A formal procedure for data quality management in life cycle inventory is described. The procedure is applied to the example of an energy inventory for 1 kg rye bread. Five independent data quality indicators are suggested as necessary and sufficient to describe those aspects of data quality which influence the reliability of the result. Listing these data quality indicators for all data gives an improved understanding of the typical data quality problems of a particular study. This may subsequently be used for improving the data collection strategy during a life cycle study. To give an assessment of the reliability of the overall result of a life cycle inventory, the data quality indicators are transformed into estimates of the additional uncertainty due to the insufficient data quality. It is shown how a low data quality can both increase the uncertainty and change the mean value. After assigning additional uncertainties to all data in the study, a calculation of the uncertainty of the overall result is made by the use of simulations. The use of default estimates of additional uncertainties is suggested as a way to both simplify and improve the procedure.
Weidema and Wesnæs demonstrate the procedure with an energy inventory for 1 kg of rye bread. They propose five independent indicators to describe data-quality characteristics that affect the reliability of an inventory result. Considering these indicators across the data can help a study team identify recurring weaknesses and use them to improve its data-collection strategy.
The authors describe translating insufficient data quality into estimates of additional uncertainty for individual inventory data. These estimates can then be combined through simulation to assess uncertainty in the overall life cycle inventory result. The paper also notes that poor data quality may affect both the uncertainty and the mean result. Default estimates of additional uncertainty are proposed as a way to simplify and improve the procedure. The approach links data-quality review to quantitative uncertainty analysis, while its conclusions remain dependent on the indicators and estimates selected for the study.
