Vic Barnett

Setting Environmental Standards: A Statistical Approach

The EU, governments and agencies throughout the world seek to regulate and protect the state of the environment by setting environmental standards to control the level of certain pollutants present in various media of concern. It has become evident in recent years that many such standards are set without due consideration of uncertainty and variation and many are based on poorly defined principles. Barnett and O’Hagan (1997) conducted a study of current standards, concluding that most standards can be classified into one of two unsatisfactory categories; ideal and realisable standards. They proposed a new conjoint and more statistically meaningful concept; the statistically verifiable ideal standard(SVIS). We consider the implications of this approach with the aim of applying the results to standards formulation for practical problems of pollution in air, water and soil, working in co-operation with relevant bodies. We will discuss the case of a simple pollutant distribution, developing SVISs from various standpoints within a hypothesis-testing framework. We consider a best linear unbiased quantile estimator(BLUQE) for use in the compliance assessment of a SVIS and describe an approximate significance testing procedure using simulated critical values. Finally, we develop a SVIS for a practical example involving water quality and illustrate the use of the BLUQE.