Generalized Linear (Mixed) Models of Post-earthquake Ignitions

Keywords: Statistical analyses.  Regression models.  Post-earthquake ignitions.  Data analyses.  California.  Ground shaking.  Generalized linear mixed models.  Goodness-of-fit analyses.  Census tracts.  Fire hazards.  Geographic information systems (GIS).

Abstract: This report presents a comprehensive approach to statistical modeling of post-earthquake ignitions and to data compilation for such modeling, and applies it to present day California. Specifically, regression models are developed that can be used to estimate the number of ignitions per census tract as a function of tract characteristics and the ground shaking experienced in a specified earthquake. The new approach recognizes the discrete nature of ignition counts by using generalized linear and generalized linear mixed models for the first time in this type of application. It includes careful model selection and goodness-of-fit analyses, examines multiple covariates to estimate ignitions, and uses a census tract as a unit of study to enable better estimates at a finer geographic resolution.