Mark Berliner, professor and chair, Department of Statistics, will give an invited presentation in the MIT Distinguished Speaker Series in Computational Science and Engineering, May 9. Berliner's talk focuses on Statistical Approaches to Combining Models and Observations.

Numerical models and observational data are critical in modern science and engineering.  Since both of these information sources involve uncertainty, the use of statistical, probabilistic methods play a fundamental role.

Berliner will discuss a general Bayesian framework for combining uncertain information and indicate how various approaches (ensemble forecasting, UQ, etc.) fit in this framework. 

He developed the framework in the context of using output from very large computer models with two illustrations: use of climate system models in the context of climate change analysis and incorporating various analyses in ocean forecasting. 

Berliner is a widely-recognized expert in the implementation of Bayesian analysis in complex settings, with particular attention to geophysical problems.

“The Bayesian paradigm provides opportunities for the combination of physical reasoning and observational data in a coherent analysis framework, but in a fashion that manages the uncertainties in both information sources,” Berliner said.

A key to the modeling is the hierarchical viewpoint in which separate statistical models are developed for the physical variables studied and the observations conditional on those variables.

Modeling physical variables in this way enables incorporation of scientific models across a spectrum of levels of intensity ranging from qualitative use of physical reasoning to strong reliance on numerical models.

Berliner specializes in developing modeling and computational methods to apply to problems of assessing climate change and its impacts, weather forecasting, glacial dynamics, and medium-range climate prediction.