PhD and postdoc opportunities in computationally oriented and mathematical hydrogeology

I am a faculty member at Ben-Gurion University exploring fundamental questions concerning solute transport, heterogeneity, and uncertainty in the subsurface. I am looking to collaborate with scientists of all backgrounds with strong quantitative skills and an interest in model development and fundamental conceptual understanding. Full multi-year funding is available for postdocs and PhD students interested in a number of challenging projects.

You can find out more about my research by searching my ORCID (https://orcid.org/0000-0001-8022-0123), or consulting my website (http://www.scottkhansen.info). Three available projects are listed below (this is not an exhaustive list of current projects or opportunities). To explore opportunities, please send a CV and expression of interest to skh@bgu.ac.il.

Explaining geochemical reaction rates using heterogeneity statistics

In situ geochemical reaction rates are commonly orders of magnitude slower than those measured under well-mixed conditions in the laboratory, a result of Ovchinnikov-Zeldovich reactant segregation. No general theory exists for quantifying the degree that the upscaled effective reaction rate is attenuated due to this segregation under realistic subsurface conditions. To address this important question, we are employing a combination of a combination of analysis, numerical simulation, and machine learning techniques to represent the field-scale effective reaction rate for general A + B → C aqueous-phase reactions as a function of geostatistical measures of field-scale subsurface heterogeneity and laboratory kinetic measurements.

How much information do remote point measurements contain?

Model selection and calibration is a major concern in the geosciences: high-dimensional parameter spaces representing space-random fields typically need to be estimated from data in low-dimensional observation spaces. It is not clear: (i) how much information the measurements contain about the underlying field(s), (ii) the optimal model resolution to optimize predictive validity (the so-called “information criteria” commonly used for this purpose have been shown to perform poorly on realistic hydrologic model selection problems), and (iii) how structural errors introduced by model simplification contaminate forward estimates (classical variance-based techniques assuming zero-mean additive measurement noise are not suitable). We are working on some novel analytical and numerical approaches to these problems.

Using non-Fickian transport models with deterministic information

A major obstacle to the adoption of realistic non-Fickian transport models has been their incompatibility with partial, large-scale, spatially heterogeneous deterministic flow information. We are developing both theory and the Aurora software (https://aurora-hub.gitlab.io/) to solve this problem.