Paul Duffy

Paul Duffy is an environmental statistician with over 17 years of experience consulting on applied problems. His greatest skill is the ability to communicate the crux of technical concepts to a non-technical audience. He has worked as a collaborator with scientists from a number of fields, including ecology, geology, mathematics, chemistry, and biology. He has contributed to projects in academia, the private sector, and various levels of local, state, and federal government. One of his recent projects has been assisting with the development of the sampling design for the National Ecological Observatory Network (NEON). This has involved the application of standard classical statistical technique to sample design as well as the development of a hierarchical Bayesian approach to characterize capabilities of both the network as a whole and subsystems. The hierarchical Bayesian approach accommodates non-separable spatial-temporal covariance structures, which can be a critical feature for the statistical characterization of ecological responses across space over the course of several decades. His current research related to the development of sampling designs for network observatories deals with spatial scaling and assimilation of data from different resolutions. This involves simulating data and characterizing the optimality of various network sampling strategies. His dissertation work arose out of the Joint Fire Science Program project, “Development of a Computer Model for Management of Fuels, Human-Fire Interactions, and Wildland Fires in the Boreal Forest of Alaska.” For this project, Paul developed a creative and technically defensible approach for the data collection and statistical analysis of fire-vegetation interactions in Alaska. The project was a success, and the results are published. Paul enjoys interacting with scientists from various fields and thrives on difficult challenges presented by the inter-disciplinary problems encountered in the environmental field. Paul is also proficient at programming in the R statistical computing language.