University of Washington - Department of Statistics
The field of environmental statistics has shown a steady growth of activity over the last decade or so. There has been work in a variety of areas, but in this talk I will focus on those where novel models, approaches, or ways of thinking are involved. And I will concentrate on air pollution to keep the amount of material finite.
Much of my own work has dealt with spatial statistics and modelling. In the future we will have to deal with space-time models, heterogeneous, nonstationary, nonseparable and multivariate.. We need models for global processes in space and time, and more generally models of vertical distributions of pollutants globally in space and time.
Recently much effort has gone into comparison of deterministic models to data. To go further, the model building needs to be an interaction between model output, data, stochastic components, and accurate assessment of the different uncertainties involved in all of these.
Health effect models have traditionally been dealing with acute health effects, and have been based on opportunistic studies. We need better estimates of personal exposure, and better toxicodynamic models to understand the mechanisms of health consequences of exposure.
Perhaps the biggest uncertainty in air quality modeling is associated with emissions. To understand emissions better, we must develop multivariate spatial dynamic source apportionment tools.
Finally, I will touch on the role of statisticians in environmental decision-making: where we are, and where we need to be.