Statistical properties of emission data in life cycle assessments

The objective of the present paper was to evaluate the statistical properties of emission factors as they are used in life cycle assessment (LCA) studies at present. A group of 26 sulphate pulp mills in Sweden has been used in the study, based on measured emissions to air and water in the period 1986–1993. The study showed large variations in annual emission factors within a group of companies. For most emissions to water, emission factors were independent of annual pulp production rates. Emission factors for an industrial branch can thus be used independent of the size of the plant in LCA studies. When comparing different types of statistical distributions for emission factors, the data set deviated significantly from all the tested distributions. Most emission factors tended to be best represented by a binomial distribution, indicating that emission factors should be square-root transformed before being used in statistical tests. However, not even a square-root transformation resulted in really normally distributed data sets for the emission factors from the pulp mills. The paper discusses several reasons for the large variation in emission factors. Strict regulations from authorities and development of environmental management systems seem to be most efficient in reducing emission factors and annual variations in emission factors