 Lausanne, Switzerland or by e-mail to niklas.linde@unil.ch. The application deadline is May 15, 2019.