Title: Large-scale computer model emulation, calibration and data fusion

Abstract:

This project is motivated by fusing disparate field observations and computer model simulations to estimate the natural systems and quantify uncertain hazard. We will introduce fast and accurate emulators to approximate large-scale computer model simulation, which allows the use of a computationally expensive computer model in probabilistic data inversion. In some scenarios, even sophisticated models cannot fully reproduce the complexity of natural system. New nonparametric models of the discrepancy between the reality and computer models will be discussed. We will also introduce methods of fusing the disparate number of observations in different types, and models for estimating measurement bias induced by equipment and environmental conditions. The data sets include satellite radar interferrograms, GPS, tilt data, height of lava lake and computer model output in our study of 2018 Kilauea Volcano eruption in Hawaii. We will also discuss the performance of our methods on NOAA global temperature anomalies and DNA methylation level imputation. Challenges and some ongoing research projects will be outlined in this talk.

Bio:

Mengyang Gu is an assistant professor in the Department of Statistics and Applied Probability at the University of California, Santa Barbara. His research focuses on statistical uncertainty quantification, computer model emulation, inverse problem, dimension reduction, tensor methods and spatio-temporal models. He received his BS degree in statistics at Zhejiang University in 2012, MS degree and PhD in statistical science at Duke University in 2014 and 2016, respectively.

Website