Life cycle impact assessment research developments and needs
- First Online:
- Cite this article as:
- Bare, J.C. Clean Techn Environ Policy (2010) 12: 341. doi:10.1007/s10098-009-0265-9
- 633 Views
Life cycle impact assessment (LCIA) developments are explained along with key publications which record discussions which comprised ISO 14042 and SETAC document development, UNEP SETAC Life Cycle Initiative research, and research from public and private research institutions. It is recognized that the short list of impact categories has remained fairly constant, even after extensive discussions. The termination point of impact assessment modeling (e.g., inventory, midpoint, endpoint, damage, single score) has been discussed extensively, and the advantages and disadvantages of these different levels are well published. Early LCIAs were conducted independent of system location, but now site-specificity has been a research topic for many of the local and regional categories (e.g., acidification, eutrophication, and smog formation). In reality, even though many advances have been made in site-specific analysis, the life cycle assessment (LCA) case studies are often limited to their inventory data, and as a result, most LCAs are still site-generic even though the LCIA methodologies exist to allow for site-specific analysis. Pollutant-based impacts have received the most research effort and support in the past, but resource depletion categories (e.g., land use and water use) are now recognized as being highly complex, site-specific, data intensive, and important for contributing toward the sustainability of the planet. Efforts in these categories are still in the neophyte stages and are expected to have the greatest advances in the upcoming years.
KeywordsLife cycle assessment (LCA)Life cycle impact assessment (LCIA)Sustainability metricsEnvironmental standardsImpact assessment
Key deliberations and publications
Life cycle impact assessment (LCIA) quantifies the potential for environmental impacts over all stages involved in the delivery of a product or service, including resource extraction, feedstock manufacturing, product manufacturing, transportation, use, recycling, and/or disposal phases.
One of the earliest significant publications in LCIA provided characterization factors for the following impact categories: depletion of abiotic resources, depletion of biotic resources, enhancement of the greenhouse effect, depletion of the ozone layer, human toxicity, ecotoxicity, photochemical oxidant formation, acidification, nitrification, and odor (Heijungs et al. 1992a, b). Although some of the impact categories had been developed specifically for Europe, by the mid-1990s, most life cycle assessment (LCA) practitioners worldwide were either using CML 1992, or not conducting an LCIA at all, but simply presenting results as a list of the life cycle inventory (LCI).
Over the next few years, impact assessment methodology developed. Two independent SETAC working groups gradually developed documents in the mid-1990s and later began to merge in the late 1990s, when both groups were working together on crafting the ISO 14042 standard. Initially, guidance came from SETAC-North America, which provided the earliest documents on A Technical Framework for Life-Cycle Assessment (Fava et al. 1991, 1993) and A Conceptual Framework for Life-Cycle Assessment (Fava et al. 1993). SETAC-Europe’s early contribution was titled Towards a Methodology for Life-Cycle Impact Assessment (Udo de Haes 1996). A 1997 SETAC-North American publication was titled Life-Cycle Impact Assessment: The State-of-the-Art (Barnthouse et al. 1997), and a publication which included contributions from both North Americans and Europeans LCIA: Striving Towards Best Practice (Udo de Haes et al. 2002) was published in 2002. Taken together, these documents provide a good snapshot in time of developments within this area. The most recent SETAC document (Udo de Haes et al. 2002) has the advantage of occurring after the discussions which were integral to the ISO 14040 series development and a series of international workshops on the varying levels of LCIA sophistication (Bare et al. 1999, 2000b), LCIA midpoint and endpoint analysis (Bare 2000; Bare et al. 2000a), and the taxonomy of LCIA (Bare 2003; Gloria and Bare 2003).
ISO 14042 development in the late 1990s was a fruitful time for LCIA discussions and consensus building globally (International Standards Organization 2000). Important issues discussed at these meetings and several ensuing discussions prompted a number of journal articles and letters to the editors. One of the most controversial issues surrounded the area of comparative assertions [i.e., an “environmental claim regarding the superiority or equivalence of one product versus a competing product which performs the same function” (International Standards Organization 1997, 2000)]. Other controversies focused on the integration of science-based and normative factors when supporting environmental decision making (Hertwich 1999; Hertwich and Pease 1998; Marsmann et al. 1999). The details of these debates will be presented later in this article.
Depletion of abiotic resources
Depletion of biotic resources
Enhancement of the greenhouse effect
Depletion of the ozone layer
Photochemical oxidant formation
Damage to ecosystems and landscapes
Although there have been numerous discussions on the best set of impact categories, the categories presented by Heijungs and Guinee have remained fairly constant over the years. The taxonomy of a comprehensive impact assessment was discussed at the 1998 LCIA Sophistication Workshop, during the ISO 14042 development, and prior to publication of the SETAC 2002 document (Udo de Haes et al. 2002). An email exchange then followed over the course of several months (Bare et al. 2002) which captured the controversy and questioned the basis for the impact categories which were originally proposed by Heijungs and Guinee and had become “de facto” starting points for every LCA. This entire email discussion has been preserved over the course of 43 electronic pages on the Global LCA Village website to allow outsiders to understand the high level controversy which previously only those involved had access (Bare et al. 2002). The discussion began with questions about goal and scope and how impact categories should be chosen—especially if some practitioners were more eco-centric or human-centric. Inconsistencies which were discussed included the proposed inclusion of wildlife which are killed in transportation accidents, but the exclusion of humans killed during these same events. Life support functions were also discussed. There was a proposal by some for midpoint categories such as ozone depletion and global warming to be considered a new area of protection or life support system based on their value in support of human and other life on earth. Others believed that maintaining the results as midpoint categories was sufficient and should allow significant focus during the valuation process, if included (Bare et al. 2002). Taxonomy was a common theme, and several within this discussion asked for a further exploration of a more comprehensiveness list of potential impacts (Bare 2003). Although consensus was not reached during this email exchange, issues were discussed, and the SETAC document was completed (Udo de Haes et al. 2002). In 2002, an international workshop on the taxonomy of impacts was held in Hamburg, Germany (Bare 2003; Gloria and Bare 2003), and Bare presented a taxonomy which was accepted as the most comprehensive listing of impacts to date including human, environmental, and resource use in a hierarchical structure (Bare and Gloria 2008).
resource and land use;
stratospheric ozone depletion;
To date, more recent UNEP SETAC LC Initiative documents have continued to maintain this list.
Although the impact categories have remained relatively constant since their introduction despite all of the controversy, the characterization models underlying these and many of the other categories have undergone significant development and improvement.
The earliest impact categories which achieved relative scientific consensus in characterization were those categories which are site independent such as ozone depletion (WMO—World Meteorological Organization 1999, 2003, 2007) and global warming (Houghton et al. 1995; IPCC—Intergovernmental Panel on Climate Change 1996, 2001a, b, 2005; Solomon et al. 2007; UNFCCC—The United Nations Framework Convention on Climate Change 2000). These impact categories were the simplest to implement since the location of the chemical emission was relatively inconsequential. Very little argument existed when evaluating these categories at the midpoint level (Bare 2000; Bare and Gloria 2008; Bare et al. 2000a; Hertwich and Hammitt 2001a, b; Jolliet et al. 2004) with the exception of the timeframe which should be used. Since endpoint and damage analysis includes more value choices, modeling assumptions, and forecasts not based on scientific consensus, quantification at the endpoint and damage level continues to focus on the loss of comprehensiveness and the uncertainty involved in quantifying the impacts (Bare et al. 2000a).
In order to simulate the potential impacts in a manner closer to what is expected to occur; some impact categories which are influenced by chemicals of much shorter lifetimes and travel distances such as acidification, eutrophication, and smog formation have a much stronger spatial and temporal dependence than the categories of ozone depletion and global warming. Models have been developed regionally to allow a better agreement between what is predicted in LCIA and what is expected when conducting a temporally and spatially specific analysis. Factors such as fate, transport, background, and buffering capacity have been included within LCIA acidification models. Eutrophication models have been modified to include fate and transport analysis even to the level of utilizing stream analysis data to determine the ultimate fate, and thus limiting nutrient in that environment. Smog formation modeling can include empirical data from various cities to tailor the model with existing background concentrations, biogenic sources, weather patterns, and seasonal variations (Andersson-Skold et al. 1992; Bare et al. 2003; Bellekom et al. 2006; Derwent et al. 1996, 1998; Hauschild et al. 2006; Hettelingh et al. 2005; Huijbregts and Seppala 2001; Jenkin and Hayman 1999; Labouze et al. 2004; Margni et al. 2008; Norris 2003; Potting and Hauschild 2006; Potting et al. 1998; Seppala et al. 2004, 2006; Van Zelm et al. 2007).
Controversies which continue to exist in acidification, for example, include whether and/or how to include the background concentrations and buffering capacity within the model. While some advocate that “less is better” everywhere, others promote prioritizing locations which are most sensitive to change. Whichever perspective is utilized for a specific study, the perspective and the ramifications of these decisions should be identified and made transparent. Finally, it should be noted that although numerous research communities have published guidance allowing the inclusion of more temporally and spatially specific parameters and models which allow a more definitive quantification of potential impacts, the majority of the LCA case studies are still site-generic simply because it is more difficult to maintain the site-specific character of the LCI data throughout the LCIA process. Also, as discussed above, with many of the impact categories, attempts at quantifying these impact categories at the endpoint or damage level remain controversial because of the lost comprehensiveness, the additional uncertainty, and the value choices and assumptions which become embedded in the models (Bare et al. 2000a).
Although presented, there was no specific mention of how to characterize land and/or water resources (Heijungs et al. 1992a, b). The above equation (Eq. 1) was utilized to characterize and quantify the depletion of energy sources, such as fossil fuels and metals. Since its introduction, there have been several attempts to characterize and quantify land and water resources based on various factors for land use including biodiversity, scarcity, and ecosystem services, but no consensus within the LCIA community has been reached (Anton et al. 2007; Baitz et al. 2000; Goedkoop and Spriensma 1999; Heijungs et al. 1997; Koellner 2002; Koellner and Scholz 2007, 2008; Lindeijer and Alfers 2001; Lindeijer et al. 1998; Michelsen 2008; Mila i Canals et al. 2006, 2007a, b; Schmidt 2008; Spitzley and Tolle 2008; Swan 1998; Udo de Haes 2006; Vogtlander et al. 2004; Wagendorp et al. 2006; Wurtenberger et al. 2006). Even less attention has been given to the category of water use to date (Arpke and Hutzler 2006; Heuvelmans et al. 2005; Owens 2001). Research is on-going within the UNEP SETAC Life Cycle Initiative to make advances in both these categories (Mila i Canals et al. 2007a).
Much attention has been given to the categories of human health and ecotoxicity. While the original work was simply based on toxicity, more recent models have continued to build upon the work of the research community and have utilized fate and transport models, risk assessment guidelines, and human exposure factors available from the risk assessment arena (Bare et al. 2003; US Environmental Protection Agency 1997, 1989a, b). These categories have also remained controversial because countries had different protocols, policies, and regulations to target various chemicals of concern (Brand et al. 1998; Goedkoop 1995; Goedkoop et al. 1996; Goedkoop and Spriensma 1999; Hauschild and Wenzel 1998; Hauschild et al. 2006; Itsubo and Inaba 2003; Jolliet et al. 2003; Steen 1999a, b; Toffoletto et al. 2007; Wenzel and Hauschild 1997). Data variability and uncertainty analysis indicate that the input parameters which contributed the most to data uncertainty and variability were related to the uncertainty of the individual chemical toxicities and lifetimes (Hertwich et al. 1999), while, the differences in human lifestyle patterns, weather, and landscape conditions were not as influential to the relative characterization of the chemicals. Consistent with these findings, the UNEP SETAC Life Cycle Initiative began an effort to develop a consensus model for human health and ecotoxicity in 2003 (Hauschild et al. 2008). Although research is on-going, interim versions of the USEtox model, a result of these collaborations, are available at www.usetox.org (Hauschild et al. 2008; Rosenbaum et al. 2008).
As previously mentioned, there are some impact categories which are currently utilized specific to an application. For example, because buildings are responsible for over 60% of the US electricity consumption (Energy Information Administration—U.S. Department of Energy 2003) and as a result, are recognized as being a significant contributor for all environmental impacts, LCIA has been employed to address and quantify the environmental impacts associated within the building industry (Bare and Gloria 2005). Recently, the US Green Building Council has employed the use of the Tool for the Reduction and Assessment of Chemical and other environmental Impacts (TRACI) (Bare et al. 2003) to characterize environmental impacts on a life cycle basis for Leadership in Energy and Environmental Design (LEED) weighting credits (US Green Building Council 2008a, b, c). Indoor air pollution has also been an issue of concern for many building occupants, and so weighted credits for indoor air pollution are incorporated into LEED even though it was not a traditional TRACI impact category.
Normalization “relates the magnitude of impacts in different impact categories to reference values” (Finnveden et al. 2002), or “relates the micro world of an LCA study to the macro world in which the product/service is embedded” (Lindeijer 1996). Normalization should be conducted with the grouping or weighting steps in mind so consistency is achieved in perspective. Early efforts in this area relied on internal normalizations (i.e., a comparison to one of the analyzed options), but practitioners were cautioned about the need for congruence with the valuation process when conducting internal normalization (Norris 2001).
The above calculations may also be quantified on a per capita basis, or quantification may cease at the spatial and temporal perspective.
In the past, the primary difficulties with external normalization have been the lack of data for the selected spatial and temporal scale, the inconsistency in a data set with characterization methodology and the stressors of concern (e.g., the chemical emissions available do not cover all of the chemicals or sources which should be evaluated), and the contribution of uncertainty related to the normalization database (Bare et al. 2006). Normalization databases are available for various countries and continents around the world (Bare et al. 2006; Blonk et al. 1997; Breedveld et al. 1999; Huijbregts et al. 2003a; Lundie et al. 2007; Seppala 2007; Seppala and Hamalainen 2001a, b; Sleeswijk et al. 2008; Strauss et al. 2006; van Oers and Huppes 2001), but high quality data sets are primarily compiled from European and North American sources. A high quality global normalization database is not expected to be accurate for most impact categories at this time since there is very little consistency of reporting, ease-of-access, or comprehensiveness in terms of stressor coverage.
Grouping and weighting
As mentioned in the earlier ISO 14042 discussions, the role of value choices, modeling assumptions, and forecasts has been and remains one of the most divisive issues of the LCIA community. While it can and has been argued that value choices and modeling assumptions are an inherent part of LCIA modeling and cannot be avoided (Hertwich 1999; Hertwich and Pease 1998), it has also been pointed out that there are ways to minimize value choices and forecasting through the use of midpoints instead of endpoints and damage level modeling (Bare 2000; Bare and Gloria 2008; Bare et al. 2000a, 2003). Value choices are unavoidable when the study is being reduced to a minimal set of environmental impact indicators such as a single score or one score to represent human health, one score to represent the environmental health and one score to represent resource use. However, some methodology developers have attempted to categorize human beings into various perspectives and allow LCIA users to select one of these perspectives (Goedkoop and Spriensma 1999; Hofstetter 1998). In these cases, it is difficult, but important, to understand the proposed perspectives, to discern which perspective is most closely aligned with the commissioners and stakeholders of the LCA case study.
The most highly contested issue within the development of ISO 14042 was whether weighting should be allowed for use in comparative assertions. Several international meetings individually focused on this issue until the June 1998 ISO Development meeting in San Francisco, California. This issue was key since there was (and continue to be) a sharp divide between what is acceptable in inclusion of value choices within environmental modeling (Hertwich 1999; Hertwich and Pease 1998; Hertwich and Hammitt 2001a, b; Marsmann et al. 1999). Researchers on one side argue for scientific defensibility within the modeling and voted for not allowing disparate impact categories and for potential effects to be summed up to a simplistic environmental indicator. Researchers on the other side wanted to acknowledge that LCA belonged within a decision making framework and state that it would be easier to make a decision if weighting were allowed to combine impact categories and potential effects. The US representative (and author of this paper) and the Dutch representative were tasked with finding a solution since their countries represented the opposite ends of the spectrum. Their proposal which allowed grouping but not weighting within comparative assertions was accepted immediately when taken to the international ISO community (Udo de Haes 1998).
Results of a valuation process conducted for the US National Institute of Standards and Technology
Fossil fuel depletion
Criteria air pollutants
Indoor air quality
Many of the discussions and concerns presented by the panelists have been reported and preserved in the paper documenting the process (Gloria et al. 2007). The weights which were developed via the process are not expected to reflect the perspective of the companies, consultancies, or government bodies which were represented, but simply the panelists’ view of the importance of each impact category. It was also recognized that all the panelists maintain biases and perspectives and have varying degrees to which they can be influenced during the workshop.
Data quality within an LCA includes not only the data’s uncertainty and parameter variability of the LCI and LCIA phases, but also model uncertainty of the LCIA. An early SETAC document (Fava et al. 1994) focused on data quality, however this topic has not been revisited by an international group since 1994. Instead, individual research groups have addressed the topic as it applies towards a specific impact category (Hertwich et al. 1999, 2000; Hertwich and Hammitt 2001a; Huijbregts 1999; Huijbregts et al. 2000; Potting and Hauschild 2006), or in total perspective (Ciroth et al. 2004; Geisler et al. 2005; Hertwich and Hammitt 2001a, b; Hofstetter 1998; Huijbregts 2001; Huijbregts et al. 2003b; Lenzen 2006; Lo et al. 2005; Ross et al. 2002). One of the most advanced efforts in this area is presented by Hertwich and Hammitt (Hertwich and Hammitt 2001a, b). In this two-part article, they recognize that many of the current impact assessment methods include discussions of data uncertainty, but none presents a full picture that allows for a complete quantification of all uncertainty and variability to fully analyze comparable case studies. They support a midpoint modeling perspective, since more uncertainty is introduced in the endpoint and damage modeling stages, but allow that endpoint models may be useful “to support ‘judgments about facts’ needed to evaluate the importance of different impact categories (or means objectives) in the means-ends objectives network”(Hertwich and Hammitt 2001a, b).
The major environmental impacts which are typically included within an LCIA have remained fairly constant over the years with a few additions for special applications (e.g., indoor air for green buildings). Significant advances have been made in impact assessment modeling for nearly every impact category, but it is recognized that some categories such as land use and water use are still in the early stages of development.
Various impact assessment models exist which are dependent upon temporal or spatial perspectives, or the perspective or view of the world. This variability can be very useful when tailoring a study to a particular location and set of regulations (e.g., using TRACI for US case studies), but the variety can also be perplexing when the analysis involves many global suppliers and governments. While for many of the categories it does not seem to matter which methods are used (ozone depletion, global warming), for other categories the location and perspective is extremely important to the point where an LCIA cannot be properly conducted without this knowledge (e.g., land use).
Midpoint impact assessment continues to support more scientifically based decision analysis. Endpoint and damage analysis provides additional support when a smaller or single environmental indicator is desired. Many of the recent proposals attempt to develop both midpoint and endpoint analyses which are consistent in framework.
Data quality continues to remain an issue because of the complexity of the problem and the lack of uncertainty knowledge to support uncertainty analysis. While many studies attempt to provide some guidance about uncertainty on specific data, none provides a complete picture of the issue.
The UNEP SETAC Life Cycle Initiative is currently the most significant effort to achieve advances in the area of LCIA. The attempt to achieve consensus on some impact categories ended up supporting the various research efforts which were more spatially and temporally accurate. Other attempts, such as developing a consensus model on human toxicity and ecotoxicity, appear to be making advances in this area. Still, other research in the areas of land and water use is too early to judge.