Measuring and modelling pollution for risk analysis.

Abstract

The great scale and complexity of environmental risk analysis offers major methodological challenges to those engaged in policymaking. In this paper we describe some of those challenges from the perspective gained through our work at the University of British Columbia (UBC). We describe some of our experiences with respect to the difficult problems of formulating environmental standards and developing abatement strategies. A failed but instructive attempt to find support for experiments on a promising method of reducing acid rain will be described. Then we describe an approach to scenario analysis under hypothetical new standards. Even with measurements of ambient environmental conditions in hand the problem of inferring actual human exposures remains. For example, in very hot weather people will tend to stay inside and population levels of exposure to e.g. ozone could be well below those predicted by the ambient measurements. Setting air quality criteria should ideally recognize the discrepancies likely to arise. Computer models that incorporate spatial random pollution fields and predict actual exposures from ambient levels will be described. From there we turn to the statistical issues of measurement and modelling and some of the contributions in these areas by the UBC group and its partners elsewhere. In particular we discuss the problem of measurement error when non-linear regression models are used. We sketch our approach to imputing unmeasured predictors needed in such models, deferring details to references cited below. We describe in general terms how those imputed measurements and their errors can be accommodated within the framework of health impact analysis.