Jef Caers

Biography

Jef Caers received both an MSc (’93) in mining engineering / geophysics and a PhD (’97) in engineering from the Katholieke Universiteit Leuven, Belgium. Currently, he is Professor of Geological Sciences (since 2015) and previously Professor of Energy Resources Engineering at Stanford University, California, USA. He is also director of the Stanford Center for Reservoir Forecasting, an industrial affiliates program in reservoir modeling and geostatistics with ~20 partners from the Energy Industry. Dr. Caers’ research interests are in the area of geostatistics, spatial modeling and modeling uncertainty applied to various areas in the Earth Sciences. He was awarded the Vistelius award by the IAMG in 2001, is Editor-in-Chief of Computers and Geosciences and served as chairman for the IAMG 2009 conference. Dr. Caers has received several best paper awards and written three books entitled "Petroleum Geostatistics” (SPE) “Modeling Uncertainty in the Earth Sciences” (Wiley-Blackwell) and "Multiple-point Geostatistics: stochastic modeling with training images" is published with Wiley-Blackwell in 2014. Dr. Caers was awarded the 2014 Krumbein Medal of the IAMG for career achievement.

Research Interest

My research occurs at the intersection of statistical science, computer science and the geological sciences. What is the fundamental research question I want to address? I believe that from a data-scientific point of view, most geological data and modeling questions can be broadly classified as problems that are high-dimensional but have small sample size. Data are often sparse and computer experiments we run are CPU demanding, resulting in some low sample size. Yet the understanding we attempt to develop requires complex physical or geochemical models, analysis of multivariate, spatial problems over potentially large areas, require aggregation of data at various scale (in space and time) and hence are high dimensional problems. How do we formulate such problems? What are fundamental mathematical and computer science methods for analyzing such problems? How can we build predictive models for such problems? How do we integrate the various disciplines involved? Most of machine learning and statistics research currently does not take place in this setting.