Abstract

The session consists of four talks on cutting-edge Bayesian models and methods in environmental applications. The speakers, three male and one female, are at different stages of their careers, come from three different continents, and bring perspectives from both academia and scientific agencies.

Michael Bertolacci is Senior Lecturer with the University of Western Australia with expertise in spatio-temporal modelling for environmental applications. He will present GeoWarp, a hierarchical 3D spatial model for characterising seabed sediment that is fitted using geotechnical data inside a Bayesian framework. The approach provides a way forward to reduce costs in site characterisation for large offshore projects such as wind farms.

Dan Pagendam is Senior Research Scientist with the Commonwealth Scientific and Industrial Research Organisation. He is leader of the Hybrid Modelling team within CSIRO's Data61 with expertise in physical-statistical modelling in environmental settings. He will present CQUESST, a physical-statistical Bayesian model developed by an inter-disciplinary team for making inference in the area of soil-carbon sequestration.

Julia Walchessen is a final-year PhD student at Carnegie-Mellon University. She investigates how neural networks can be used to accelerate conditional simulation of latent spatial processes, which is a key component in many Bayesian algorithms that involve complex spatial process models. Her work employs so-called "diffusion models" and is used for making inference on ocean variables from Argo data.

Raphaël Huser is Associate Professor with the King Abdullah University of Science and Technology (KAUST) in Saudi Arabia, with expertise on modelling spatial extremes. His talk will centre on the use of neural networks to accelerate inference from spatial extremal data. He applies the methodology to analysing air pollution extremes in Saudi Arabia.