Toxic Data
Bias and the Challenges
of Using LCA in the Design Community
Presented at GreenBuild 2003 –
Published in the Proceedings of GreenBuild 2003 Conference
And in
“Sustainable Development and Society”,
US General Services Administration,
Tom Lent
Principal • Energy & Environmental
Consulting •
510-845-5600 •
1. Overview
Awareness is growing in the design community of the need to account for impacts throughout the life cycle when assessing the environmental characteristics of materials. Until recently, Life Cycle Assessment (LCA) tools to help rationalize this intricate task have been inaccessible to most due to the need for complex software tools and expensive proprietary databases. Now with the development of software tools such as BEES 3.0, relatively simple affordable tools that compare LCA information for different building materials are becoming available to the design community and the USGBC is considering incorporating LCA into LEEDä.
This paper outlines some important inherent structural constraints on the ability of LCA to address a range of toxic chemicals and their related human health issues. It particularly focuses on toxicity hazards that are known by science to be serious environmental health problems but are as yet poorly quantified or otherwise not readily managed in an LCA framework. It explores how these LCA constraints can guide the user away from a good understanding of the full environmental health impacts and can lead to materials decisions that do not actually reflect the user’s environmental goals. It also suggests approaches to overcome these problems to consider before LCA tools are incorporated into LEED or otherwise used broadly by the design community. The goal is to ensure that these tools serve the design community reliably and assure that their use does not undermine the environmental and health goals they seek to promote.
2. LCA: Power of the double edged sword
Quantitative LCA tools have progressed tremendously and have become an effective means for systematic internal industrial design analysis. When carried out by an individual manufacturer using datasets they understand and manage, these analyses can provide excellent insight into the impacts of alternative design pathways and be powerful tools for identifying environmental impacts and selecting optimal design directions.
The application of these tools to material selection by the design community, however, presents significant challenges that have not yet been overcome. The power of the LCA tool lies in the wide ranging scope of its analysis, encompassing a large number of factors through which a building material can impact the environment throughout its life cycle. At this state of development, however, this scope can be a double-edged sword. Descriptions of LCA typically imply that the analysis is complete, describing LCA as the analysis of the total environmental impact of a product through every step of its life.[*] While LCA designers are striving mightily to improve the accuracy of their estimates and approximations, LCAs can never live up to the expectation of total analysis set by such descriptions. By definition, LCAs must always have boundaries limiting the impacts they attempt to model and are highly dependent upon industry and science to provide useful data to drive the models. If the users don’t understand these limitations, LCA tools can give the user a false sense of security that the tool is providing a comprehensive, unbiased and final analysis of all of the environmental impacts resulting from production, use and disposal of a material, ending the need to ask further critical questions. In reality, serious data and analysis limitations inherent in current tools can lead them to strong but hidden biases for materials with major environmental health impacts that are as yet inadequately quantified or where acceptable health and safety threshold limits are in flux and dispute. Persistent, bioaccumulative toxic (PBT) chemical releases are one key area that is highly problematic to accurately represent in a quantitative assessment.
3. Reversals through uncertainty: Vinyl vs. LINOLEUM CASE study
A case study will demonstrate how LCA results can be dramatically affected by the effects of data uncertainty. In fact results can be totally reversed.
The Vinyl Institute and several vinyl flooring manufacturers have been quick to promote that one popular LCA tool – the BEES model (NIST 2002) - appears to rate vinyl composition tile (VCT) as much more environmentally sound than linoleum. The Institute further implies that the USGBC has endorsed this evaluation:
“The results show VCT ranks 20 to 30 percent higher in environmental performance and 90 to 170 percent higher in economic performance. Criteria for the rating include indoor air quality, solid waste, acid rain, global warming and natural resource depletion. The BEES model for evaluating building products has been adopted as an official tool of the U.S. Greenbuilding Council (sic), and is used by architects, builders, contractors and other specifiers who want to select environmentally friendly products.” (Vinyl Institute 1998)
Indeed
the comparison looks even more dramatic than the Vinyl Institute suggests. With
no modification of parameters, BEES gives normalized environmental performance
score results of .0521 for generic linoleum and .0131 for generic VCT (where a
lower score represents a lower environmental impact). From this analysis it appears
that linoleum has 3.98 times the environmental impact of VCT. BEES allows for
the fact that all environmental impacts are not of equal import. BEES provides
the option of using one of two different weighting schemes developed by a US
EPA Scientific Advisory Board or
A
deeper look at the numbers within the BEES model and study of the controversy
around vinyl, however, reveal a different story. A careful comparison of each
of the impact categories across the two floorings reveals that BEES actually
calculates that VCT performs worse than linoleum in every category in which
either of the materials has a significant impact, with impact factors that are
anywhere from 1.4 to 8 times higher in every category, except eutrophication –
excess nutrient runoff[†].
Linoleum even outperforms VCT on fossil fuel depletion despite the fact that
linoleum currently is imported from
A look at one of the health impact concerns at issue in the life cycle of vinyl will help make clear the challenges faced by LCAs in properly evaluating the health impacts of materials. Dioxins - the most potent carcinogens known to science - are an unavoidable byproduct of the manufacture of polyvinyl chloride (PVC) feedstock for VCT and of the combustion of PVC products (Thornton 2002).
For an LCA to accurately capture and evaluate the health impacts of dioxin releases from the life cycle of VCT, LCA planners must have access to validated quantifications of the amount of dioxin emissions resulting from an average pound of PVC though its life cycle. Science is still very early in its efforts to quantify dioxin flows in the environment. The vinyl industry points to the US EPA Dioxin Assessment and notes that the latest draft identifies only 12.3g TEQ[‡] as coming from ethylene dichloride/vinyl chloride manufacturing. They point out that this is a tiny fraction of the total EPA estimated dioxin flows of 3,252g TEQ (USEPA 2001a). Unfortunately, this is likely just a tiny fraction of the actual flows of dioxin from the PVC life cycle. The EPA Dioxin Assessment inventory represents only the flows that have been reliably characterized and quantified, not an estimate of total flows. A tremendous volume of projected dioxin flows are known to exist but have not yet been quantified reliably enough to be included in the official EPA assessment. For example, the EPA estimates that landfill fires alone (with PVC the primary chlorine donor and therefore a major dioxin factor) may contribute 1000g TEQ of dioxin per year (USEPA 2001a). This is only one of a significant number of yet to be quantified dioxin flows from the life cycle of PVC, including as yet poorly or unquantified combustion sources in which PVC is a major chlorine donor (such as structural and vehicle fires) and disposal sites for dioxins produced in the manufacture of PVC (such as landfills and injection wells) and better quantified sources for which the fraction contributed by PVC has not yet been estimated (such as burn barrels at 628 g TEQ , copper smelters and incinerators).
The result of the uncertainty in estimating dioxin flows is that the actual dioxin flows resulting from PVC production and use are indeterminate. They likely total somewhere between one and two orders of magnitude above the quantities listed in the inventory of the EPA assessment. With a potent flow like dioxin this is potentially a very significant issue. The human health factor rating in BEES for dioxin is more than 10,000 times higher than the next highest chemical (diethanol amine) and a million or more times greater than the remainder (Lippiatt 2002). A small difference in estimates of dioxin flows can have a massive impact on the outcomes. Increasing the dioxin flow factor in BEES for VCT only by a factor of three would be sufficient to totally eliminate the environmental advantage that BEES indicates for VCT over linoleum in the initial comparison.
If only
a portion of the EPA estimates eventually are confirmed, the end result for a
VCT material comparison will be dramatically changed – if the flow is included
in the LCA. Researching the origins of the dioxin flows in BEES, brought me to
the PriceWaterhouse Coopers DEAM database that drives the BEES analysis. On
inquiry at PriceWaterhouse, I was surprised to learn that that PVC flow model
comes via
It should be noted that there are similar data concerns (except in the opposite direction) on the linoleum side about whether the massive eutrophication impact that BEES displays is correct. Resolution of these possible discrepancies could also reverse the results of the BEES analysis in linoleum’s favor[**].
LCAs as currently designed can only calculate and compare impacts based upon flows that are well understood and quantified. For at least some prominent materials, only a small portion of the total chemical flows that result in known human health impacts are well quantified. As is made dramatically clear from the case of dioxin emissions from the PVC life cycle, the magnitude of the missing or poorly characterized flows may well overwhelm all other impacts, reversing the LCA analytical judgment of relative environmental impact when comparing materials.
This is not meant to dismiss LCAs as a useful tool for the designer. LCAs can be very helpful in assisting the designer to understand many significant environmental impacts of a material. This case study demonstrates that use of BEES throws a strong spotlight on the enormity of the eutrophication impact and the importance of addressing the agricultural practices involved. The case study, however, also demonstrates that any use of LCAs as a tool to compare materials must be informed by knowledge of where flows are poorly characterized or totally missing and the potential impact on the LCA of those flows. Failing to do this ensures that, at least in some cases such as PVC, the LCA comparison will serve to mask the worst environmental impacts – in this case the carcinogenic impact of dioxins - rather than clarify the tradeoffs and relative merits of the materials. In so doing, uninformed LCA use can be expected to lead to environmentally detrimental material choices.
4. More missing data: Maintenance challenges
Dioxins are not the only significant flows unmeasured by LCAs. Another example is the emissions from flooring materials over the lifetime of the installation. Most LCAs base their flow assumptions about flooring emissions on the total of volatile organic compounds (VOCs) emitted during the first 72 hours after installation. The theory of using this as a proxy for all emissions from the floor is that VOC emissions from flooring products rapidly decline after installation and become insignificant so only the initial emissions are meaningful. This approach misses several other significant flows from installation that may far exceed the impact of the 72 hour post-installation VOC emissions, including both maintenance flows, such as VOCs and sewage loads from wax and strip cycles and semi-volatile organic compound (SVOC) emissions such as phthalates from PVC (Lundgren 2002), that occur over a much longer time frame than the VOCs.
The emissions from cleaning, stripping and waxing maintenance activities are likely to be orders of magnitude higher than the 72 hour post-installation VOC emissions. A recent LCA study found that the amount of VOCs emitted from a single waxing of a floor is comparable to the amount of VOCs emitted from the flooring itself over its entire life (Norris 2003). Two materials with different maintenance requirements are likely to have vastly different lifetime VOC emissions that are likely to dramatically alter the relative health impact balance between the materials.
Accurately including the maintenance flows and properly weighting the impact of the exposures in an LCA intended for universal use is very difficult, as maintenance procedures and exposures vary widely from building to building. Rates of maintenance will vary widely depending upon traffic, preferences, and budgets. Staff and patients in a healthcare facility that is fully operational 24/7 are going to have far higher exposure to cleaning chemicals than occupants of a 9-5 office building where maintenance can be scheduled for evenings and weekends when occupancy is low. To account for this, LCA programs will need to gather input from the user on building occupancies and maintenance procedures and build that in to the model. Lacking that, users must be made aware that not only are the maintenance-related flows excluded, but that the impact of those flows could radically impact the indoor air quality results for floorings and other maintained surfaces. Otherwise, again the results of a comparative analysis between two materials with substantially different maintenance regimes will sometimes be quite counter to the true comparative environmental impact.
5. How do we really select materials? Screens versus weightings
Most LCAs are currently configured to be completely based upon relative weightings. Every impact is assumed to have a value that can be exchanged against another impact. For example, two materials will be score about equally in an LCA using the EPA weighting if one has three times the indoor air quality impact but only half the global warming impact of the competing material.
Real world materials selection, however, doesn’t work that way. Specifiers work with a variety of weightings, limits and absolute screens. Minimum standard criteria are established by building codes for concerns such as flame spread. The building code is absolute on this issue, not allowing more flame spread in exchange for more of another value like structural strength. Owner values may also determine absolute criteria ranging from color and pattern, to life expectancy and maintenance requirements. Owners may also place a combination of criteria on other factors, such as cost. These might include both absolute criteria limit (must cost no more than $X/yd) and a weighting (below that maximum price a cheaper product may outweigh another value like ease of maintenance). Environmental impact analyses can and should be similarly subject to more than just comparative weighting. Just as fire safety provided a strong rationale for setting absolute standards for flame spread and screening out inappropriate materials, other environmental and human health and safety concerns establish a strong rationale for setting standards and screens on the chemical flows resulting from materials selection.
In some cases the rationale may
be for a not-to exceed chemical standard, as in the case of the
In other cases, chemicals have
been clearly determined to be sufficiently harmful to warrant outright
elimination. CFCs and PCBs have been banned for use in the
6. Accommodating Data realitIES: Suggestions for LCA enhancementS
To avoid the pitfalls identified here, LCA tools like BEES should consider incorporating a number of enhancements:
Data uncertainties: Identify the significant uncertainties and quantification controversies in the data and flag these uncertainties numerically and graphically to show the end user the effect they may have on the end results through error bars and similar tools.
Use phase flows: Obtain user inputs on use patterns, preferences and maintenance procedures and build that in to the models to allow modeling of use- and maintenance-related flows.
Chemical restrictions and screens: Build absolute chemical and maximum concentration level screening into the model to allow application of legal limits, health research based standards, international treaty obligations and precautionary specifications to the environmental flows. Prime chemicals for absolute limits are persistent bioaccumulative toxics and others whose manufacture, use or disposal results in generation or release of carcinogens (cancer causing chemicals), teratogens (chemicals inducing birth defects in the developing fetus), reproductive toxicants (chemicals that damage the functions of the reproductive system), developmental toxicants (chemicals that stop or misdirect human development), or endocrine disruptors (chemicals that disturb the operation of the endocrine system, affecting development and other key bodily functions). This process will need to be one that continues to be updated for a long time with ample use of precautionary principles for protective insurance as we await testing on thousands of uncharacterized chemicals[††].
7. Conclusion
LCAs are truly a double-edged sword. On the one hand they have the potential to provide the design community with highly important information in the search for the most environmentally friendly and healthy materials. On the other hand, the current reality is that they provide just one portion of the picture in any comparative analysis of materials. LCAs, such as BEES, do not generally incorporate information into their analyses about environmental health issues that are not yet well quantified, are affected by user patterns, are precautionary or are subject to maximum limits or absolute restrictions. By being portrayed as total analyses of environmental flows, LCAs run the danger of lulling the design professional user into thinking that the provided material comparison is complete and does not require any additional analysis.
A series of enhancements have been described that would help flag uncertainty issues, incorporate use variability and acknowledge chemical restrictions and screens, bringing each of these explicitly into the LCA analysis.
Even with these enhancements, it remains critically important for the user to understand the limitations of quantitative analysis in the face of scientific uncertainty. Put simply, this means that science understands the existence and importance of many significant environmental health hazards - like dioxin emissions - for which it can not yet provide reliable numbers to plug into LCA type analyses. The lack of those numbers means that these issues do not show up in a quantity based LCA as currently designed. The health impacts, however, do not go away and we can ill afford to ignore them. For many of these issues, preemptive precautionary action to limit or exclude use of materials involving certain targeted chemicals is the wiser, more responsible and - considering potential liability issues - sometimes more economically conservative course than waiting for certain scientific quantification.
Even with the best data available and more enhancements, LCAs will remain a tool that can only provide one important part of the puzzle - the well quantified part. It will remain critical for the design community to recognize the limitations on what quantification can do to guide materials decision making. We must continue to take responsibility for understanding the challenges and controversies around our materials choices and continue to make value judgments on the precautionary issues at the boundaries of scientific certainty.
8. References
Lent, Tom. 2003. Review
of the
Lippiatt, Barbara C. 2002. BEES 3.0
Technical Manual and User Guide. NISTIR 6916. National
Lundgren, B., et al.
2002. “Small Particles Containing
Phthalic Esters in the Indoor Environment – A Pilot Study”. Proceedings: Indoor Air
2002. p.153.
Montgomery, Margaret. 2003.
“Life Cycle Assessment Tools” Architecture
Week. 20 August 2003 P. E2.1.
NIST (U.S. National Institute of Standards and Technology). 2002. BEES (Building for Environmental and
Economic Sustainability). 9/20/03. http://www.bfrl.nist.gov/oae/software/bees.html.
Norris, Greg, et al. 2003. “Indoor
Exposure in Life Cycle Assessment: A Flooring Case Study. life-cycle assessment.” Harvard School of Public Health unpublished
paper. quoted in “Floorcoverings: Including
Maintenance in the Equation.” Environmental Building News. Vol. 12,
No. 5, p. 12.
Thornton, Joe, PhD. 2002. Environmental
Impacts of Polyvinyl
UNEP. 2000. United Nations Environment Programme. Persistent Organic
Pollutants,. http://www.chem.unep.ch/pops/.
Vinyl Institute. 1998. Website. “Environmental Attributes of
Vinyl, Vinyl Flooring Comes First in Life-cycle Assessment”. Environmental
Briefs. December 1998.
9/5/2003. http://www.vinylinfo.org/attributes/resource.html.
[*] A typical description states: “LCA analyzes the total environmental impact of all materials and energy flows, either as input or output, over the life of a product from raw material to end-of-life disposal or rebirth as a new product.” (Montgomery 2003)
[†] Eutrophication is the addition of mineral nutrients to the soil or water – in this case primarily from agricultural practices used to raise the flax seed - which can increase algae growth, which in turn can lead to lack of oxygen, impacting aquatic life.
[‡] Dioxins are actually a family of chemicals – all potent but of varying toxic strength. “g TEQ” refers to “grams toxic equivalent," a quantitative measure of the combined toxicity of a mixture of dioxin-like chemicals in reference to 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD).
[§] Alexandre Rossin, PricewaterhouseCoopers LLP, Personal communication 10/17/03.
[**] Analysts from linoleum manufacturer Forbo have done their own internal LCAs and assert that the factor for eutrophication from linoleum used in BEES is 2 orders of magnitude (100X) too high and suspect a misunderstanding of functional units may have contributed to the error. If the current BEES eutrophication category value (.0454) is reduced by 99%, the total environmental impact of linoleum drops from .0521 to .0071, flipping the results. Linoleum ends up beating VCT with only 55% of the measured environmental impact – even before having accounted for all of VCT’s dioxin related flows.
[††] The task of identifying chemical hazards and screening our materials choices for environmental and human health and safety will be an ongoing one. Complete basic publicly available toxicity information (the Screening Information Data Set, or SIDS) is available for less than 10% of the roughly 2,800 high production volume chemicals (those produced in volumes over one million pounds per year). No toxicity information at all is available for more than 40%. Even less is known about the tens of thousands more chemicals that are produced in smaller volumes of for any of these chemicals in combination with each other. (US EPA 1999, Lowell 2003)