A Flexible Class of Spatio-Temporal Model for Mountain Pine Beetle Damage

• Activity Number: 646
• Type: Contributed
• Date/Time: Thursday, August 13, 2015 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistics and the Environment
• Author(s): Kimberly Kaufeld* and Sujit Ghosh
• Companies: SAMSI/North Carolina State University and SAMSI/North Carolina State University
• Keywords: spatio-temporal ; beta regression ; imputation

Abstract

The forests of the western region of the United States have changed dramatically over the last ten years due to the increase in bark beetle damage. Statistical modeling is needed to both understand and predict the occurrence of bark beetle outbreaks. We use data collected from an aerial detection survey in the Rocky Mountains in Colorado in the years 2001-2014 to model damage using a spatio-temporal beta regression model for the percent of the region damaged. A sparse conditional autoregressive model is used to capture any spatial information not modeled by spatially varying covariates. A dynamic linear model is used to capture the temporal evolution of the amount of bark beetle damage occurred in the previous year. We compare the zero augmented spatio-temporal model to an imputed spatio-temporal beta model where areas that were originally recorded as not damaged are noted as detection errors. We find that the imputed model predicts better in a several regions in the Rocky Mountains.