Abstract

The analysis of count data is extremely important in environmental risk assessment, particularly when developing control limits for pollutants which are harmful to resident biota and which enter the environment from point or non-point sources. Poisson and mixed Poisson models and their extensions provide the backbone for inferential methods which deal with various complexities encountered in the analysis of count data. Here we show how to adopt these methods in the presence of over and under dispersion and lack of independence in real data sets. Examples considered are from the Canadian Effects Monitoring Program, The ELF Ecological Monitoring Program of the U.S Navy, and the 2024 Canadian wildfire in the town of Jasper, Alberta.

Keywords: Toxic contaminants, Impact assessment, Poisson regression; Negative binomial, Lognormal, Taylor’s power law