Cleaner Technology Projects in Denmark 1997
Data Quality and Statistical Analyses in Life-cycle Assessment
Datakvalitet og statistisk analyse i livscyklysvurdering
Arbejdsrapport nr. 49, 1997, Miljøstyrelsen
The purpose of this project is to describe and use methods to characterise data quality and to describe and use statistical analyses in life cycle assessment. Furthermore the project will identify areas with needs for further development and/or testing of methods. The use of the methods has been tested in a number of chosen cases.
This project includes:
 | A systematic collection of information on statistical analyses and life cycle assessment |
| Further development of a method to characterise data quality and test of the method on three selected cases |
| Statistical analysis of data from three cases; in one case Taylor approximation has been used as well as simulation |
The collection of information has been done by searching in available literature, in databases and by contact to people working with handling of high amounts of data and statistical analyses. The collected literature has been described under the headlines:
| Life-cycle assessment and environmental data |
| Uncertainties on industrial and environmental data |
| General statistical methods in relation to life cycle assessment |
| Understanding/interpretation of uncertainties |
Uncertainties in life-cycle assessment can be divided into three types: technical, methodological and epistemological uncertainties. The uncertainties can be introduced in different steps in the life cycle in relation to process data, system data, unit data, characterisation data and valuation data.
Different "roles of thumb" can be used in comparison and assessment of industrial data and hereby reduce the needs for complicated statistical calculations. The methodology requires comprehensive documentation like e.g. statistical assessment of existing data sets.
Principles from economic theory can be adapted in order to analyse data in life cycle assessment by using reliability analysis, validity analysis, dominance analysis and marginal analysis. These methods involve comprehensive and complex calculations and therefore, the need for automatised computation.
The collection of information has only resulted in data from one source: PWMI (European Centre for Plastics in the Environment) that has published data from the European plastic industry. Average data are available for e.g. consumption of raw material, energy consumption and different emissions. The gross energy consumption is given as a range, expressing the spread in the data as a consequence of different operational conditions in different countries.
The statistical methods can be divided in methods based on simulations and methods using exact calculations and approximations.
Simulations are described in relation to risk assessment in connection with public health investigations. The input parameters are described by statistical distributions and a response function can be determined by using relevant input parameters. The uncertainty on the response function (the output variable) can be determined by e.g. "Monte Carlo" simulations.
The variation of sums and differences can be determined by relatively simple formulas especially if the variable is not correlated.
Calculation of variation of products or ratios of two or more variables can be done by e.g. Taylor approximations.
Uncertainties on industrial data are often expressed as: "uncertainty means the variable is in the range with a confidence of 95 %", "the uncertainty is 10 %" or "the uncertainty is 2 MJ". The first statement is supposed to express confidence limits, the second statement is supposed to express a coefficient interval, and the third statement is supposed to express an absolute number, e.g. a spread.
A method to characterise data quality is described theoretically, and the method is tested in three cases.
In life-cycle assessment, the term environmental data expresses data from the considered process, system data expresses data on the flow of raw materials, energy and products, and unit data expresses information on the functional unit. In a complete life-cycle assessment o characterisation and valuation data are also included.
The data quality can be described by a data quality index describing the following parameters:
| Reliability of data |
| Completeness of data |
| Time-related correlation |
| Geographical correlation |
| Technological correlation |
The individual parameters are assigned with a score between 1 (the best) and 5 (the worst).
The described method of characterising data quality might seem overwhelming. In praxis "Description of the case" and "Assessment of the data quality" can be included in the reporting and "Description of the data quality" can be placed in annex or omitted.
Author/ institution
Bo P. Weidema, Instituttet for Produktudvikling
Annette K. Ersbøll, Kongelige Veterinær- og Landbohøjskole
Leif Hoffmann, COWI
This report is subsidised by the National Council for Recycling and Cleaner Technology
ISSN no. 0908-9195
ISBN no. 87-7810-818-7