Best Proxy: New Methodology for Selection of LCI Dataset to Achieve Regionalized LCA

The common practice of using generic or country-specific LCI datasets for LCAs for which site-specific LCI datasets are not available can result in inaccuracies and affect the relevance of the LCA results. On the other hand... [ view full abstract ]

The common practice of using generic or country-specific LCI datasets for LCAs for which site-specific LCI datasets are not available can result in inaccuracies and affect the relevance of the LCA results. On the other hand comprehensive site-specific LCI datasets require considerable time and effort, and data that are rarely available at the level of desired detail.

We present a new best proxy methodology for systematic selection of the most appropriate LCI dataset for a specific site, out of the available LCI datasets for a specific background process. The aim of the methodology is to select the dataset that will result in the LCA impact scores that are closest to "true values", at only a small fraction of the effort needed to generate a comprehensive site-specific LCI dataset. When used as a background process for an LCA of a product/service at that site, the selected dataset will evidently lead to better estimations of LCA results than a generic or random country specific LCI dataset. The selection process is based on the concepts of characteristics associated with each dataset for the background process and of "distance" between LCI datasets in the characteristics space, where the missing LCI dataset of the analyzed site is also represented by a set of descriptive characteristics. A rigorous mathematical approach is used to define the "distance" between any two datasets for a specific process in the characteristics space.  The dataset with the minimal distance to the site with the missing dataset is the selected as the best proxy dataset for that site.

The methodology is general and can be applied to various background processes. The methodology is demonstrated and validated on a model of water supply systems that serves as a case study using a harmonized set of 23 published LCA studies and corroborated on a harmonized set of electric power stations fired by coal. The results demonstrate the validity and the predictive power of the methodology. The methodology has an incorporated learning capability. This capability is demonstrated with the case of the Israeli water system LCA. Adding it to the basic set improves accuracy of predicted impact values.

The cost-effectiveness of the methodology is demonstrated by comparing the effort needed to carry out a site-specific LCA of the Israeli water supply system to the effort of implementing the model for one site.

The model developed for water supply systems can be used "as is" in sites for which site-specific LCAs are not available. The methodology can be used to develop similar models for other background processes.  It is expected that models of other background processes will be developed, research will lead to better characteristics sets, and new site-specific LCAs will be published, leading to increased robustness of the predictive power of the methodology and eventual adoption by LCA practitioners.