Data quality in Life Cycle Assessment of waste management: Importance for result interpretation and decision-making

Abstract

Despite the data quality requirements in life cycle assessment (LCA) standards and existence of data quality assessment (DQA) frameworks there is still need for a further integration of data quality in LCA. Main challenges are the choice of appropriate life cycle inventory (LCI) data, subjectivity in the DQA process, and the lack of a standardized approach to include data quality alongside with the quantitative uncertainty when interpreting the LCA results. The scope of this PhD project was on the role of data quality in LCA of solid waste management (SWM) systems. The objectives, methods, results, and conclusions of the PhD project are summarized. The objectives were to make methodological contributions to 1) representative foreground modeling concerning data choices versus specificity of scope of study and concerning the occurrence of data gaps, 2) DQA methods concerning the assessment of data representativeness relative to technology characteristics, and the systematic assessment of process completeness, and 3) integration of uncertainty and data quality analysis to identify the critical data in a model. The same approach was followed for the four publications: development of a principle method and test on a case study. The case studies involved the construction of SWM LCA models. Furthermore, waste-to-energy (WtE) technology data was collected and applied as an illustrative example of modification of the criteria to data representativeness. In the LCA modeling benefits from the generation of usable by-products were accounted for by system expansion. The LCA models were constructed in EASETECH and the data analysis was done in Excel and RStudio. Concerning objective 1, firstly, data choices relative to the specificity of the scope of study was analyzed for a landfill model. The number of representative landfill datasets ranged from 52 to 1 for the least specific and most specific scope of study. This was reflected by the global warming potential (GWP) results, which ranged from -457 to 293 kg CO2-eq for the least specific scope of study and a single point value for the most specific scope of study. The results highlighted the need for compatibility between the scope of study and LCI model. Secondly, the use of surrogate data to fill data gaps was analyzed for a model of household waste management. Alternative proxy values were identified for data gaps in the waste treatment processes, and the critical data gaps were those influencing the results > 5% when not filled by a proxy value; these were the CH4 release from composting (up to 40%), WtE efficiency (>100%), sorting efficiencies at the material recovery facility (up to 29%), and composition of the plastic, metal, and paper fractions in the waste (up to 25%). The selection of proxies for the critical data gaps were based on a comparison of the data quality of all candidate proxy values. Concerning objective 2, firstly, a general format of modified criteria to assess data representativeness in a pedigree matrix was developed. The modified criteria consisted of a time interval according to the temporal development of the technology, and identification of relevant geographical and technical factors, which should be modeled accurately. An example with the flue gas emissions of the WtE technology resulted in 4 year time intervals and the worst pedigree score given for data older than 15 years. The relevant geographical and technical factors were the design of the scrubber, NOx removal, and dioxins removal, and the recipient of the flue gas emissions. Secondly, a systematic identification of the expected flows for calculation of the process completeness score was tested on the WtE process. A completeness score of 78% was obtained for the material consumption, and the missing flows were auxiliary fuels and precipitation chemicals. The completeness score for the air emissions ranged from 38 to 50% depending on the inclusion of expert judgment. Consideration of the relevance of the air emissions to GWP resulted in an adjusted score of 67%. Furthermore, applying weighting factors reflecting the greenhouse gas contribution within the geographical context of the LCA adjusted the completeness score to 94%. Concerning objective 3, an integrated importance and data quality analysis was suggested. Firstly, a comprehensive data quality evaluation was used to identify issues influencing the overall LCA results. Secondly, if not provided with the data, the estimation of the input uncertainty was correlated with the evaluation of the acquisition method and statistical properties of the data. Thirdly, the strength of the data was calculated as the average data quality score and plotted with the uncertainty contribution of the data to identify critical data in an LCI model. A preliminary example with an LCA model of paper waste management was provided, which will be further elaborated. Based hereon, the principle conclusions of the PhD project are: A. To avoid bias of LCA results, due to data inaccuracies, the range in representative discrete data choices should be included in the LCI model. B. Data quality and quantitative uncertainty are independent data attributes that are equally important in the interpretation of the LCA results. C. Systematic consideration of the technology characteristics in the assessment of representativeness and process completeness is a prerequisite of relevant and credible data quality scores and LCA results. D. An improved DQA framework is suggested that include the aspects of B and C.