Header

UZH-Logo

Maintenance Infos

Computational frameworks to increase effectiveness and efficiency of data collection in Life Cycle Assessment


Reinhard, Jürgen. Computational frameworks to increase effectiveness and efficiency of data collection in Life Cycle Assessment. 2017, University of Zurich, Faculty of Economics.

Abstract

Life Cycle Assessment (LCA) is perhaps the leading technique for assessing the environmental impact of products over their entire life cycle, and currently plays a decisive role in the field of environmental management. The technique is highly data-intensive and would be practically impossible without using the background unit process datasets provided by Life Cycle Inventory (LCI) databases. A typical product life cycle covers thousands of unit processes, each of which needs to be described with exchange flow data. Although the increased availability of LCI databases has substantially decreased the amount of work involved in conducting an LCA, the high costs associated with data collection still impose substantial limitations on the quality of LCA. These limitations either take the form of uncertain or incomplete background data, which must be gathered and maintained at great effort, or an excessively generic representation of highly context-dependent activities, for example, agricultural work.
The overall goal of this dissertation is to increase the efficiency and effectiveness of data collection in LCA by means of two novel and complementary solutions based on information systems (IS). The first solution is a statistical prioritization approach that directs LCI database improvement efforts toward datasets of key importance in terms of their potential influence on overall database quality. This approach aims to increase the effectiveness of data collection for LCI databases. The second solution is a computational framework that facilitates the automated generation of regionalized cultivation datasets on the basis of publicly available spatial (raster) data. This approach aims to increase the efficiency and quality of unit process dataset generation in the important domain of agricultural LCAs.
We examine the first, statistical prioritization approach in detail in two research articles. Article I (Chapter 2) presents a computational framework for prioritizing LCI database improvements. We demonstrate, then evaluate, a method for database-wide contribution analysis (CA) and corresponding summary measures that facilitate the identification of key processes, that is to say, unit processes with consistently large relative contributions throughout all product systems in the database. We show that prioritizing the improvement efforts is very useful because a tiny, robust nucleus of unit processes proves to be consistently important across all product systems and many Life Cycle Impact Assessment (LCIA) indicators. Focusing research efforts on these processes makes it possible to improve the LCI database effectively.
Article II (Chapter 3) applies the same statistical prioritization framework to the ecoinvent databases in a comprehensive case study. We identify the most important unit processes according to a set of 19 selected LCIA indicators using a newly developed ranking algorithm. Our study shows that a relatively large proportion of the overall database quality is dependent on a small set of key processes. Overall, 300 (out of 11,000) datasets cause 60% of the environmental impacts across all LCIA indicators, while just three datasets cause 11% of all climate change impacts. We present a ranking of key processes that adds a new perspective to database improvements, in that it makes it possible to allocate resources according to the structural dependencies in the data.
We examine the second, regionalization approach in another research article. Article III (Chapter 4) presents a computational framework that allows the automated, site-specific (regionalized) generation and assessment of cradle-to-gate agricultural unit process datasets. The framework facilitates the transformation of publicly available spatial (raster) data into comprehensive unit process datasets using default data from Version 3.2 of the ecoinvent database and the emission models from the World Food Life Cycle Database (WFLDB) guidelines. To illustrate the application of our framework, we describe its key features and present a case study on rapeseed production in Germany. Our study shows that automatically generating regionalized cultivation datasets harbors great potential for improving the accuracy of agricultural unit process datasets in LCA applications. With 580,000 datasets, the case study presented is likely the most comprehensive cradle-to-gate LCA on the climate change and eutrophication impacts of rapeseed cultivation in Germany.
Our research demonstrates that both prioritization and regionalization are valid, useful approaches to improving the data foundation of LCA-based decision-making. The prioritization framework makes it easier to align data collection efforts in LCI databases with those datasets that are the most important in terms of their overall influence on the quality of the database. Focusing research efforts on these processes makes it possible to effectively improve the datasets that play a dominant role in nearly every LCA application. Our research provides valuable new design knowledge in the form of operational principles that can be applied and adapted in other, as yet unstudied fields. Our framework for regionalized LCI modeling makes it easier to automatically generate high-resolution agricultural process datasets for all major crops and all regions around the world. The framework improves the geographical representativeness and reproducibility of agricultural datasets and offers new possibilities for their aggregation and analysis. Finally, the design knowledge that we develop provides a starting point for enhancing the utility of LCA, particularly in the context of other environmental system analysis tools.

Abstract

Life Cycle Assessment (LCA) is perhaps the leading technique for assessing the environmental impact of products over their entire life cycle, and currently plays a decisive role in the field of environmental management. The technique is highly data-intensive and would be practically impossible without using the background unit process datasets provided by Life Cycle Inventory (LCI) databases. A typical product life cycle covers thousands of unit processes, each of which needs to be described with exchange flow data. Although the increased availability of LCI databases has substantially decreased the amount of work involved in conducting an LCA, the high costs associated with data collection still impose substantial limitations on the quality of LCA. These limitations either take the form of uncertain or incomplete background data, which must be gathered and maintained at great effort, or an excessively generic representation of highly context-dependent activities, for example, agricultural work.
The overall goal of this dissertation is to increase the efficiency and effectiveness of data collection in LCA by means of two novel and complementary solutions based on information systems (IS). The first solution is a statistical prioritization approach that directs LCI database improvement efforts toward datasets of key importance in terms of their potential influence on overall database quality. This approach aims to increase the effectiveness of data collection for LCI databases. The second solution is a computational framework that facilitates the automated generation of regionalized cultivation datasets on the basis of publicly available spatial (raster) data. This approach aims to increase the efficiency and quality of unit process dataset generation in the important domain of agricultural LCAs.
We examine the first, statistical prioritization approach in detail in two research articles. Article I (Chapter 2) presents a computational framework for prioritizing LCI database improvements. We demonstrate, then evaluate, a method for database-wide contribution analysis (CA) and corresponding summary measures that facilitate the identification of key processes, that is to say, unit processes with consistently large relative contributions throughout all product systems in the database. We show that prioritizing the improvement efforts is very useful because a tiny, robust nucleus of unit processes proves to be consistently important across all product systems and many Life Cycle Impact Assessment (LCIA) indicators. Focusing research efforts on these processes makes it possible to improve the LCI database effectively.
Article II (Chapter 3) applies the same statistical prioritization framework to the ecoinvent databases in a comprehensive case study. We identify the most important unit processes according to a set of 19 selected LCIA indicators using a newly developed ranking algorithm. Our study shows that a relatively large proportion of the overall database quality is dependent on a small set of key processes. Overall, 300 (out of 11,000) datasets cause 60% of the environmental impacts across all LCIA indicators, while just three datasets cause 11% of all climate change impacts. We present a ranking of key processes that adds a new perspective to database improvements, in that it makes it possible to allocate resources according to the structural dependencies in the data.
We examine the second, regionalization approach in another research article. Article III (Chapter 4) presents a computational framework that allows the automated, site-specific (regionalized) generation and assessment of cradle-to-gate agricultural unit process datasets. The framework facilitates the transformation of publicly available spatial (raster) data into comprehensive unit process datasets using default data from Version 3.2 of the ecoinvent database and the emission models from the World Food Life Cycle Database (WFLDB) guidelines. To illustrate the application of our framework, we describe its key features and present a case study on rapeseed production in Germany. Our study shows that automatically generating regionalized cultivation datasets harbors great potential for improving the accuracy of agricultural unit process datasets in LCA applications. With 580,000 datasets, the case study presented is likely the most comprehensive cradle-to-gate LCA on the climate change and eutrophication impacts of rapeseed cultivation in Germany.
Our research demonstrates that both prioritization and regionalization are valid, useful approaches to improving the data foundation of LCA-based decision-making. The prioritization framework makes it easier to align data collection efforts in LCI databases with those datasets that are the most important in terms of their overall influence on the quality of the database. Focusing research efforts on these processes makes it possible to effectively improve the datasets that play a dominant role in nearly every LCA application. Our research provides valuable new design knowledge in the form of operational principles that can be applied and adapted in other, as yet unstudied fields. Our framework for regionalized LCI modeling makes it easier to automatically generate high-resolution agricultural process datasets for all major crops and all regions around the world. The framework improves the geographical representativeness and reproducibility of agricultural datasets and offers new possibilities for their aggregation and analysis. Finally, the design knowledge that we develop provides a starting point for enhancing the utility of LCA, particularly in the context of other environmental system analysis tools.

Statistics

Downloads

2 downloads since deposited on 14 Feb 2018
2 downloads since 12 months
Detailed statistics

Additional indexing

Item Type:Dissertation
Referees:Hilty Lorenz, Finnveden Göran
Communities & Collections:03 Faculty of Economics > Department of Informatics
Dewey Decimal Classification:000 Computer science, knowledge & systems
Language:English
Date:2017
Deposited On:14 Feb 2018 15:37
Last Modified:19 Mar 2018 10:38
Number of Pages:164
OA Status:Closed
Other Identification Number:merlin-id:16018

Download