Title: Forest Fires: Learning to Survive

Speaker: Thierry Duchesne (Université Laval)

Abstract: In this talk I will survey the statistical methods that my collaborators and I have used over the last 6 years and are planning to use over the next few years in our analysis and modeling of forest fire data, from survival analysis to machine learning. The first part of the talk will focus on methods used to analyze fire size and duration data and include standard survival analysis techniques, deconvolution methods, joint modeling of longitudinal and survival data as well as causal inference. In the second part I will talk about fire spread simulation models and the methods that we are using and developing to calibrate these models from observed data; these range from standard maximum likelihood estimation to percolation processes to image encoding algorithms. My intent is not to get into the details of each method, but rather to give an idea of the vast array of methodological challenges associated with the understanding and modeling of forest fires.

Date/Time: Tuesday February 2, 2021, 3:45 - 4:45 (NB - This Seminar will start 15 minutes later than usual to accommodate Thierry's teaching schedule)

Location: Virtual