United States
          Environmental Protection
          Agency
            Office of Research and
            Development
            Washington DC 20460
EPA/600/R-00/023
July 2000
US EPA Office of Research and Development
An International
Workshop on Life Cycle
Impact Assessment
Sophistication

                       eli^t^^^rA
                  ^n.1 fr '1 * iVS^
                  n&'CT&Tk

-------

-------
                                                        EPA/600/R-00/023
                                                             July 2000
An  International Workshop on  Life Cycle
    Impact Assessment Sophistication
                          Co-Organized by:
                            Jane C. Bare
                  U.S. Environmental Protection Agency (EPA)

                         Helias A. Udo de Haes
                    Centre of Environmental Science (CML)

                          David W. Pennington
          U.S. EPA Oak Ridge Institute for Science and Education (ORISE) Fellow

            in co-operation with United Nations Environment Programme (UNEP)
                  National Risk Management Research Laboratory
                     Office of Research and Development
                     U.S. Environmental Protection Agency
                         Cincinnati, Ohio 45268
                                                     Printed on Recycled Paper

-------
                                       Notice
    The views expressed in these Proceedings are those of the individual authors and do not neces-
sarily reflect the views and policies of the U.S. Environmental Protection Agency (EPA). Scientists in
EPA's Office of Research and Development have prepared the EPA sections, and those sections
have been reviewed in accordance with EPA's peer and administrative review policies and approved
for presentation and publication.
                                          11

-------
                                     Foreword
    The U.S. Environmental Protection Agency is charged by Congress with protecting the Nation's
land, air, and water resources. Under a mandate of national environmental laws, the Agency strives to
formulate and implement actions leading to a compatible balance between human activities and the
ability of natural systems to support and nurture life. To meet this mandate, EPA's research program
is providing data and technical support for solving environmental problems today and building a sci-
ence knowledge base necessary to manage our ecological resources wisely, understand how pollut-
ants affect our health, and prevent or reduce environmental risks in the future.

    The National Risk Management Research Laboratory is the Agency's center for investigation of
technological and management approaches for  reducing risks from threats to human health and the
environment. The focus of the Laboratory's research program is on methods for the prevention and
control of pollution to air, land, water and subsurface resources; protection of water quality in public
water systems;  remediation of contaminated sites and ground water; and prevention and control of
indoor air pollution. The goal of this research effort is to catalyze development and implementation of
innovative, cost-effective environmental technologies; develop scientific and engineering information
needed by EPA to support regulatory and policy decisions; and provide technical support and informa-
tion transfer to ensure effective implementation of environmental regulations and strategies.

    This publication has been produced as part of the Laboratory's strategic long-term research plan.
It is published and made available by EPA's Office of Research and Development to assist the user
community and to link researchers with their clients.
                                             E.Timothy Oppelt, Director
                                             National Risk Management Research Laboratory
                                             in

-------
                                        Preface
   This is the report of the second workshop on a topic in the field of Life Cycle Assessment organized
 under the auspices of UNER The first one was held in June 1998 in San Francisco under the title
 Towards a Global Use of LCA". An important characteristic of'that workshop was the strong input
 from experts of developing countries.

   The second workshop was in Brussels in November 1998, and focused on the sophistication of Life
 Cycle Impact Assessment. Since its coming out in the early seventies, the focus of LCA has been on
 the Inventory Analysis. Life Cycle Impact Assessment, or LCIA, was for a long time regarded "to be in
 an early stage of development".This changed with two SETAC working groups, both in North America
 and Europe, and with the unifying guidance of the ISO process. In the coming ISO standard on LCIA
 only the framework of LCIA will be defined. This still allows for very diverging types of analysis that all
 can claim to be ISO compatible. In itself there is nothing wrong with that, but a second SETAC-Europe
 working group on LCIAhas begun the process of providing more guidance in this area.

   An important issue in this field regards the present topic, the appropriate level of sophistication of
 LCIA. As will be clear from the present report, this involves quite a number of issues A major point
 concerns the extension of the characterization modeling to include also the fate of the substances
 and not only their effects. Another issue concerns a possible differentiation in space and time. Studies
 can include impact models that use  data just at world level  and do not specify time periods; in
 contrast, more recent options involve spatial differentiation of impacts and distinguish between differ-
 ent time periods. A further point concerns the type of modeling. Traditionally LCIA uses linear model-
 ing, like in the  Inventory Analysis; but  more sophisticated possibilities arise which take background
 levels of substances into account and make use of nonlinear dose-response functions An important
 question here is whether there are real science based thresholds, or whether these thresholds are
 always of a political origin. A further question  relates to the role and practicality of including uncer-
 tainty analysis. Sensitivity  analysis is  increasingly included in LCA studies; but  this is not yet the
 case for uncertainty analysis. Finally, there is the question of how to apply these different options for
 sophistication of LCIA; which applications can afford to keep it simple, and for which applications is a
 more detailed analysis needed?

  Like with the first workshop in this series, the partners in the process were the US EPA and Centre
 of Environmental Science (CML). We can expect that it will not be the last one. An issue that could
 only be touched upon indirectly  during the second workshop concerns  the choice of the category
 indicator in the environmental mechanism of an impact category. Should it be chosen at midpoint
 level, as is done traditionally in many categories, or should the modeling proceed up to the level of the
category endpoints?  We expect that this will  be  a topic for a third workshop  in  this series  to be
organized in  May 2000 in Brighton.
                                            IV

-------
                               Table of  Contents
Pre-Workshbp Summary	

Summary of Technical Issues Discussed
.VI

..1
Agenda	9

Remarks by Co-Chairs
Life Cycle Impact Assessment Sophistication (Jane Bare, U.S. EPA, USA)
LCIA Sophistication Issues Overview (David Pennington, ORISE Fellow, U.S. EPA, USA)
Bio of Helias Udo de Haes, CML.The Netherlands

Determination of Sophistication
ISO Standards: Life Cycle Assessment & Comparative Assertions	17
(J.Willie Owens, Procter & Gamble, USA)

Impact Assessment for Ecodesign	25
(MarkGoedkoop, Pre Consultants, The Netherlands)

Application Dependency of LCIA	31
(Henrik Wenzel, Technical University of Denmark)

Impact Categories & Uncertainty Analysis within HumanToxicity & Ecotoxicity
Types of Uncertainty in the Human Toxicity Potential	35
(Edgar Hertwich, University of California, Berkeley, USA)

The Different Levels of Uncertainty Assessment in LCI A; The Case of Carcinogenic
Effects	39
(Patrick Hofstetter, Environmental Sciences: Natural and Social Science Interface, ETH Zurich)

Human Toxicity and Ecotoxicity - Modeling versus Scoring	46
(Olivier Jolliet, EPFL, Switzerland)

Priority Assessment of Toxic Substances in LCA—A Probabilistic Approach	54
(MarkHuijbregts, IVAM, Netherlands)

Acidification, Eutrophication, andTropospheric Ozone
Implications of Inventory Structure for Life Cycle Impact Assessment and Uncertainty
Analysis	61
(Gregory A. Norris, Sylvatica, Maine, USA)

Levels of Sophistication in Life Cycle Impact Assessment of Acidification	67
(Jose Potting and Michael Hauschild, Inst. for Product Development, Technical U. of Denmark)

Eutrophication — Aquatic and Terrestrial — State of the Art	..77
(Goran Finnveden, fms and Stockholm University)

Appendix	83

-------
                                     Pre-Workshop  Summary

                              Jane Bare, Helias Udo de Haes, and David Pennington
  In meetings and journals worldwide, practitioners have
debated the utility of conducting Life Cycle Assessments
(LCAs).The debate has often hinged on the appropriate level
of sophistication. While some have advocated abandoning
LCA altogether, since it is not achievable in its most so-
phisticated form, others have supported the concept of con-
ducting LCAs while making the limitations and uncertain-
ties transparent.

  This proposed workshop is needed to address the appro-
priate level of sophistication in Life Cycle Impact Assess-
ment (LCIA).The workshop will be formulated to help evalu-
ate available approaches in terms  of framework, scope,
impact assessment methodologies and uncertainty analy-
sis. Additionally this workshop will serve to educate poten-
tial practitioners in developing countries about issues con-
fronting LCIA, and to include developing countries in small
working group discussions. A number of prominent practi-
tioners and researchers will be invited to present a critical
review of the associated factors, including the current limi-
tations of available impact methodologies and a compari-
son of the alternatives in the context of uncertainty. Each
set of presentations will be followed by a discussion ses-
sion to encourage an international discourse with a view to
improving the understanding of these crucial issues. Dis-
cussions will be focused around small working groups of
LCA practitioners and experts including representatives from
industry, government, and academia.

Introduction
  A UNEP Workshop titled Towards Global Use of LCA was
held on June 12 -13,1998 in San Francisco. The purpose
of this workshop was to develop recommendations and an
action plan that would lead toward greater use of LCA in the
context of sustainable development. Fifty LCA users and
experts from 26 different countries attended the workshop.
At the end of the workshop, each of the participants was
asked what actions could lead to greater development and
use of LCA for sustainable development decision making.
One of the many ideas discussed included providing a fo-
rum for an  International discussion of the appropriate levels
of sophistication within LCIA.

  Practitioners of  LCA are faced with the task of trying to
determine the appropriate level of sophistication to provide
a sufficiently comprehensive approach to assist in environ-
mental decision making. The impact assessment stage of
LCA, termed LCIA, helps decision-makers to interpret in-
ventory data in the context of a number of impact catego-
ries.

  In an ideal world, an LCIA would be based on high quality
data. All impact categories and all steps in each cause-
effect chain would be considered using state-of-the-art meth-
ods, which include fate and account for spatial variation. In
this ideal world, decisions could be made based on these
environmental assessments with a high level of certainty.
In the real world,  practitioners  have had to deal with the
questions of inaccuracy and uncertainty in LCA and LCIA
in particular. Hence, many simplifications are made within
the LCIA stage. Simplifications may include: reduction in
spatial discrimination (or ignoring spatial variation altogether),
ignoring fate, ignoring background levels of pollutants, as-
suming linear dose-response curves, and/or eliminating an
impact category altogether because the appropriate assess-
ment methodologies do not exist.

  While ideally an impact assessment should be sophisti-
cated in all dimensions, obviously, high levels of sophisti-
cation require exhaustive time, data, and resources, or can-
not be reached at all. Hence, the scope of the assessment
needs to be defined, possibly iteratively, to provide the ap-
propriate level of sophistication. Defining this scope appro-
priately  is key to effective environmental decision making.

  Conducting an uncertainty analysis can be one approach
to determining the appropriate level of sophistication. Un-
fortunately, many of the studies conducted in the past were
deterministic and therefore, may have misled commission-
ers into believing that decisions could be based on the stud-
ies, when in reality, the compared options were essentially
non-differentiable.This workshop will explore the sophisti-
cation of LCIAs including the discussion of uncertainty analy-
sis involved.

Factors Involved in the Determination of
     Sophistication
  The important issue of deciding the appropriate level of
sophistication is not typically addressed implicitly in LCA.
Often, the determination of sophistication is based on con-
                                                     VI

-------
siderations which may, or may not, be appropriate, but which
may include practical reasons for limiting sophistication (e.g.,
the level of funding). A discussion of the most appropriate
ways of determining sophistication will include:

   • project objective

   • an uncertainty and/or sensitivity analysis

   • the inventory data and availability of accompanying pa-
    rameters

   • depth of knowledge and comprehension in each impact
    category

'  • perceived value placed on the specific impact catego-
    ries

   • the anticipated level of impact

   • the quality and availability of modeling data

   • available supporting software

   • the level of funding

Hope for the Future
   There have been a number of advances made in LCIA in
recent times.

   • The framework for LCIA is becoming standardized, en-
    hancing the comparability and avoiding unnecessary
    variation between studies.

   • The fate of substances is increasingly taken into ac-
    count, in particular using multimedia modeling as a basis
    for characterization.

   • The results of different characterization procedures for
    the same category are compared amongst each other
    and they show convergence.

   • Better distinctions are being made between scientific
    information and value choices.

   These developments are leading to advances in the prac-
tice of LCIA, but the major limitations and uncertainties can-
 not be expected to go away in the near future. Instead, practi-
tioners must learn how to best address the concerns and
 limitations of the methodologies. Among the questions that
 remain, the following list will be tentatively posed within small
 group discussions:

 Determination of Sophistication
 Methods for
   What are the most common methods by which the level of
 sophistication is determined?

   Which methods are considered more acceptable? Why?

   What are the barriers to using the acceptable methods?
What can be done to overcome these barriers?

Various Uses of LCIA
  Is LCIA application dependent? Why or why not?

  What are the expectations in sophistication for the vari-
ous LCA uses (e.g., government, industry/public communi-
cation, and internal use)?

  When should LCIAs be as detailed as possible, because
every increase of detail increases accuracy and removes
uncertainty? And when is it better to limit the scope of LCA
to addressing questions on a macroscopic scale,  leaving
spatial and threshold consideration to other analytical tools?

  How do practitioners determine the value of the trade-offs
necessary when sophistication and comprehensiveness are
"at odds"  (e.g., choosing a modeling approach may limit the
number of pollutants that can be characterized vs. a scor-
ing approach which may limit the sophistication of the mod-
eling)?

Uncertainty Analysis
  What case studies are available using uncertainty analy-
ses within LCIA? And what are the major findings to date
(levels of uncertainty discovered)? When is the uncertainty
determined to be unacceptable?

Specific Impact Categories & Uncertainty
     Analysis within LCIA

Human Toxicity and Ecotoxicity
  In human toxicity and ecotoxicity, when is spatial and/or
temporal differentiation necessary? If necessary, what spa-
tial and/or temporal details are recommended (e.g., indoor/
outdoor, and height of emission point)?

  In ecotoxicity what is the best approach to addressing
multiple species? If suggested, what are the recommended
representative species?

  In human toxicity and ecotoxicity, what are the greatest
barriers to conducting uncertainty analysis?

  What are the recommendations for research and/or de-
velopment in these impact categories?

Eutrophication, Acidification, and
      Tropospheric Ozone Formation
  In each category, when  is spatial and temporal differen-
tiation necessary? If so, what spatial and temporal scales
are recommended (e.g., regionally, or differentiating terres-
trial, sweet water and marine)?

   In each category, what are the greatest barriers to con-
ducting uncertainty analysis? What are the recommenda-
tions for research and/or development in these impact cat-
egories?
                                                     vn

-------

-------
                         Summary of Technical  Issues  Discussed
                       Jane C. Barei, David W. Penningtonz, and Helias A. Udo de Haes3
Abstract
  On November 29 - 30,1998 in Brussels, an international
workshop was held to discuss Life Cycle Impact Assess-
ment (LCIA) Sophistication. Approximately 50 LCA experts
attended the workshop from North America, Europe, and
Asia. Prominent practitioners and researchers were invited
to present a critical review of the associated factors, includ-
ing the current limitations of available impact assessment
methodologies and a comparison of the alternatives in the
context of uncertainty. Each set of presentations, organized
into three sessions, was followed by a discussion session
to encourage international discourse with a view to improv-
ing the understanding of these crucial issues. The discus-
sions were focused around small working groups of LCA
practitioners and researchers, selected to include a balance
of representatives from industry, government and academia.

  This workshop provided the first opportunity for Interna-
tional experts to address the issues related to LCIA So-
phistication in an open format. Among the topics addressed
were: 1) the inclusion or exclusion of backgrounds and thresh-
olds in LCIA, 2) the necessity and practicality regarding the
sophistication of the uncertainty analysis, 3) the implica-
tions of allowing impact categories to be assessed at "mid-
point" vs. at "endpoint" level, 4) the difficulty of assessing
and capturing the comprehensiveness of the environmental
health impact category, 5) the implications of cultural/philo-
sophical views, 6) the meaning of terms like science-based
and environmental relevance in the coming ISO LCIA stan-
dard, 7) the dichotomy of striving for consistency while al-
lowing the incorporation of state-of-the-art research, 8) the
role of various types of uncertainty analysis,  and 9) the role
of supporting environmental analyses (e.g., risk assess-
ments). Many of these topics addressed the need for in-
creased sophistication in LCIA, but recognized the conflict
1 Corresponding author: U.S. Environmental Protection Agency, National Risk Man-
agement Research Laboratory, Cincinnati, Ohio 45268, USA. Email:
bare.jane@epa.gov
2 ORISE Research Fellow, U.S. Environmental Protection Agency, National Risk
Management Research Laboratory, Cincinnati, Ohio 45268, USA.
3 Centre of Environmental Science (CML), Leiden University, P.O. Box 9518, NL-
2300 RA Leiden, Netherlands.
this might have in terms of the comprehensiveness and
holistic character of LCA, and LCIA in particular.

Introduction
  A UNEP Workshop titled 'Towards Global Use of LCA"
was held on June 12-13,1998 in San Francisco. The pur-
pose of the San Francisco workshop was to develop rec-
ommendations and an action plan that would lead towards
a greater use of LCA in the context of sustainable develop-
ment. At the end of the San Francisco workshop, each of
the participants was asked what actions could lead to greater
development and use of LCA in sustainable development
decision making. One of the many ideas suggested was to
provide a forum for an International discussion of the appro-
priate practice of LCIA. LCIA Sophistication was taken up
as subject of the workshop held in Brussels, which was
attended by approximately 50 LCA practitioners and experts
from various countries.

  Practitioners of LCA are faced with the task of trying to
determine the appropriate level of sophistication in order to
provide a sufficiently comprehensive and detailed approach
to assist in environmental decision making. Sophistication
has many dimensions and dependent upon the impact cat-
egory, may simulate the fate and exposure, effect and tem-
poral and spatial dimensions of the impact. (Udo De Haes,
1999a, Owens, et al., 1997, Udo de Haes, 1996, Fava, et
al., 1993). In the context of the Brussels workshop, sophis-
tication was considered to be the ability of the model to
accurately reflect the potential impact of the stressors, or
in language more consistent with recent ISO publications,
the ability to reflect the environmental mechanism with sci-
entific validity. (ISO, 1999).

  The impact assessment phase of LCA,  termed LCIA,
helps decision-makers interpret inventory data in the con-
text of a number of impact categories and to bring them into
a more surveyable format. Ideally, an LCIA would be based
on high quality data. All impact categories and processes
in the environmental mechanism of each  of these catego-
ries would be considered using state-of-the-art techniques,
which would fully account for spatial and temporal variation.
In such an Ideal World, decisions would be made based on

-------
 these assessments with a high level of confidence and cer-
 tainty. However, real world practitioners have to deal with
 limitations (e.g., budget, and poor quality data) and simplifi-
 cations are made. Some modifications may include: 1) re-
 duction in spatial and temporal discrimination (or ignoring
 these dimensions altogether), 2) ignoring fate, 3) assuming
 linear dose-response curves and/or 4) eliminating an im-
 pact  category because appropriate data or assessment
 methodologies do not exist.

  While ideally an impact assessment should be sophisti-
 cated in all dimensions, this high level of sophistication re-
 quires exhaustive time, data, and resources and generally
 cannot  be reached due to limitations in methodology and
 data available. Hence, the scope of the assessment needs
 to be defined, possibly iteratively, to provide the appropriate
 level of sophistication, including the required level of detail
 and accuracy, together with an uncertainty analysis practi-
 cal for individual studies, and the specification of value
 choices within the framework of the LCA. Appropriate defi-
 nition of this scope,  including sophistication, uncertainty
 analysis, and comprehensiveness is the key to effective
 environmental decision making.

  Many practitioners in the past have attempted to evalu-
 ate impacts to support broad LCA-based decisions, but have
 oversimplified the impact assessment step. Unfortunately,
 limitations in simulation sophistication lead to a reduced
 ability of the study to answer the questions at hand with a
 high degree of certainty. In the absence of accompanying
 uncertainty analysis, and validation (which addresses model
 uncertainty), many LCAs are conducted at such a low level
 of simulation sophistication that they are ineffectual in dif-
 ferentiating the very options they are trying to evaluate.
 (Coulon, 1997; Potting, 1997, Udo de Haes, 1996). Work-
 shop participants also discussed the dichotomy of sophis-
 tication and comprehensiveness. As an example, very sim-
 plistic methods such as relying solely on toxicity data may
 allow a larger chemical  database set than a more sophisti-
 cated approach which would require additional chemical and
 physical properties to determine the relative human health
 potentials.

  More recently, researchers are recognizing the many types
 of uncertainty involved  in environmental decision making.
 Two types of uncertainty discussed at this workshop were
 model uncertainty and  data uncertainty. Data uncertainty
 may be  estimated by the propagation of uncertainty and
 variability of the input parameters. Model uncertainty can
 only be characterized by comparison of the model predic-
tion with the actual response of the system being addressed.
As  data uncertainty is relatively  easy to characterize,
whereas model uncertainty is difficult, especially in a field
 like LCIA, the presentation of data uncertainty alone may
 not appropriately be used to compare two methodologies.
 For example, a simplistic approach utilizing only persistency,
 bioaccumulation, and toxicity data may appear to be more
certain when compared in terms of data uncertainty to a
 more complex multimedia/human exposure approach, but
the unaddressed model uncertainty may significantly over-
shadow the data uncertainty.
  The specification of value choices has a bearing on the
 level of sophistication and has been the subject of many
 recent papers. (Owens, 1998, Finnveden, 1997, Volkwein,
 1996a, Volkwein, 1996b, Powell, 1996, and Grahl, 1996).
 Some practitioners are uncomfortable with the subjectivity
 of the Valuation Process, but fail to recognize the role of
 subjectivity in other phases of the LCA framework. All LCAs
 are conducted under the influence of subjective decisions.
 In fact, subjective decisions, value choices, or scientific or
 engineering judgements are made throughout the LCA pro-
 cess.Thus, the selection, aggregation, or disaggregation of
 impact categories and the determination of the methodolo-
 gies to quantify the potential impacts are all influenced by
 value choices. The Brussels workshop was chosen to ex-
 plicitly address the incorporation of value choices within
 the LCA process.

  Unfortunately, the important issues of deciding the ap-
 propriate level of sophistication often remain unaddressed
 in LCIA. The determination of the level of sophistication is
 often not based on sound and explicit considerations, but
 on practical reasons (e.g. the level of funding, level of in-
 house knowledge). The workshop was therefore formulated
 to allow a more  explicit discussion of the many factors out-
 lined above that can influence the choice of the level of
 sophistication of a study, including:

  • The project objective

  • The perceived value placed on the specific impact cat-
    egories

  • The availability of inventory data and accompanying
    parameters

  • The depth of knowledge and comprehension in each
    impact category

  • The quality  and availability of modeling data

  • The uncertainty and sensitivity analyses

  • The level  of validations

  • The available supporting software

  • The level  of funding

  This paper provides a summary of the results of this work-
shop, including discussion on many of the above topics. An
attempt is made to provide short reviews of the presenta-
tions and discussions.  However, in documenting the work-
shop it was not possible to capture the full detail of the
many points raised.

Workshop  Logistics
  On the 29th and 30th of November, 1998 in Brussels,
Belgium an international workshop was held to discuss Life
Cycle Impact Assessment (LCIA) Sophistication. Approxi-
mately 50 LCA experts attended the workshop, coming from
Europe, Asia and the USA. Several prominent practitioners
and researchers were invited to present a critical review of

-------
the associated factors, including the current limitations of
available impact methodologies and a comparison of the
alternatives in the context of uncertainty. Each set of pre-
sentations, organized into three sessions, was followed by
a discussion session to encourage international discourse
with the aim to improve the understanding of these crucial
issues. The discussions were focused around small work-
ing groups of LCA practitioners and researchers, deliber-
ately selected to include a balance of representatives from
industry, government and academia. Each group was given
the charge to address the questions that most interested
them, as opposed to assigning specific groups with spe-
cific questions.

Introductory Session
  Jane Bare of the U.S. Environmental Protection Agency
(EPA) opened the workshop noting that many of the partici-
pants had been involved in previous meetings as LCIA ex-
perts, sometimes even discussing related issues in the
development of ISO 14000 series and SETAC Working
Groups on LCA and LCIA. Requirements are being devel-
oped under IS014042 to specify a high level of sophistica-
tion for Comparative Assertions, including language con-
cerning the scientific validity, environmental relevance, and
the role of value choices. Within SETAC-Europe efforts are
on going to develop a document related to the selection of
the "state-of-the-art" impact assessment methodologies.
Bare asked that participants consider the present workshop
as a more open format than either of these settings to allow
a completely uninhibited technical exchange. She stressed
that Life Cycle Impact Assessment can be effective in sup-
porting environmental decision making, but only if the data
and methods are sufficiently scientifically defensible. Sci-
entifically defensible was defined as being dependent upon
the level of sophistication, the level of certainty (including
both data and model certainty), the level of comprehensive-
ness, and data availability. The participants were challenged
to address several additional questions throughout the two
days of discussions including: What is "scientifically defen-
sible"? In the sphere of determining whether impact assess-
ment is based on sound science, where does one draw the
line between sound science and modeling assumptions?

   Garrette Clark from the  United Nations  Environment
Programme (UNEP) then provided a short history of UNEP's
involvement in the area of LCA, which includes providing
technical assistance to developing countries and the devel-
opment of an associated guidance document for LCA (UNEP,
 1996). She stated that LCA is considered by UNEP to be an
 important tool for achieving cleaner production and consump-
tion. She also summarized findings from the recent LCA
workshop in San Francisco in June 1998 (UNEP, 1998).

   David Pennington discussed two extremes of LCIA so-
 phistication. One extreme he called the "Contribution or
 Burden" approach, which is comparable to what has been
 historically used in LCIA (reflecting the Precautionary Prin-
 ciple and the combinatory potential to cause  impacts).The
 other extreme, the "Consecutive Risk Assessment" ap-
 proach, he noted as being particularly recommendable for
use in areas with high stakes, such as comparative asser-
tions, but as often limited to the assessment of chemicals
in isolation. He introduced the question concerning the need
for spatial differentiation and asked when site-specific dif-
ferentiation was appropriate. He also pointed out that the
category indicators are chosen at different points in the
environmental mechanism (or cause-effect chain), and
stated that the U.S. EPA has been  using the term of "mid-
point" to address indicators that stop short of expected ef-
fects on the final "endpoint" of the  environmental mecha-
nism. He presented acidification as an example of a cat-
egory with the indicator at "midpoint" level and human health
as a possible example of a category with the indicator at
"endpoint" level. He concluded by asking about the different
levels of sophistication. What is possible? What is required?
When to use the various levels of sophistication?

Session One: Overview
  Willie Owens of Procter and Gamble spoke about com-
parative assertions (i.e., public comparisons between prod-
uct systems) and the requirements  for LCIA under ISO
14042. He stated that ISO 14042 requires a sufficiently
comprehensive set of category indicators, a comparison
conducted indicator by indicator (i.e. no weighting) and that
LCIAs should not be the sole basis for comparative asser-
tions. Current language in IS014042 states that subjective
scores, such as weighting across categories, shall not be
used for comparative assertions; that category indicators
be scientifically defensible and environmentally relevant and
that sensitivity and uncertainty analyses shall be conducted.

  Mark Goedkoop of Pre Consultants discussed LCIA for
ecodesign. He pointed out that the point of conducting an
LCA is typically to determine whether A is better than B. He
then presented three problems with LCA and ecodesign: 1)
LCA studies are too time consuming, 2) LCA studies are
hard to interpret, and 3) Designers never become experts,
but remain dependent upon experts. His proposed solution
for these problems was to calculate predefined single scores
for the most commonly used materials and processes, and
to incorporate uncertainty into the modeling. He also dis-
cussed the sometimes hidden role  of societal  values in
characterization modeling, even for internationally agreed
models. As an example, he presented the three classes of
carcinogens (proven, probable and possible)  and pointed
out that the practitioner must make a decision about whether
to include one, two, or all three classes. He proposed that a
single truth does not exist and that modeling is dependent
upon the chosen perspective. He then introduced three dif-
ferent views of the world based on values: egalitarian, hier-
archical and individualist. (A topic  discussed later in more
detail by Patrick Hofstetter.) He pointed out that if A is not
better than B in all three cases then the result is dependent
 upon the perspective.

   Henrik Wenzel of the Technical University of Denmark
discussed the application dependency of LCIA. He men-
tioned several applications including life cycle management,
 strategic planning, product development, process design,
 green procurement and public purchasing, and marketing.

-------
 In addition, he discussed three main variables governing
 application dependency: the environmental consequence of
 the decision (including spatial and temporal scale), the so-
 cioeconomic consequence and the decision context. He dis-
 cussed the application dependency of uncertainty, trans-
 parency, documentation and the inclusion of temporal and
 spatial resolution. He stated that the need for sophistication
 of LCIA is largest in decisions with the highest requirements
 for certainty. He also stated that the decision-maker might
 impact the choice of normalization and weighting. (Wenzel,
 1998)

   Helias Udo de Haes wrapped up this first session by pro-
 viding a summary of some  of the key points covered and
 challenging the participants to address the questions pro-
 vided during the small group discussions. Workshop par-
 ticipants were asked to address the following questions and
 to provide additional questions to aid discussion.

 Determination of Sophistication and Uncertainty Analysis

   1. What are the most common methods by which the
     level of sophistication is determined?

   2. Which methods are considered more acceptable? Why?

   3. What are the barriers to using the acceptable meth-
     ods? What can be done to overcome these barriers?

   4.To what extent should LCIA be application dependent?

   5. What are the expectations  regarding the level of so-
     phistication for the various LCA applications (e.g., by
     government, by industry, for public communication, and
     for internal use)?

   6. When should LCIAs be as detailed as possible, aiming
     at the maximum level  of accuracy? And when  is it
     better to limit the scope of LCA to addressing ques-
    tions on a  macroscopic scale, leaving spatial and
    threshold considerations to  other analytical tools?

   7. How do practitioners deal with the trade-offs neces-
    sary when sophistication and comprehensiveness are
    "at odds" (e.g., choosing a detailed modeling approach
    that may limit the comprehensiveness vs. a scoring
    approach that may limit the sophistication)?

   8. What case  studies are available using uncertainty
    analyses within LCIA?  And what are the major find-
    ings to date (levels of uncertainty discovered)? When
    is the uncertainty determined to be unacceptable?

Questions Added at Workshop

   9. What is scientifically and technically valid, as included
      in the requirements of IS014042?

   10.  If LCIA is an iterative process, what drives the deci-
     sion on the level of sophistication (e.g., uncertainty
     analysis, relevance, and existence of trade-offs)?
    11. Define uncertainty in the context of LCIA. What pa-
       rameters must be analyzed?

    12. How do we incorporate background levels into LCIA?
       Should  we define working points (as in Mark
       Goedkoop's presentation)? Should this be done for
       individual chemicals or combined?

    13. What is the best currently available method to repre-
       sent the combined effect of chemicals without double
       counting, or inappropriately allocating?

    14. How do we incorporate (or should we incorporate) the
       differing philosophical views in characterization?

 First Session Discussion Summary
   An aggregation of the resultant views is presented below:

   Determination of Sophistication-Many different groups
 commented on  the appropriate level of impact assessment
 sophistication.  One group commented that some sound
 decisions may  be/have  been made on the basis of LCA
 studies, which did not have very sophisticated LCIAs, but
 these tended to  be more obvious cases.They recommended
 using the most sophisticated impact assessment models
 that provide information closest to the endpoint. Another
 group commented that sophistication is dependent upon a
 number of things including: inventory data availability, the
 availability of characterization models and data to support
 these models, objective, the application dependency, the
 decision maker's sphere of influence and the impact cat-
 egory. A third group stated that the choice of sophistication
 depends upon an iterative process, where the iterations may
 be dependent upon uncertainty, the environmental relevance
 of the results and the minimum level of certainty required to
 support a decision. Several participants commented that
 sophistication is often limited by budget, inventory data
 availability, ease of use of  impact assessment methods
 and in-house knowledge.

  These participants stressed the practical side of LCA and
 recognized the difficulty in data collection and the structur-
 ing of public databases to support more sophisticated analy-
 ses.

  Application Dependency-There was a general belief that
 LCIA sophistication is application dependent, according to
 the type of application and not the individual user. For ex-
 ample, screening level LCA studies may not require the rig-
 orous use of sophisticated impact assessment techniques
 but final comparative assertions may require much more
 rigor, particularly if the benefits are not apparent. LCIA stud-
 ies should be performed based on the type of question or
 decision at hand and the purposes that the LCIA may be
 serving.

  Validating the Results of LCIAs- There was agreement
that one cannot  validate the results of a single LCIA study,
because of the  lack of temporal and spatial specification
associated with the inventory data, and an inability to accu-
rately model complex interactions in the environment, in-

-------
 eluding the combinatory effects of chemical mixtures. How-
 ever, input data can be quality checked, and elements in
 the models can be compared with models developed in the
 context of other applications such as environmental risk
 assessment. It was also noted that validation might not be
 as important in  the context of LCIA since models simply
 reflect a relative comparison as opposed to an absolute
 assessment.

   Backgrounds and Thresholds - Practitioners have tried to
 incorporate background levels in LCA studies in the past
 but there was a lot of discussion that this practice may or
 may not be appropriate. One of the questions at hand is
 whether emissions do occur in above or below threshold
 situations. Another issue concerned the fear that defining
 backgrounds and thresholds will lead to treating many envi-
 ronments as infinite sinks (e.g., for acidic chemicals) when
 in reality nature's ability to absorb the impact may be ex-
 ceeded at some future time. The distinction was also made
 that thresholds may be less strict, because of the presence
 of very sensitive species or human individuals.Thresholds
 may also not be protective enough in many environments in
 which the  combined effects of chemicals may cause ef-
 fects at a level much lower than the threshold effect. Fi-
 nally, practitioners were cautioned not to use LCIA to the
 exclusion of recognizing the problem of hot spots surround-
 ing facilities. (See the following point for more information
 on mixtures). On the other hand, some participants believed
 that thresholds might be valuable indicators of relative po-
 tency for many chemicals and that thresholds had been
 derived with statistically sound methods. Further clarifica-
 tion of the  decision-making context may be necessary to
 determine the value of thresholds and backgrounds in par-
 ticular applications of LCIA.

  Mixtures - One of the basic limitations of the current state-
 of-the-science of LCIA of human and ecotoxicity is the in-
 ability to effectively deal with potential combinatory effects
 of chemical mixtures. Toxicologists operate  under the as-
 sumption that chemicals acting on the same organ can be
 considered to have an additive effect, but often LCIA im-
 pact categories are much  broader than a focus on target
 organs. Therefore, the same assumptions used in risk as-
 sessment are not applicable to LCIA. This is especially an
 issue when practitioners try to incorporate threshold levels
 for individual chemicals into LCIA. Because mixtures are
 not well characterized in LCIA, effects may be occurring at
 much lower levels than the accepted threshold levels of the
'individual chemicals. Practitioners often try to compensate
 for these and other model deficiencies by adopting the Pre-
 cautionary Principle.

  Data Gaps-There was a concern that data gaps can be
 significant. Particularly in human and ecotoxicity, availabil-
 ity and quality of both inventory and chemical data to sup-
 port the modeling of a large number of chemicals can be
 frustrating. These impact categories are a good example of
 where less sophisticated screening techniques may, with
 an appropriate degree of caution prove useful.

  Uncertainty Analysis - LCIA still faces great challenges
 before fully addressing uncertainty analysis. Some of these
 challenges include the lack of awareness, lack of associ-
 ated methodology, and the perceived difficulty of present-
 ing the results to decision-makers. Specifically, practitio-
 ners need better knowledge of uncertainties in existing
 methods within the different impact categories and of the
 potential for improvement, if any, by using methods with
 greater sophistication. Many participants acknowledged a
 need for a better understanding of the uncertainty involved
 in each of the impact assessment methodologies for each
 of the impact categories, noting that uncertainty is associ-
 ated with the models as well as the input data. The potential
 trade-off in available models between increased sophistica-
 tion (i.e., detail) and  reduced comprehensiveness (e.g.,
 number of stressors simulated) was again noted.

   Unnecessary Rigor? - There was a belief that the ISO
 standard on LCIA, specifically for the comparative asser-
 tions to be disclosed to the public, is too demanding in the
 areas of scientific validity and certainty. Examples were given
 of some other modeling arenas that face the same chal-
 lenges  (e.g., economic modeling, risk assessment stud-
 ies). In  these fields large uncertainties are accepted, ex-
 pected and (sometimes) clearly documented.There was also
 a concern that the rigor expected of the impact categories
 without a working international acceptance (e.g., human
 toxicity) exceeds the rigor and certainty requirements com-
 pared with the impact categories that benefit from having
 international consensus (e.g., global warming potentials).

  Model uncertainty vs. data uncertainty - Some partici-
 pants commented that the  current disparity in levels of un-
 certainty analysis may have led to the false impression that
 the more sophisticated models have increased uncertainty
 when compared to less sophisticated techniques.Typically
 this is not the case. Usually,  with a more sophisticated model
 the model uncertainty has decreased and the ability to model
 data certainty quantitatively has increased. Deceptively
 (since model uncertainty is not typically characterized) the
 increased characterization of data certainty may have
 seemed to increase total uncertainty. (Additional details on
 uncertainty analysis may be found in Edgar Hertwich's pre-
 sentations.)

  Standardization- While  it was recognized that the level
 of sophistication might depend upon the type of application
 and the availability of data, there was a belief that consis-
 tency of approach  or methodology may be an important
 priority to allow comparability between studies. Some  par-
 ticipants pointed out that certain studies may only require
 Life Cycle Thinking and therefore, should not be subject to
 the standardized methodologies. Others addressed the idea
 of approach hierarchies that differentiate between screen-
 ing and more intensive techniques  but noted that the ap-
 proaches could be consistent within these tiers. It was simi-
 larly noted that there could be a trade-off between sophisti-
cation and comprehensiveness, while one approach pro-
vides a more complete picture but with low level of detail,
another may provide a higher level of detail but at the ex-
pense of comprehensiveness. It was further noted that there
 is continual development of methods and standardization
should not discourage further research efforts.

-------
  More Focused Research - More energy needs to be ex-
pended to ensure that LCA research is focused on areas
that will have the greatest impact. Research needs to be
conducted in deriving better methodologies for more relevant
indicators. Specifically, land  use, habitat alteration, and
environmental toxicity were mentioned as examples of im-
pact categories requiring much more research.

Session Two: Human Health and Ecotoxicity
  Edgar Hertwich of the University of California, Berkeley
opened the session on Human and Ecotoxicity with his pre-
sentation: "A Framework for the Uncertainty Analysis of the
Human Toxicity Potential". He presented the purpose of
uncertainty analysis: "to develop confidence in an analyti-
cal result, as an input to formal decision analysis techniques
and as a tool to  refine impact assessment methods." He
noted that uncertainty analysis includes: parameter uncer-
tainty, model uncertainty, decision rule uncertainty and vari-
ability. He then presented various examples of each of these
as they might pertain to modeling for human toxicity impact
assessment in LCIA. Finally, he pointed out that simply
conducting a sensitivity analysis can often provide valu-
able insights about the significance of the multiple uncer-
tainties involved in the decision and can help refine impact
assessment techniques. (Hertwich, et al., 1993; Hertwich,
1999)

  Patrick Hofstetter of the Swiss Federal Institute of Tech-
nology in Zurich  addressed the question of "What is sci-
ence?" in the presentation: 'The Different Levels of Uncer-
tainty Assessment in LCIA: The Case of Carcinogenic Ef-
fects." He stated that the development of models is depen-
dent on the perspective of the modeler. Three perspectives
were described: hierarchist, individualist and egalitarian. An
individualist optimizes the spending of resources based upon
the known or certain types of harm that can be modeled
(e.g., only choosing to include IARC Group 1 Carcinogenics
in an analysis). A hierarchist could be closest to the operat-
ing positions typically held by government and international
organizations and would include Group 1 and Group 2 Car-
cinogens. Egalitarians tend to take a more risk aversive
and preventive standpoint and thus would include Groups
1,2, and 3 in a carcinogenic analysis. Similarly, these dif-
ferent perspectives would derive different discount rates for
use within an assessment in terms of the Disability Ad-
justed Life Years (DALY). An illustration showed the combi-
nation of the assumptions of all three cultural perspectives
in an eco-index probability graph. Finally, he concluded that
LCIA could be made simple to use and yet robust by incor-
porating the values associated with various perspectives
and allowing an analysis of the related technical, method-
ological and epistemological uncertainties. (Hofstetter, 1998)

  Olivier Jolliet of the Swiss Federal Institute of Technol-
ogy in Lausanne discussed "Human Toxicity and Ecotoxicity
Modeling vs. Scoring." He opened by saying 'Tell me your
results and I will tell you who paid you!" Then he called for
the identification of best available practice regarding impact
assessment methods to reduce the ability to provide LCAs
that support such malpractice. He also proposed that this
process should try to meet the ISO 14042 requirements to
be "scientifically and technically valid" and "environmentally
relevant." After comparing different human toxicity model-
ing efforts, he pointed out parameters and model character-
istics that are important in human and ecotoxicity model-
ing, including exposure and fate uncertainties, that can be
responsible for significant uncertainty and which open op-
tions for reduction of modeling uncertainty by proper em-
pirical or experimental validation. He concluded by saying
that modeling comparisons should be made based on model
characteristics and consistent data.

  Mark Huijbregts of the University of Amsterdam presented
a paper on "Priority Assessment of Toxic Substances in
LCA: A Probabilistic Approach." Citing previous publications
(e.g., Guinee, et al., 1996 and Hertwich, et al., 1998), he
suggested that the following  specific improvements  are
needed: a review of default values with the possibility of
using more realistic values, an inclusion of all relevant envi-
ronmental compartments and inclusion of a Monte Carlo
type of uncertainty analysis. He  presented a probabilistic
simulation of weighted human, aquatic and terrestrial Risk
Characterization Ratios (RCRs) for 1,4-dichlorobenzene and
2,3,7,8-TCDD and demonstrated that only a few substance-
specific parameters are responsible for the uncertainty in
results. Finally, Huijbregts concluded that variability is not
of significance if it is identical for all options being com-
pared and asked that researchers continue to explore the
issue of when data uncertainty/variability cancel in relative
comparison applications.

Second Session Discussion Summary
  Workshop participants were asked to address the follow-
ing questions and to provide additional  questions to aid  dis-
cussion.

   1.  In human toxicity and ecotoxicity, when is spatial and/
      or temporal differentiation necessary? If necessary,
      what spatial and/or temporal details are recommended
      (e.g. indoor/outdoor, height of emission point)?

   2.  With respect to ecotoxicity what is the best approach
      to addressing multiple species? If suggested, what
      are recommended representative species?

   3.  With respect to human toxicity and ecotoxicity, what
      are the greatest barriers to conducting uncertainty
      analysis?

   4.  What are recommendations for research and devel-
      opment in these impact categories?

  An aggregation of the groups'views  is presented below:

  Standardization - Again the question of standardization
was discussed. Specifically, if the practitioner or study com-
missioner can have such a strong  influence on  the final
results of the study, then perhaps some standardization
would be useful to provide comparability between studies.
However, what perspective or aggregation of perspectives
should be represented in  a standardized approach? Should

-------
central tendency assumptions or worst-case assumptions
be used? Some participants stated that additional time was
needed to ferment an opinion in this area. Others contended
that "allowing" for too many methods and approaches could
undermine the credibility of LCIA. However, many believed
that now is the time to capture the state-of-the art in a docu-
ment, while still allowing room for advances in the future.
Several participants expressed interest in being involved in
the current SETAC-Europe Working Group on Life Cycle
Impact Assessment. (Udo de Haes, et al, 1999a and Udo
deHaes, etal,  1999b).

  Midpoint vs. Endpoint Level- In further discussion of the
concepts of midpoints vs. endpoints, many participants dis-
cussed the advantages of making all impact assessment
models as close as possible to the final endpoints of the
environmental  mechanism of the impact categories (e.g.,
quantifying fish kills and trees lost as opposed to the acidi-
fication potential of the substances). One benefit of this
approach would be to allow more common endpoints for the
valuation process, perhaps even opening the door to allow-
ing more economic valuation of endpoints. Others pointed
out that this might be unnecessary in a relative comparison
context. They stated that extending the models to the end-
points will narrow down the comprehensiveness of the im-
pacts considered, and will include many more assumptions
and value judgements into the assessment. This may sub-
sequently increase the uncertainty of the results and re-
duce credibility by further mixing "science and value judge-
ments."

  Ecotoxicity- There was a strong call for research in this
area.There was a recognized need to extrapolate ecotoxicity
in a manner similar to human toxicity with representative
species but also a realization that representative species
may vary within different areas. However, there was also
some discussion that LCA is a very macroscopic tool and,
can not be expected to accurately model local issues. Per-
haps, ecotoxicity is so specific to the locality affected that
an attempt should not even be made to  include it as an
impact category. The most widely held view on this topic
seemed to be that ecotoxicity should continue to be included,
for the sake of providing a more holistic picture, and that
the potential for more site-specific approaches should be
considered further.

  Potentially Affected Fraction of Species (PAFs) - Mark
Goedkoop gave an impromptu presentation on PAFs. He
stated that PAFs are different from PNECs in that they take
the background level of the substances into account and
thus enable nonlinear modeling of impact on the species
composition. Many in principle liked the idea  of PAFs and
combined PAFs that represent the combined effect of chemi-
cals. However,  there were concerns related to the possibil-
ity of identifying PAFs, due to the limited availability of dose
response curves and of background concentration data for
so many chemicals. A discussion of Eco-lndicator 98's re-
lationship to PAFs was held. (Goedkoop, 1998)

  Borrowing from Risk Assessment - Concern was voiced
that LCIA for human toxicity is often based on typical risk
assessment practice (e.g., the use of toxicological bench-
marks). Caution was particularly high in the context of de-
terministic safety factors used in the toxicity component of
the characterization factors, many of which compensate for
low test species numbers. As this reduces the equity and
comparability of chemicals, participants suggested that LCIA
must be careful when adopting deterministic risk assess-
ment perspectives.

  Research into Increasing Sophistication and the Role of
Other Assessment Techniques - One group asked for in-
creasing temporal modeling, real ground concentration mea-
surement, incorporation of population density into simula-
tions and better representation of food webs. In this group,
there was a concern that the current direction of research in
multimedia modeling would not address these areas. How-
ever this  must be viewed in the context of the aims which
are to be met by LCA as opposed to the types of analytical
tools. Thus, another group stressed that perhaps practitio-
ners are too concerned with detail. Perhaps the focus should
remain on macro differentiation of substances in terms of
their persistent, bioaccumulative, and toxic (PBT) proper-
ties. This could be subsequently complimented (if required)
by local scale analysis using other tools, and would help to
include a larger set of  chemicals  at a sufficient level of
differentiation.

Session Three: Acidification, Eutrophication
    and Inventory
  Greg Norris of Sylvatica,  North  Berwick, Maine, USA,
presented a "Value-of-lnformation Approach." [He pointed
out that uncertainty analysis allows some additional infor-
mation (e.g., confidence intervals associated with data un-
certainty) within the decision-making framework.] Norris
stated  that the level of  sophistication should be  partially
dependent upon the inventory data and its uncertainty, upon
the appropriate models  and upon decisions about weight-
ing. He suggested using Input/Output-based upstream LCI
databases to answer many of the common questions that
practitioners face, such as "How many sites, with how much
geographic dispersion, contribute significantly to inventory
totals?" And "What are the expected shapes of these distri-
butions?" He also cautioned participants against trying to
draw conclusions about the preferability of more  detailed
LCIA, based on a Probability Density Function (PDF) or
Cumulative Density Function (CDF) diagram, pointing out
that further simulations may be required. Finally, he dis-
cussed the difference between  analyzing uncertainty in
weighting and in characterization modeling and the need to
treat these issues jointly in the determination of the level of
sophistication and decision support.

  Jose Potting of the Technical University of Denmark pre-
sented "Levels of Sophistication in Life Cycle Impact As-
sessment of Acidification." Potting presented a case study
comparing alternative locations for copper production and
demonstrated the potential need for site-specific simula-
tions, including emission dispersion and deposition patterns,
background depositions on receiving ecosystems, and the
sensitivity of receiving ecosystems. [She used the Regional

-------
Air pollution information System  (RAINS) model (from
IIASA) with calculations based on Critical Loads provided
by the National Institute of Public Health and the Environ-
ment (RIVM) in the Netherlands and transfer-matrices from
EMEP MSC-Watthe Norwegian Meteorological Institute.]
She announced that easy-to-use acidification factors had
been established for 44 European regions and suggested
that utilizing this site dependent approach for acidification"
resulted in a significant reduction in uncertainty.

  G6ran Finnveden of Stockholm University presented two
topics - "Eutrophication—Aquatic and Terrestrial - State of
the Art," and "Thresholds/No Effect Levels/Critical Loads."
Finnveden discussed the site dependency of eutrophica-
tion in three models, developed since 1993. He presented
additional topics for discussion  and research related to
eutrophication. In his second presentation, Finnveden pro-
posed that thresholds may, at the macrolevel, have no sci-
entific basis and in fact may just be "acceptable" levels of
risk and thus constitute value choices. Acidification and
human toxicity were used as examples of impact catego-
ries that should not ignore "below threshold values." In line
with this, he proposed that threshold values should not ex-
ist in LCIA for any impact category.

  The third session was concluded with the large group
documenting some of the earlier topics and discussing the
value of conducting future similar workshops. An  on-site
workshop summary was presented by two of the co-chairs.

Conclusions
  In meetings and journals world wide, practitioners have
debated the utility of conducting Life Cycle Assessment
studies. The debate has often hinged on the appropriate
level of sophistication. While some have advocated aban-
doning LCA altogether, since it is not achievable in its most
sophisticated form, others have supported the concept of
conducting LCA studies at a more holistic level, while mak-
ing the limitations and uncertainties transparent. This work-
shop discussed many of the issues of dealing with the ap-
propriate level of sophistication  in the  Life Cycle  Impact
Assessment phase of an LCA study.

  A number of prominent practitioners and researchers pre-
sented a critical review of the associated factors, including
the current limitations of available impact methodologies
and a comparison of alternatives in the context of model
and data uncertainty. On the one hand the workshop ad-
dressed the various factors which are connected with an
increase  of sophistication in LCIA. Examples include: a)
the need for better fate and effect models, b) the role of
spatial and temporal differentiation, c) the identification of
background levels and thresholds, and d) the need to specify
value-laden aspects such as connected with different cul-
tural perspectives. On the other hand, the holistic and com-
parative character of LCA was stressed. In this context,
many questioned whether LCA should aim to conduct so-
phisticated site specific risk assessments, particularly when
this high level of detail may give a false impression of great
confidence, especially when it is not presented with a strin-
gent uncertainty analysis. Moreover, it was recognized that
thresholds reflect value choices about what is regarded
acceptable, rather than science based parameters. And fi-
nally, increasing level of detail can increase model certainty,
but, in some cases, may reduce the comprehensiveness.

  Workshop speakers and participants discussed the way
that philosophical views may affect not only the valuation
process, but also the impact assessment phase by includ-
ing assumptions that include values based on the differing
perspectives. This further complicates the question of what
is "science-based" and what are "reasonable" modeling as-
sumptions. Arguments were raised both for and against striv-
ing for consistency at this time in the effort to standardize
some of the methods and assumptions to allow compara-
bility between studies.

  There was much discussion about the decision-making
framework and the role of other environmental analyses,
such as risk assessment. From the sophisticated uncer-
tainty analyses presented it was obvious that great advances
are being made,  but there are many very basic principles
that still lack consensus (e.g., the use of threshold values
and background  concentrations). As in risk assessment,
there is great attention to being true to the science, but in
the interest of practicality, a great need for simplifying as-
sumptions.

  There was consensus that the workshop was very valu-
able and that this  exchange should be continued through e-
mail discussions and periodic workshops (next target work-
shop in Brighton, U.K. in May 2000). Several topics  were
mentioned for future workshops, including: LCIA at strate-
gic levels of decision making (including sustainable devel-
opment decision support), community planning using LCIA-
type indicators, the role of value choices  in  characteriza-
tion modeling, and the state-of-the-science for characteriz-
ing ecotoxicity in  LCIA.

References
Coulon, R., Camobreco, V., Teulon, H., Besnainou, J.,
   (1997): Data Quality and Uncertainty in LCI. Interna-
   tional Journal of LCA,  Vol. 2, No. 3, pp. 178 - 182.

Fava, J.,  Consoli, F., Denison, R., Dickson, K., Mohin,
   T, Vigon, B., (1993): A Conceptual Framework for
   Life-Cycle Impact Assessment. Society of Environ-
   mental Toxicology and Chemistry: Pensacola, FL.

Finnveden, G., (1997): Valuation Methods Within LCA-
   Where are the Values?. International Journal of  LCA,
   Vol.2, No. 3, pp. 163-169.

Goedkoop, M., Hofstetter,  P., Muller-Wenk, R.,
   Spriensma,  R., (1998): The  Eco-lndicator 98 Ex-
   plained, International Journal  of LCA, Vol. 3, pp. 352
   -260.

Grahl,  B., and Schmincke, E.,  (1996): Evaluation and
   Decision-making Processes in Life Cycle Assess-
   ment. International Journal of  LCA, Vol. 1, No. 1, pp.
   32-35.

-------
Hertwich E.G., Pease W.S., and McKone T.E., (1993):
    Evaluating Toxic Impact Assessment Methods: What
    Works Best? Environmental Science and Toxicology,
    32,138A-145A.

Hertwich, E. G. (1999) Toxic Equivalency: Accounting for
    Human Health in Life-Cycle Impact Assessment. Ph.D.
    thesis, Energy and Resources Group, University of Cali-
    fornia, Berkeley.

Hertwich, E. G., T. E. McKone and W. S. Pease (in press)
    "Parameter uncertainty and variability in evaluative fate
    and exposure models." Risk Analysis.

Hofstetter, Patrick, (1998): Perspectives in Life Cycle Im-
    pact Assessment: A Structure Approach to Combine
    Models of the Technosphere, Ecosphere and
    Valuesphere. Kluwer Academic Publishers, Norwell,
    Massachusetts.

International Organization for Standardization, 1999: ISO/
    DIS 14042: Environmental Management - Life Cycle
    Assessment - Life Cycle Impact Assessment.

Owens, J.W., (1998): Life Cycle Impact Assessment: The
    Use of Subjective Judgements in Classification and
    Characterization. International Journal of LCA, Vol. 3,
    No. 1, pp. 43 - 46.

Owens, J.W., Amey, E., Barnthouse, E., Consoli, F., Coulon,
    R., Fava, J.,  Humphreys,  K., Haaf, B.,  Hunt,  B.,
    Laibson,  L., Mones, E., Noesen, S., Norris, G.,Todd,
    J.A., Vigon, B.,Weitz, K., Young, J., (1997): Life-Cycle
    Impact Assessment: State of the Art. Society of Envi-
    ronmental Toxicology and Chemistry: Pensacola, FL.

 Potting, J., Hauschild, M., (1997): Spatial Differentiation in
    Life-Cycle Assessment via the Site-Dependent Char-
    acterization of Environmental Impact from Emission.
    International Journal of LCA, Vol. 2, No. 4, pp. 209 -
    216.

 Potting, J., Hauschild, M., Wenzel, H., (1999): "Less is Bet-
    ter" and "Only AboveThreshold":Two Incompatible Para-
    digms for Human Toxicity in Life Cycle Assessment.
     International Journal of LCA, Vol. 4, No. 1, pp. 16 - 24.

 Powell, J., Pearce, D.W., Craighill, A.L., (1996): Approaches
    to Valuation in LCA Impact Assessment. International
     Journal of LCA, Vol. 2, No. 1, pp. 11 -15.

 Udo de Haes, H., Jolliet, O., Finnveden, G., Hauschild, M.,
     Krewitt, W., and Muller-Wenk, R., (1999a): Best Available
     Practice Regarding Impact Categories and Category In-
     dicators in Life Cycle Impact Assessment- Part 1. Inter-
     national Journal of LCA, Vol. 4, No. 2, pp. 66 - 74.

  Udo de Haes, H., Jolliet, O., Finnveden, G., Hauschild, M.,
     Krewitt, W., and Muller-Wenk, R., (1999b): Best Available
     Practice Regarding Impact Categories and Category In-
     dicators in Life Cycle Impact Assessment - Part 2. In-
     ternational Journal of LCA, Vol. 4, No. 3, pp. 167 -174.
Udo de Haes, H., (ed.,) (1996):Towards a Methodology for
   Life Cycle Impact Assessment, SETAC-Europe Work-
   ing Group on Impact Assessment, Brussels.

UNEP Industry and Environment (1996): Life Cycle Assess-
   ment: What is it and How to Do It. United Nations Pub-
   lication Sales no. 9C-III-D.2, Paris, France.

UNEP  Industry and Environment (1998): Draft Workshop
   Summary -Towards Global Use of LCA, Paris, France.

Volkwein, S., Gihr, R., Klopffer, W., (1996a): The Valuation
    Step Within LCA. Part I: General Principles. Interna-
   tional Journal of LCA, Vol. 1, No. 1, pp. 36 - 39.

Volkwein, S., Gihr, R., Klopffer,  W., (1996b):The Valuation
    Step Within LCA. Part II: A Formalized Method  of
    Prioritization by Expert Panels. International Journal of
    LCA, Vol. 1, No. 4, pp. 182 - 192.

Wenzel, H., (1998): Application Dependency of LCA Meth-
    odology: Key Variables and Their Mode of Influencing
    the Method. International Journal of LCA, Vol. 3, No. 5,
    pp. 281-287.

Workshop Agenda
Sunday, Nov. 29

15:00 -15:45     Introduction of Workshop by Co-Chairs

 15:45 - 21:30     Determination of Sophistication

 15-45 -16-15     ISO Standards: LCA & Comparative As-
                 sertions (Willie Owens, P&G, USA)

 16:15 -16:45     Impact Assessment for Ecodesign (Mark
                 Goedkoop, Pre, Netherlands)

 16:45 -17:15     Application Dependency of LCI A (Henrik
                 Wenzel, Tech U of Denmark)

 17:15-19:00     Dinner

 19:00 - 20:30     Small Group Discussion of Determina-
                 tion of Sophistication

 Questions Provided

 20:30 - 20:45    Coffee Break

 20:45 - 21:30    Small Group Presentations

 Monday, Nov. 30

 08:00 -17:30    Specific Impact Categories & Uncer-
                 tainty Analysis within LC1A

 08:00 -12:00     Human Toxicity & Ecotoxicity

 08:00 - 08:30     Types of Uncertainty in the Human Tox-
                  icity Potential
                  (Edgar Hertwich, UC Berkeley, USA)

-------
 08:30-09:00
 09:00-09:30
 09:30-10:00
 10:00-11:15
The Different Levels of Uncertainty As-
sessment in LCIA;The Case of Carcino-
genic Effects (Patrick Hofstetter, Swiss
Federal Institute of Technology)

Human Toxicity and Ecotoxicity - Model-
ing vs. Scoring (Olivier Jolliet, EPFL,
Switzerland)

Priority Assessment of Toxic Substances
in LCA: A Probabilistic Approach (Mark
Huijbregts, IVAM, Netherlands)

Small Group Discussion of Human Health
and Ecotoxicity
 Questions Provided

 11:15-12:00     Small Group Presentations

 12:00-13:30     Lunch

 13:30 -16:30     Acidification, Eutrophication, andTro-
                 pospheric Ozone

 13:30-14:00     Implications of Inventory Structure for
                 LCIA and Uncertainty Analysis (Greg
                 Norris, Sylvatica)

 14:00 -14:30     Levels of Sophistication in LCIA of Acidi-
                 fication  (Jose Potting, Tech U of Den-
                 mark)

14:30 -15:00    Eutrophication - Aquatic and Terrestrial-
                State of the Art (Goran Finnveden, fms
                and Stockholm University)

15:00-15:15    Coffee Break
 15:15-16:45     Large Group Discussion

 16:45-17:30     Workshop Revelations and Lessons
                 Learned

 Closing Remarks by Co-Chairs

 Bare's Biography

   Bare is the Impact Assessment and Measurement (I AM)
 Team Leader within the Sustainable Technology Division of
 the National Risk Management Research Laboratory. The
 role of the I AM Team is to influence environmental deci-
 sion making in Sustainable Development, Life Cycle Im-
 pact Assessment, and Pollution Prevention through the re-
 search, development, and application of comprehensive
 impact assessment and progress measurement. Bare has
 been active in ISO 14042 (Life Cycle Impact Assessment)
 providing written and verbal comments to the U.S. position
 and representing the U.S. position on several issues. She is
 also the U.S. EPA representative and an active member of
 the American Institute of Chemical Engineers (AICHE) Cen-
 terfor Waste Reduction Technologies (CWRT) Sustainabil-
 ity Metrics Working Group which is designing a framework
 for Sustainable Development decision making within large
 U.S. chemical  companies,  including Monsanto, DuPont,
 Rohm and Haas, Eastman Kodak, and others. She has given
 numerous U.S. and International presentations related to
 Life Cycle Impact Assessment and Sustainable Develop-
 ment. Bare and other members of the I AM team conduct
 research into improved impact assessment methodology
 development and impact assessment application to sup-
 port environmental decisions. One tool - TRACI - Tool for
the Reduction and Assessment of Chemical and other en-
vironmental Impacts will provide improved impact assess-
ment methodologies for five of seven major chemical im-
pact categories. Bare is a chemical engineer with 15 years
of experience within the U.S. EPA.
                                                   10

-------
Bare's Overheads
      Life Cycle Impact Assessment
            Sophistication
                   UNEP
                                CML
                                Ceadeof Eavironmentil Science
                 Organized by:
           Jane C. Bare, U.S. EPA ORD
            Helias A. Udo de Haes, CML
      David W. Pennington, U.S. EPA ORISE Fellow


           in co-operation with UNEP
  Life Cycle Impact Assessment
       can be effective in
 environmental decision making,
              but

      only If the      and
                 support a
                 defensible
                  point
         Scientifically Defensible
              Decision Point


               Dependent Upon:

             Level of Sophistication
          Level of Uncertainty Analysis
            Level of Comprehension
                Data Availability
                 Data Quality
     How is LCIA Sophistication
            Determined?

To Be Addressed

• project objective or application
• state-of-art for the impact category

• uncertainty and/or sensitivity analysis
• quality and availability of modeling data
          How is LCIA Sophistication
                 Determined?

     Recognized but Not Addressed

     •  perceived value of the impact category
     •  anticipated level of impact
     •  inventory data available ,
     •  supporting software available
     •  awareness of available models
     •  the level of funding
     Impact Category Trends

   • Simulations are becoming more
   sophisticated as technology advances.

   • Researchers are increasingly aware of
   developments in other countries and are
   testing and comparing methodologies.

   • More discussions are addressing the
   limitations of LCIA and the difference
   between scientific and non-scientific
   model simplifications.
                                          11

-------
                   Your Challenge

    To discuss:

    current LCIA sophistication, including
    temporal and spatial resolution,

    major findings to date concerning the
    scientific validity of current LCIAs, and

    the research  needs in this area.
                                                    Sunday Afternoon,  Nov. 29
                                             15:00-15:45
                                             15:45-21:30
                                             15:45-16:15


                                             16:15-16:45
                                             16:45-17:15


                                             17:15-17:30
                                             17:30-19:00
                                             19:00-20:30


                                            20:30 - 20:45
                                            20:45-21:30
                   Introduction of Workshop by UNEP & Co-Chairs
                   Determination of Sophistication
                   Industrial Use: Comparative Assertions
                   (Willie Owens, P&G)
                   Internal Use Only (Mark Goedkoop, Pre Consultants)
                   Application Dependency of LCIA -
                   (Henrik Wenzel, Tech U of Denmark)
                   Summary Remarks by Co-Chair
                   Dinner
                   Small Group - Determination of Sophistication
                   Questions Provided
                   Coffee Break
                   Small Group Presentations
          Monday Morning,  Nov. 30
 08:00-17:30

 08:00-12:00
 08:00-08:30

 08:30-09:00
 09:00-09:30

 09:30-10:00

 10:00-11:15

 11:15-12:00
 12:00-13:30
Specific Impact Categories & Uncertainty
Analysis within LCIA
Human Toxicity & Ecotoxicity
Discussion of Uncertainty Types & Human Health
Example (Edgar Hertwich, UC Berkeley, USA)
Structured Approach with Quantitative Uncertainty
Assessment (Patrick Hofstetter, Swiss Fed. Inst. of
Tech.)
Human Toxicity and Ecotoxicity Modelling vs. Scoring
(Olivier Jolliet, EPFL, Switzerland)
Uncertainty in Toxicity (Mark Huijbregts, IVAM,
Netherlands)
Small Group - Human Health and Ecotoxicity
Questions Provided
Small Group Presentations
Lunch
                                                   Monday Afternoon,  Nov. 30
13:30-16:30

13:30-14:00


14:00-14:30


4:30 -15:00


5:00-15:15
5:15-16:00


6:00-16:45
6:45-17:30
Acidification, Eutrophication, and Tropo. Ozone
Smog Formation Characterization Analysis (Greg
Norn's, Sylvatica)
Levels of Soph. Of Acidification (Jose Potting, Tech U
of Denmark)
Eutrophication (Goran Finnveden, fms and Stockholm
University)
Coffee Break
Small Group - Three Impact Categories
Questions Provided
Small Group Presentations
Workshop Revelations and Lessons Learned
Closing Remarks by Co-Chairs
       Determination of Sophistication
 Methods for
 1. What are the most common methods by which the level of sophistication is
 determined?
 Which methods are considered more acceptable? Why?
 2. What are trie barriers to using the acceptable methods? What can be done
 to overcome these barriers?

 Various Uses o( LCIA
                                                 Human Toxicity and Ecotoxicity
                                            ncertalntv Analysis
3. Is LCIA application dependent? Why or why not?
4, What are the expectations in sophistication for the various LCA uses (e.g.,
government, Industry/public communication, internal use)?
5. Whan should LCIAs be as detailed as possible, because every increase of
detail increases accuracy and removes uncertainty? And when is it better to
limit the scope of LCA to addressing questions on a macroscopic scale, leaving
spatial and threshold consideration to other analytical tools?
6. How do practitioners determine the value of the trade-offs necessary when
sophistication and comprehensiveness are "at odds" (e.g., choosing a modeling
approach may limit the number of pollutants that can be characterized vs. a
scoring approach which may limit the sophistication of the modeling)?
                                             What case studies are available using uncertainty analysis within LCIA? And
                                            hat are the major findings to date (levels of uncertainty discovered)? When is
                                            e uncertainty determined to be unacceptable?

                                            uman Toxicitv and Ecotoxieity
                                             In human toxicity and ecotoxicity, when is spatial and/or temporal
                                            fferentiation necessary? If necessary, what spatial and/or temporal details
                                            e recommended (e.g., indoor/outdoor, height of emission point)?
                                             In ecotoxicity what is the best approach to addressing multiple species? If
                                            ggested, what are the recommended representative species?
                                             In human toxicity and ecotoxicity, what are the greatest barriers to
                                            nducting uncertainty analysis?
                                             What are the recommendations for research and/or development in these
                                            pact categories?
                                                             12

-------
        Eutrophication, Acidification, and
         Trope-spheric Ozone Formation

Uncertainty Analysis
1.  What case studies are available using uncertainty analysis within LCIA? And
what are the major findings to date (levels of uncertainty discovered)?  When is
the uncertainty determined to be unacceptable?
Eutrophication. Acidification, and Trooospheric Ozone Formation
2.  In each category, when is spatial and temporal differentiation necessary?
so, what spatial and temporal scales are recommended (e.g., regionally, or
differentiating terrestrial, sweet water and marine)?
3.  In each category, what are the greatest barriers to conducting uncertainty
analysis?
4.  What are the recommendations for research and/or development in these
impact categories?
                                                                 13

-------
Pennington's Biography
  Pennington began his career at the University of Surrey in
England, where he obtained an Honors Degree in Chemical
Engineering in 1990. Aftertwo years with the consultant com-
pany Advanced Mechanics and  Engineering (AME),
Pennington accepted a scholarship at the Hong Kong Univer-
sity of Science & Technology to perform research in the De-
partment of Chemical Engineering under the supervision of
department head, Professor P.L. Yue. Completing development
of a software tool and the associated Ph.D. thesis "A Pollution
Prevention Tool for Continuous Chemical Processes (P2TCP)"
in 1997, he then accepted a Post Doctoral Research Fellow-
ship at the Systems Analysis Branch of the US EPA National
Risk Management Research Laboratory (NRMRL) in Cincin-
nati, Ohio. There, Pennington has continued his research and
development activities in the areas of environmental impact
comparison and screening. A key interest in his R&D activi-
ties has been the simplification and use of multimedia mod-
els in the application of predicting persistence, relative im-
pact potentials and long-range transport capabilities of chemi-
cals. He is currently a participant in three key activities: the
US EPA Persistent Bioaccumulative and Toxic Pollutants Ini-
tiative (PBTI) working group; Jane Bare's development of Pol-
lution Prevention Progress (P2P) and the Tool for Reduction
and Assessment of Chemical and other environmental Im-
pacts (TRACI); and the SETAC Europe LCA working groups
on chemical fate and exposure, human toxicological impacts
and ecological toxicological impacts.
                                                   .14

-------
Pennington's Overheads
       LOiA Sophistication
           Issues Overview

               Dr. David W. Pennington
                ORISE Research Fellow
            US EPA Systems Analysis Branch
                       Cincinnati
                                   TwoExtrensliutwhatisappriale?
                                   "Contributions or Burden" Approach
                                   • Typical LCIA Approach
                                   • Not actual impacts associated with specific emissions
                                   • More than one chemical and one emission
                                   "Consecutive Risk Assessments" Approach
                                   • Comparative assertions (?) and LCSEA
                                   • Actual impacts associated with a specific emissions
                                   • Thresholds ?
                                                                                         ^ฃ<2aK&*t
                Subjectivity!

       What level of subjectivity acceptable ?
       Need to distinguish science vs. non-science based?
       •Disability Adjusted life Years
       • Average human inhalation rate in models
                                    Sophistication of Model

                                   Screening: PBT Scoring Approaches
                                   • Scientifically combined with release quantity ?
                                   • Qualitative ranking (high, medium, low) ?
                                   • More chemicals ?
                                   Continuous scales: Potentials
                                   • Potentials modify release quantities
                                   • Ability to be comprehensive (state-of-the-art) ?
                                   • All pathways, species, interaction of chemicals...? /j
           Spatial  Differentiation
          Global
        •Human & Eco?
 Regional/Remote

•Human & Eco?
•What boundaries?
•Advection rates?
•What Species?
•Other Fate Parameters ?
•Which Models?
•Acute Impacts?
   Local

•Human & Eco?
•What is local?
•InLCA?
•Acute Impacts?
•Which Models?
   Midpoints vs Endpoints

Midpoints: Acidification Potential - # of H+ ions
• Simplicity & reflective of "contributions" LCIA approach
• Lower uncertainty ?
ซ How to measure it?
• Lack of spatial and temporal variation-Problem? How?
Endpoints: Human Health Potentials - fate, exposure and effects
• Complex but not time intensive?
• Uncertainty and variation quantifiable ?
• High?
• Can consider spatial and temporal variation-Do we need to?
                                                   15

-------
         ISO Standards:  Life Cycle  Assessment &  Comparative Assertions
                                               J.Willie Owens,
                                              Procter & Gamble
                                             Cincinnati, OH, USA
 Summary

 ISO Standards: Life Cycle Assessment &
           Comparative Assertions
   IS014040 defines a comparative assertion as an "envi-
 ronmental claim regarding the superiority or equivalence of
 one product versus a competing product which performs
 the same function." The IS014020 series establishes sev-
 eral principles for any environmental claim, including com-
 parative assertions: information will be accurate, verifiable,
 relevant, and non-deceptive; scientific methods will used to
 generate the results; the process is open and participatory;
 the information is transparent and available to all (e.g., pur-
 chasers, interested parties, etc.); any claim is based on
 measurable differences including consideration of variations
 and uncertainty; and clear explanatory statements justify
 and qualify the claim. The 14040 LCA standards and draft
 standards meet these principles.

   JS014040 and 14041 provide a detailed framework cov-
 ering the LCA study goal definition and scope; functional
 unit and system boundaries; data inclusion and exclusion;
 data quality; data collection, handling, and calculations; al-
 locations; and reporting.There are also specific requirements
 when the comparative assertion is disclosed to the public:
 data quality parameters that consider time-related,  geo-
 graphical, and technological variations; data precision, com-
 pleteness and representativeness for different processes;
 consistency and reproducibility of the methods;  data
 sources; and uncertainty. Public comparative assertions also
 require equivalence between compared systems for func-
 tional unit, methodological considerations, performance, sys-
 tem boundaries, data quality, allocation procedures, deci-
 sion rules on inputs and outputs, and the impact assess-
 ment phase. Together, these requirements establish a fair
 comparison in the inventory phase that is technically sound
 transparent, and non-deceptive.

  ISO 14040 requires that public comparative assertions
 be based on a full LCA, including an impact assessment
 phase. The mandatory portions of impact assessment in
 FDIS14042 are Figure 1 .The concept of an LCIA category
 indicator and its relationship to the impact category and the
characterization model is shown in Figure 2.
   ISO 14042 requires categories, indicators, and models
 used be: consistent with study goal and scope, sources for
 models, etc., be referenced; selections be transparently jus-
 tified and explained; a comprehensive set of environmental
 issues within the goal and scope be covered; and describe
 environmental mechanism and model so that the LCI  re-
 sults are related to the indicator, and the appropriateness of
 model is understandable. A scientific approach  in the im-
 pact category model is taken by basing the model on a
 defined environmental mechanism, i.e., a system of physi-
 cal, chemical, and biological processes linking category LCI
 results to category indicators and to category endpoints. A
 highly simplified example for an acidification environmental
 mechanism is shown in Figure 3.

   ISO 14042 requires for public comparative assertions: a
 sufficiently comprehensive set of indicators, a comparison
 conducted indicator by indicator, that LCIAs not be sole
 basis for comparative assertions as additional information
 may be necessary to overcome  limitations in clause 8 of
 14042), that subjective scores such as weighting shall not
 be used; that category indicators used shall be both 1) sci-
 entifically and technically valid, and 2) environmentally rel-
 evant, and sensitivity and uncertainty analyses of results
 shall be conducted.

  Thus,  each of the mandatory steps used to derive an
 indicator must be scientifically and technically valid: group-
 ing into impact categories (Classification), converting LCI
 results (1st Characterization step), and aggregating con-
 verted LCI results (2nd step). This may present difficulties
 for several current practices such as aggregation of differ-
 ent effects, aggregation  of a similar  effect from  different
 places and times, and the use of subjective scores. For
 example, an ILSI expert  panel to toxicologists states that
 "It is inherently impossible to make a purely scientific com-
 parison  of qualitatively or quantitatively different toxicity
 impacts. This type of aggregation is equivalent to  compar-
 ing the impacts of global warming and  acid rain. Such com-
 parisons cannot be done scientifically."

  For environmental relevance, IS014042 establishes key
criteria to meet: the indicator will reflect the actual conse-
                                                     16

-------
quences of the  system  operation on  the category
endpoint(s), at least qualitatively. In addition, the category
model must incorporate environmental data or information,
including:  environmental condition  of  the category
endpoint(s); intensity of environmental changes; spatial and
temporal aspects such as duration, residence time, persis-
tence, timing, etc.; reversibility; and uncertainty.

  ISO also established a critical review of any study used
for public comparative assertions so that methods used will
be consistent with ISO, all methods are scientifically and
technically valid, the data are appropriate and reasonable,
study interpretations reflect the limitations, and that the study
report is transparent and consistent. The panel members
should be familiar with ISO and have scientific/technical
expertise to address the impact categories covered.There
are also extensive reporting requirements for the conduct
of a study to ensure transparency, the critical review panel
report must included in study report, and the report must be
made available to all upon request.

  As a result, the ISO 14040 series clearly complies with
the 14020 series principles for environmental claims made
in the opening paragraph.

Thought Piece: Comments on  Science
  First, science is the exploration of the physical world and
the body of knowledge that results from that exploration.
This knowledge is gained  via a number of processes in-
cluding  basic research, applied  research, and applications
such as environmental monitoring.

  Second, this body of knowledge is both gained and as-
sessed  by  applying the scientific method. The tenets or
characteristics of  the scientific method are:

   • The method begins with defined and framed observa-
    tions or questions.

   • A testable hypothesis is generated to explain the spe-
    cific observation or address the specific question. The
    hypothesis is then used to design physical experiments.
    This includes clear descriptions of assumptions to be
    used and variables to be tested.

   • The experiments include positive and negative controls.
    In many cases, statistical tests of the  results and the
    variables are also included.

   • The experimental observations are based on results that
    are physically measurable and obtained using validated
    measurement techniques.

   • The hypothesis, methods and results are transparently
    described so  that

    1. Both the hypothesis and the experiment can be un-
       derstood, reproduced and the results verified.

    2. The hypothesis can be tested in other experiments
       with different designs and measures.
  • An interpretation consistent with the experiment is of-
    fered, including care to distinguish the statistical sup-
    port, to note alternative interpretations where they ex-
    ist, and to carefully qualify additional hypotheses or
    speculations that are not directly supported by the re-
    sults.

  • Independent researchers can then explore the hypoth-
    esis and findings where new experiments may support
    or challenge the results and interpretation.The ultimate
    outcome is acceptance or rejection of the hypothesis.

  Third, the last point has major implications. Science rarely
offers absolute proof, and cannot prove a negative or null
hypothesis. The body of knowledge generated by the scien-
tific method is subject to assessment with new experimen-
tal evidence, and even periods of controversy when mat-
ters are disputed. In fact, certain findings can be highly
contentious when established models and thinking are chal-
lenged by new experiments.The acceptance of these new
experiments then awaits testing, verification, and under-
standing. For example, a strong debate continues among
evolutionary scientists between continuous evolutionary
change and another model termed punctuated equilibrium.
In another case, proposals for a synergistic effect among
so-called endocrine disrupters could not be reproduced ei-
ther in other laboratories or the original laboratory, so the
authors retracted their findings. Thus, science and its body
of knowledge is not dogma. Science operates by construct-
ing and assessing the weight of evidence.

  Fourth, scientific weight of the evidence is often a gradual
accumulation of data  and knowledge. For  example, the
empirical observation of PCB body residues in Baltic fauna
in the mid-1960s led to the reproduction of this observation
in other areas and species as well as with other compounds
such as DDT. As a result, we  now have detailed body of
knowledge about:

  - the physical and biological properties underlying the fate
    and transport of substances,

  - the degradation and persistence of substances,

  - their many mechanisms of toxicity,

  - the processes of bioaccumulation and biomagnification,
    and

  - the means to measure and model these processes as
    well as how to validate the models.

Owens' Biography
  Owens is a Principal Scientist at Procter & Gamble and
has responsibility for LCA development. He has a B.S. in
Biology (summa cum  laude, Christian Brothers College,
Memphis, TN) and Ph.D. in Molecular Biology (US National
Science Foundation Fellow, Washington University, St. Louis,
MO). His prior experience contributing to his LCA background
includes product development, manufacturing, and consumer
testing within P&G, such as upstream product development
                                                      17

-------
for both disposable medical products and industrial cellu-
lose products, pilot plant experience in modified cellulose
fiber production, in-plant process development experience
at pulp mills, pilot scale production and consumer testing of
consumer paper products, service as a human toxicologist
for pulp and paper products, and environmental assessment
studies at pulp and paper mills. The largest environmental
study involved the design and the execution of the
multidisciplinary and multi-year effort in Alberta, Canada.
He has participated in the conduct of several life-cycle stud-
ies within P&G. He is a member of SETAC's LCA Advisory
Board, chair of the SETAC LCA Work Group on Impact
Assessment (North America), and is a member of the U.S.
delegation chair for the IS014000 TC 207 LCIA Work Group.
                                                    18

-------
Owen's Overheads
                ISO Standards:
            Life Cycle Assessment
            Comparative Assertions
                    J. William Owens
                    Procter & Gamble
                  Cincinnati, OH, USA


                  UNEP LCIA Workshop
                    Nov. 29-30, 1998
                       Brussels
             Definition of

      Comparative Assertions

             74040 - Definition 3.2


  environmental claim
   regarding the superiority or equivalence
    of one product versus a competing product
      which performs the same function
            ISO 14000 Standards

                Normative Terminology

        shall - requirement:
            critical review ฅ/ou!d note whether or not
            study was in compliance

        should- recommendation;
            critical review would note whether or not
            recommendation was followed

        may- suggests.conslderatton
          IS014020 Series

             Declarations & Labels

Necessary elements for claims

  ป Accurate, verifiable, relevant, and non-deceptive
  ป Scientific method used where results are
     are accurate and reproducible
  • Open, participatory process
  * Information available to purchasers, interested
     parties, etc. (transparent)
                IS014020 Series

                  Declarations & Claims

          Other elements:

            • Based on measurable difference;
               analysing data ranges and variability
            • Explanatory statements where needed
            • Relevant to the geographic area

          •/ Life cycle assessment standards are
            consistent with these requirements
         IS014040 and 14041

                Coverage

             Brief Overview of Areas
           definition and scope
       Functional unit and system boundaries
       Data inclusion and exclusion for inventory
       Data quality
           collection, handling, and calculations
       Energy data handling and reporting
       Allocations
       Reporting
                                                         Accurate, technically sound, verifiable, & transparent
                                                   19

-------
       Goal  Definition & Scope

    Clearly define application & audience as well as:

  - the functions of system or systems
  • the functional unit;
  - the product system to be studied
  - the product system boundaries;
  • allocation procedures;
  - categories, methodology, and interpretation of impact
    assessment  (also see Appendix A of 14042);
  - data requirements;
  - assumptions;
  • limitations;
  • Initial data quality requirements
  • type of critical review, if any;
  • type and format of study report.
            Data Quality

          Comparative assertions

All the data quality parameters must be addressed
and included in the study

- time-related
- geographical
- technology
- precision, completeness and representativeness
  -of data
- consistency and reprodycibiilty of the methods
- data sources and their representativeness
- uncertainty
   Equivalence of Comparisons

          Comparative Assertions -14040

Equivalence  necessary to make valid comparisons:
- same functional unit;
- equivalent methodological considerations:
   performance,
   system boundaries,
   data quality,
   allocation  procedures,
   decision rules on inputs and outputs and
   Impact assessment.
v'  Non-deceptive & based on measurable difference
 Life Cycle Impact Assessment
4 phases of LCA                 LCI     LCA
   - goal and scope definition     +        *
   - inventory analysis           +        +
   - impact assessment                   *
   - interpretation                +•        4-
LCIA phase necessary for comparative assertions
               Discussion for
        Mandatory Elements Only
       Category Indicator
         Selection of impact categories,
         category indictors and models

    Assignment of LCI results [Classification]  1
                                  Example

                                     HCS, etc.


                                 Aefdiffctttion
Acidifying Emissions
        ate.)
     Calculation of category indicator results
              [Characterisation]
      Category indicator results (LCIA profile)
                                                          Life cycle Inventory results
  LCI results assigned to
    mpact category
                               Proton S0fe.sss (H>)
                                                 20

-------
       Goal Definition  & Scope

   Clearly define application & audience as well as:

  -the functions of system or systems
  - the functional unit;
  - the product system to be studied
  - the product system boundaries;
  - allocation procedures;
  - categories, methodology, and interpretation of impact
    assessment (also see Appendix A of 14042);
  - data requirements;
  - assumptions;
  - limitations;
  - initial data quality requirements
  • type of erif icai review, if any ;
  - type and format of study report.
            Data Quality

          Comparative assertions

All the data quality parameters must be addressed
and included in the study

- time-related
- geographical
- technology
- precision, completeness and representativeness
- consistency     reproducibility of the methods
•     sources and their representativeness
- uncertainty
   Equivalence of Comparisons

         Comparative Assertions -14040

Equivalence necessary to make" valid comparisons:
 same functional unit:
 equivalent methodological considerations:
   performance,
   system boundaries,
   data quality,
   allocation procedures,
   decision rutes on inputs and outputs and
   impact assessment.
/  Non-deceptive & based on measurable difference
 Life Cycle Impact Assessment
4 phases of LCA
   - goal and scope definition
   - inventory analysis
   - impact assessment
   - interpretation
                             LCI
                             4-
                                      LCA
                                        +
                                        +•
 LCIA phase necessary for comparative assertions
              Discussion for
        Mandatory Elements Only

         Selection of impact categories,
         category indictors and models
        Category Indicator
    Assignment of LCI results [Classification]
                                  Example

                                kg SO,, HCl, etc.


                                  Acidiftcallsn
                               Acidifying Emissions
                                 (NO, $0ป etc.)
     Calculation of category indicator results
  v 	——
              [Characterisation]
      Category Indicator results (LCIA profile)
Life cycle inventory results
                                                          tCI results assigned to
                                                            impact category
                                Proton Rstesse (tf)
                                                 21

-------
Algebraic Formula of Calculations

   The conversion of the inventory result is
   a multiplication step - in its simplest form:
          CIR = CF MR
   whore
    CIR - converted inventory result
    CF - characterisation factor
    IR - inventory result

   or algebraically:

            y = a*x
       For LCI Result Conversion

            When seen as Clfl = CF * IR:
                    y = a*x
    Converted
    Inventory
     Result
Zero intercept
 No threshold
slope (a) is Characterisation Factor

          vป
               Linear response
                                                                      Inventory Result
                                                                      (IR) is x-axis
            AGGREGATION

  Is Aggregation of Converted LCI results Valid?


                    E?

 Do the LCI results truly act and combine together?
    Are the LCI results interchangeable across
          spatial and temporal scales?
                     or
Are the LCI results independent in the environment
          where action in one place or
       time has no connection to another?
            HUMAN TOXICITY

                  Current Issues

   issue           Toxicologists         LCA

 Thresholds          Exist            Absent
 Dose-Response    Case-Specific       Linear
 Endpoint Additivity Common Mode /    Universal
                  Common Organ
 Exposure*      Case & Circumstance  Simultaneous
                    Specific       & Aggregated

 * Assumes common place and time in aggregation
          HUMAN TOXICITY

           I LSI Panel on Aggregation

   "It is inherently impossible to make & purely
   scientific comparison of qualitatively or
   quantitatively different toxicity impacts.
   This type of aggregation is equivalent to
   comparing the impacts of global warming
   and acid rain. Such a comparisons cannot
   be done scientifically."


 Burke, Doull, McKone, Paustenbach, Scheuplein,
           Udo de Haes, and Young
 Criteria for Environmental Relevance

  indicator reflects the consequences on the category
    ertdpoint(s), at     qualitatively.

  Incorporate environmental data or information in the
     category model, including:
  - environmental condition of the category endpoint(s)
  - intensity of the-projected impacts,
  - spatial extent of projected impacts,
  - temporal aspects, duration, residence time,
   persistence, timing, etc,, of projected impacts,
  - reversibility of projected impacts, and
  - uncertainty of projected Impacts
                                               22

-------
             Critical  Review

            Summary of requirements

- methods consistent with ISO
- methods are scientifically and technically valid
- data are appropriate and reasonable
- interpretations reflect the limitations
- report is  transparent and consistent

Panel requirements - 14040 & 14042
- famiiiar with ISO and scientific/technical expertise
- may include other parties effected by study
- pane! report included In study report
- expertise for impact categories covered
             Conclusions

          The 14040 LCA standards

Procedures technics! & scientific valid so that
  analysis is credible to all parties

Results will be environmentally relevant

Procedures, assumptions, and decisions transparent

Independent review of the study and its conclusions

Flexibility for different applications and for
 current and future methods development

Open and equal treatment to industry, NGOs,
  government, consultants, universities, etc.
                                                   23

-------
                             Impact  Assessment  for  Ecodesign
                             Mark Goedkoop, PRe Consultants,The Netherlands,
                                         Email: Goedkoop@pre.nl
Introduction
  The problem with LCA and ecodesign can be summa-
rized as:

  • LCAs are too time consuming in product development
    processes.

  • The results of LCAs are hard to interpret for a designer;
    often ten different scores for different environmental
    impact categories are presented as "the" answer.

  Our solution: Develop a set of predefined scores for the
most commonly used materials and processes. The set
should include indicators for common materials, production
processes, transport, energy consumption and different dis-
posal options per material. With this set anyone can per-
form its own LCA in a matter of minutes on the back of an
envelope.

  A crucial factor in  this solution is the reliability of the
indicators. The first effort to develop this set of indicators
resulted in the Eco-indicator 95 LCIA methodology, which
now in use by many companies. Currently a follow-up project
is undertaken for the Dutch government resulting in the Eco-
indicator 99 methodology (Goedkoop 98).

Calculating standard indicators
  Standard indicators are calculated with the LCA method-
ology. This means we have to deal with the three different
aspects of any LCA: inventory, impact assessment and
valuation issues. In more abstract terms we have described
these three different issues as "spheres":

  • Technosphere, describing the technical system, or the
    inventory phase.

  • Ecosphere, describing the environmental cause and ef-
    fect mechanisms, resulting in a number of damage cat-
    egories or endpoints.

  • Valuesphere, describing the societal values in assess-
    ing the seriousness of the damages, and describing
    some of the value choices that are unavoidable in the
    modeling of cause and effect mechanisms.
  Each sphere has its own typical characteristics and prob-
lems.
            Technosphere
                  Ecosphere
Valuesphere
Subject
of the
modeling:

Verification
Main
problems:
values
Uncertainty
Concrete technical Models complex  Societal prefer-
systems        cause and effect ences and
               chains         values
Possible in many
cases

Boundaries
Allocations
Difficult or
impossible
Limited under-
standing of
mechanisms
and data
problems


A single truth
does not exist
How do we
measure
in society?
How do we
deal with
incompatible
views
low (far less then  High (sometimes High, depending
an order of magni- several orders of on procedure
tude)           magnitude      for information
                             gathering
  The problems in technosphere are relatively well understood
in LCA. The problems in ecosphere can in principle be solved
as our scientific understanding progresses, although consid-
erable research efforts will still be needed. The problems in
value sphere are fundamental, as they deal with incompatible
views of different people. These problems cannot really be
solved, but they can be managed if they are well understood.
The nature of the Valuesphere problems can be illustrated
with two examples.

  Example 1: What level of scientific proof is required?

  For only a few substances the carcinogenicity is sufficiently
proven. Many substances are generally suspected to be car-
cinogenic, but scientific proof is incomplete. Some people will
maintain that all substances under suspicion must be included,
while others would demand sufficient scientific proof if a sub-
stance were to be included. Who will decide?

  Example 2: What timeframe do we adopt?
                                                      24

-------
  In fate models an important parameter is the time per-
spective. If we take a long time perspective, substances
like heavy metals become very dominant in the analysis. In
fact the choice of the time perspective determines the char-
acterization values to a great extent. Some people will ar-
gue an indefinite time perspective must be used, while oth-
ers would argue it is nonsense to look at damages that
would occur after 100 years. Who will choose the time per-
spective, and thus decide on the relative importance of heavy
metals?

  In a system with standard indicators such questions  be-
come very important, as, unlike in a full LCA, the stake-
holders can not make  such decisions on a case by case
basis. In the Eco-indicator 99 methodology the solution is
to choose three different perspectives.This means we pro-
duce three different versions of this methodology, each with
its own consistent set of value choices. (See also [Hofstetter
1998] and [Thompson 1990] on Cultural Theory).

   • Egalitarian: Precautionary principle: include as much
    as possible, use a long time perspective. (Typical atti-
    tude of environmental groups)

   • Hierarchical: Include facts that are backed up by sci-
    entific and political bodies. (Typical attitude of scien-
    tists).

   • Individualist: Only proven cause effect relations, short
    time perspective; future damages  are far less impor-
    tant then present damages, (typical attitude of entre-
    preneurs)

   The users of the indicator system cannot make value
 choices on a case  by case basis, but they can choose
 between three sets of value choices. This underlines the
 obvious fact that a single truth is not available in LCA, even
 when indicators are used.

 Conclusions
   • It is much more efficient to calculate a large set of reli-
     able single scores for commonly used materials and
     processes. With these any designer can make their own
     LCA without the help of other experts.

   • The issue of value choices becomes very important in
     a system with predefined indicators, as the user  can-
     not influence them, while the results they get partially
     depends on such choices.

   • By separating three different perspectives we give the
     user the opportunity to choose a  perspective he can
     adhere to, and we show that a single truth does not
     exist.
References
Goedkoop, M.J., Hofstetter, P., Muller-Wenk, R., Spriensma,
    R.The Eco-indicator 98 Explained, InternationalJour-
    nal on  LCA, December 98,  ECOMED  Publishers,
    Munchen

Hofstetter,  P.  (1998): Perspectives in Life Cycle Impact
    Assessment; A Structured Approach to Combine Mod-
    els of theTechnosphere, Ecosphere and Valuesphere.,
    Kluwers Academic Publishers, 1998, Info: www.wkap.nl/
    book.htm/07923-8377-X.

Thompson M,, Ellis  R., Wildavsky A.; Cultural Theory,
    Westview Print Boulder 1990

Goedkoop's Biography
  Goedkoop is a Consultant and  Scientist, born in 1954.
After obtaining an Electronic Engineering degree at Alkmaar
Polytechnic and an Industrial Design Engineering degree at
Delft University of Technology in 1983, he established an
industrial design engineering consultancy, called Evegoed.
In 1990 he shifted the attention to Ecodesign and Life Cycle
Assessment and renamed  the consultancy as PRe
Consultancy, (seewww.pre.nl)

   • PRe was in 1990 the first Dutch design consultancy
     exclusively devoted to Eco-design and LCA. We ex-
     ecuted one of the first LCA in the Netherlands.

   • From the start we worked on the development of LCA-
     based tools for Eco-design, like SimaPro LCA software
     (introduced in 1990) and the development of Eco-indi-
     cator 95 methodology. SimaPro is by far the most widely
     used LCA software world-wide.

   • Development of ECO-it, a low cost, simple tool for de-
     signers, based on Eco-indicators.

   The most important projects run by Goedkoop are

   • Active participation in seven  Eco-design  projects un-
     der the two National Dutch Ecodesign projects, as well
     as participation in several other Ecodesign-type projects.

   • Project leader for the Eco-indicator 95 Methodology un-
     der commission from The Dutch NOH project.

   Project leader for the follow up project, the Eco-indicator
 97/98 under commission of the Dutch ministry of Environ-
 ment, in collaboration with several companies and well known
 LCA experts in the Netherlands and Switzerland.
                                                      25

-------
Goedkoop's Overheads
  Impact assessment for Ecodesign
                 Mark Goedkoop
    PR6 Consultants B.V. Plotterweg 12 3821BB Amersfoort The Netherlands
           E-mail: goedkoop@pre.nl web site: www.pre.nl
                                                   The problem with LCA and ecodesign
• LCA's are too time consuming in product
  development processes.
• The results of LCA's are hard to interpret for a
  designer; often ten different scores for different
  environmental impact categories are presented
  as "the" answer.
• Designers will neve become LCA experts.
  They may not stay dependent on experts
       Stages in the design process
 Example: LCA of a coffee machine
Stase
Product idea generation
Definition of
requirements
Idea generation
Concept development
Detailed design
LCA tool
Trends, experience,
policy, etc
LCA of reference, calc.
of new indicators
Eco-indicators
Eco-indicators
LCA for verification
                                                              disposal of  I disposal in
                                                              filters + coffee I municipal
                                                              in org. waste I waste
   Result of coffee machine analysis
   7.3 fcg     1kg     0.1kg     0.3kg    0.4kg
   paper I |polystyrenc| | aluminium|  | steel |   I glass I"
            ; injection  ' extrusion
            I. moulding  f
                11 r~
                                   :onmng
                                 375 kWh

                                electricity
  Our solution: Pre-defined single
                 scores
 1  Simplification of the methodology is not the
   solution.
   - The results will not be reliable
   - Speed improvements are not sufficient
   Pre-defined Single scores are the solution...
   Important requirement: LCAs based on
   single scores should predict results of later
   full LCAs
                                              26

-------
     Calculating single scores

• In order to make predefined scores extremely
  careful modelling is needed.
• The model should contain all three spheres *):
  - Techno-sphere,
  - Eco-sphere and
  - Value-sphere
• Different value systems can now be made
  explicit.
) Sec also: Hofstetter, P. (1998): Perspectives in Life Cycle Impact Assessment; A
  Structured Approach to Combine Models of the Technosphere, Ecosphere and
  Valuesphere., Kluwers Academic Publishers, 1998
The three Spheres
      Modelling Technosphere
                                                 Typical characteristics Techno-sphere

                                                   • Description of concrete technical systems
                                                   • Uncertainties can be low (far less then an
                                                     order of magnitude)
                                                   • Verification is possible in many cases
                                                   • Main problems:
                                                      - Boundaries.
                                                      - Allocations
        Modelling Eco-sphere
                                                   Typical characteristics Eco-sphere

                                                   • Models complex cause and effect chains
                                                   • Usually high uncertainties (sometimes
                                                     several orders of magnitude)
                                                   • Verification is difficult or impossible
                                                   • Main problems:
                                                      - Limited scientific understanding of mechanisms
                                                      - Limited availability of data
                                                      - Different perspectives are possible
                                               27

-------
       Modelling Value-sphere
                                                Typical characteristics of Value-sphere

                                                   • Models aspects natural science cannot take
                                                     into account: "seriousness"
                                                   • Uses social science
                                                   • A single truth does not exist
                                                   • Main problems:
                                                     - How do we measure values in society
                                                     — How do we deal with incompatible views
           The three Spheres
                                                          Values in Eco-sphere

                                                  Example: There are three classes of carcinogenic
                                                   substances:
                                                    - Carcinogenic effect is proven
                                                    - Carcinogenic effect is probable
                                                    - Carcinogenic effect is possible
                                                  Do we include all classes?
                                                  Who will decide this?
  Cultural theory as management
        system of subjectivity
Three types of perspectives:
   - Egalitarian: Future generations just as
    important as present. In case of uncertainty take
    the worst case.
   - Individualist: Limit time scale, In case of
    uncertainty take proven facts only.
   — Hierarchist: Base analysis on what is agreed
    upon in policies and science.
Three different ecosystem models are needed!
 Three instead of one single score

• We will develop three alternative sets of
  single scores.
• Each set represents a cultural perspective.
• If A is better then B in all perspectives, then
  the result is robust.
• If A is not better then B, the result is
  dependent on perspective
                                             28

-------
     Compatibility with other
             approaches

Based on damage approach
Themes are made explicit as intermediate results
Characterisation not always based on
equivalency, but damages
If ISO compatibility is required, final weighting
may not be used in comparative assertions.
Eco-indicator society should promote proper
application
Overview Eco-indicator 98
             Conclusions

 Single scores need very careful modelling, as
 they need to predict the outcomes of full
 LCAs
 By separating three different perspectives we
 acknowledge the fact that different views are
 .possible, and that a single truth does not exist
 (unlike most full LCAs)
 Modelling three spheres is essential
 This method can also be used in full LCA
                                            29

-------
                              Application Dependency of  LCIA
                                     HenrikWenzel,Tech U of Denmark
  The reason to perform an LCA is essentially to use it in
support of a decision. A decision gives rise to a change
somewhere in society compared to a scenario in which this
decision was not taken. The key requirement for the LCA in
any application is, therefore, that it shall reflect the environ-
mental change caused by the decision. It is found that the
need to differentiate LCA methodology for the use in differ-
ent applications is born by a few key characteristics of the
decision to be supported.

  The first key characteristic is the environmental conse-
quence of the decision, i.e. the nature and extent of the
environmental change caused by the decision. When mod-
eling the environmental change, its extent in time and space
will differ between decision types, thus giving rise to differ-
ent requirements, primarily for the scoping and inventory
phases of the LCA. Furthermore, some decisions will imply
trade-offs between different impact categories, while others
will not, thus causing different requirements for the impact
assessment. The second key  characteristic is the social
and economic consequence oi the decision, the magnitude
of which will influence the need for certainty, transparency
and documentation. The third characteristic is the context
in which the decision is taken, including the decision-maker
and interested parties, implicitly influencing the impact as-
sessment and weighting.

Wenzel's Biography
  Wenzel has expertise in environmental Life Cycle Assess-
ment of products (LCA) and environmental design of prod-
ucts. Research  and development of methods and tools.
Special experience with electronics and electro-mechani-
cal products, textiles, packaging, furniture and energy sys-
tems.

  Wenzel has experience in Cleaner Production: Environ-
mental auditing, identification of options, research and tech-
nology development, implementation and dissemination.
Special experience within water based industry including
textile industry, paper industry, industrial laundries. Exper-
tise in reclamation and reuse of water and waterborne en-
ergy and chemicals. Development of methods and proce-
dures.

  He also has expertise in environmental impact assess-
ment of industrial effluents, and industrial wastewater treat-
ment and control. Special experience with chemical indus-
try, pharmaceutical industry and pulp industry. Development
of methods.
                                                    30

-------
Wenzil's Overheads
      Application Dependency of LCIA
                      Henrik Wenzel
             Institute for Product Development
       Technical University of Denmark, Building 424
                     DK2800Lyngby
                        Denmark
                 e-mail: wenzel@ipt.dtu.dk
LCA Applications

- some of the most useful

                          Companies

Life Cycle Management              X

Strategic Planning                  X

Product Development               X

Purchase                        X

Process Design/Selection             X

Marketing
- active claims                    X
- passive information                X
                                                                                                      Authorities
X

X
                                                                                                         X
                                                                                                         X
                  Goal Definition
           The proof of the pudding lies in the eating
    Goal and Scope Definition
                                                                                      The decision
                                                                                        maker
           In plain language: you know It's good if it's used
                     More philosophically:
      the identity of a thing is borne by its being used for its purpose
                 Scope Definition
                                 No decision / no env. consequence
                             - who cares about what method was used?

                                    Decision / env. consequences?
                             - defined scope and inventory data should
                              reflect the caused environmental changes
                                             = XA-BACO2
     Application dependency of LCA
     - hypothesis

     Three main variables govern application dependency:

     1.     The environmental consequence of the decision

           - time scale
           - spatial scale

     2.     The economic/social consequence of the decision

     3.     The decision context

           - the decision maker
           - interested parties
           - implicit assessment criteria
                                                          31

-------
 The Environmental Consequence of the Decision
 - differentiation of LCA applications
                       \
x    \   \   \
        \  \   \
        .   ^    ^  Prtwuctioit (ethnology isscunicnl
        X   \   \   \
         .   '-  \  \   \
         w   \   \    '  "
          \    \ X.
                 \
                   \
                                   NMdroriKespcdficarinfbnmUiin

  Figure 4. Examples of LCA applications which differ in the time scale of the
  decision and in their need for site specific information
                                                      The Economic/Social Consequence of the Decision
                                                      - differentiation of LCA application
                                                                              Need for certainty, transparency and documentation
                                                                                       Economic or social consequence

                                                         Figure 5. Examples of LCA applications with different need for certainty,
                                                         transparency and documentation caused by difference in economic or social
                                                         consequence of the decision to be supported
 LCA Applications in 3 Governing Dimensions
                  Need for certainty, transparency
                      & documentation
                                              Societal
                                                lion plans
                                             Time scale
                                                        Simplification of LCA Methods
                                                        Consensus: simplification is about delimitation and focusing. An
                                                        iterative procedure using all phases of LCA in increasing level of detail.

                                                                               Matrix LCA
                                                                              - qualitative or
                                                                              semiquantative
           -passjy

         ' Site speciality
                                                                              Screening LCA
                                                                               - quantitative
                                                                           - readily available data
                                                                                Full LCA
                                                                               - quantitative
                                                                                - new data
                                                                                jnventory/
                                                                                       Figure 3. Three iterations of LCA
Extent of Work Required at Different LCA Levels
- rough average time estimates
                                                     LCA Simplification and Iterative Procedure
                                                     - the reasons to continue with next iteration
        Extent of Work (days)
        100
         10
                                                                              Semi-quantitative
                                                                                Matrix LCA
                                                                Quantitative LCA
                                                                   - readily    —,
                                                                 available data    /
7
Is there a trade-off?
Can you conclude without impact
assessment?

Are data readily available?
Can you conclude without new data
inventory?
            Matrix LCA   Screening LCA   Full LCA
                                                               32

-------
LCA Simplification and Iterative Procedure
- continue as far as needed

         SumE-quuuititative
          Matrix LCA
  The three iterations of the LCA supports different decisions and conclusions

       The LCA application = each single conclusion/decision
The Decision Context
1.     Implicit assessment criteria

2.     Implicit weighting criteria

       - e.g. national priorities in LCA for national strategies
         and action plans

3.     Interpretable interpretation support

       - e.g. Dutch normalisation and weighting for a Dutch
         decision maker?
      Application Dependency of LCIA

      I.      LCIA only needed when LCA is applied on decisions
              involving a trade-off

      2.      There is a need for sophistication of site dependency
              in LCIA when decisions involving trade-offs cannot be
              supported without it

      3.      The need for site dependency in LCIA is largest for
              applications of LCA to trade-off decisions for which the
              consequences are limited to few specific sites

      4.      The need for sophistication of LCIA is largest in trade-off
              decisions with highest requirements for certainty

      5.      The decision context may imply certain weighting criteria

      6.      The background and knowledge of the decision maker
              may influence the choice of normalisation/weighting
                                                                      33

-------
                 Types of Uncertainty in  the Human Toxicity Potential
                                              Edgar Hertwich
                                           University of California
                                             Berkeley, CA, USA
  A framework for uncertainty analysis, which was origi-
nally developed to risk assessment (Finkel 1990), is ap-
plied to the exposure modeling component of the Human
Toxicity Potential  (HTP) (Hertwich 1999). The HTP is a
weighting factor, similar to that goal warming potential, that
is used to multiply emissions in a life cycle inventory to
obtain a single metric representing the human health haz-
ard (Hertwich et al. 1998).The HTP presents evaluations of
hazard based on the toxic potency of a substance and the
potential dose in a so-called unit world.The toxic potency is
expressed by cancer potency factors q1 * or the inverse of
the allowable daily intake or reference dose. The exposure
is calculated using CalTOX (McKone 1993; Maddalena et
al. 1995), a risk assessment model that integrates a multi-
media environmental fate model with  a  multiple pathway
exposure model. We distinguish between carcinogens and
non-cancer effects that are expressed in benzene and tolu-
ene equivalents, respectively. For the same chemical, the
HTP will depend on the compartment of the release.  HTP
comes significantly closer to an actual  risk  assessment
than alternative approaches without requiring site-specific
input data (Hertwich et al. 1998). HTP values were calcu-
lated for approximately 300 chemicals representing > 80%
of the air emissions reported by U.S. manufacturing facili-
ties in theTRI. HTP values for air and surface water emis-
sions can be downloaded from http://www.scorecard.org/
env-releases/def/tep_gen.html

  The framework for uncertainty analysis distinguishes
among the uncertainty and variability in input parameters,
model uncertainty, and the decision rule uncertainty.

1.  Parameter uncertainty: measurement errors (resolu-
    tion of instrumentation), random errors (sampling er-
    rors-representativeness of entire population) and sys-
    tematic errors (biases introduced through experimental
    design or instruments; e.g., the "healthy worker effect"
    according to which epidemiological studies are not rep-
    resentative for the general population  because workers
    in general are of better health than the average citizen.)
    Parameter uncertainty captures the possible errors as-
    sociated with the inputs to mathematical models. These
    uncertainties can also be due to estimation methods
    and guesses.
2.
The CalTOX model uses three different sets of param-
eters: chemical-specific parameters (decay rates, vola-
tility), landscape parameters (biomass inventory, pre-
cipitation rate), and exposure parameters (diets, breath-
ing rates). For the purpose of the present assessment,
we assume that chemical-specific data is uncertain,
whereas landscape and exposure parameters are vari-
able.

The uncertainty associated with chemical specific in-
put parameters was investigated using Monte Carlo
analysis (Hertwich et al. in press). First, chemicals were
grouped according to the most important exposure route.
For each group, a representative chemical was selected.
For each of these chemicals, Monte Carlo simulations
were conducted and the importance of the input param-
eters was evaluated using rank correlation.The analy-
sis indicates that the uncertainty varies between 1/2
and 3 orders of magnitude, but is commonly around
one order of magnitude.This is little compared to a varia-
tion of potential dose among chemicals of 7 orders of
magnitude and indicates that there is a significant gain
in information from modeling the potential dose. The
most important input parameters were reaction half lives,
as a sensitivity analysis indicates. The results show
that the uncertainty due to the exposure modeling can
be as large as the uncertainty in the toxic potencies,
which has been evaluated by others (Gaylor et al. 1993;
Bairdetal. 1996).

Variability expresses the known natural variation of spe-
cific parameters,  such  as diets, proximity to emission
sources, temperatures and weather patterns, annual or
diurnal variation of emissions rates, transformation rates,
and transport. Variability differs from parameter uncer-
tainty because it is not related to how precisely we can
measure a specific quantity but it is due to natural varia-
tions of a parameter.

The influence of variability in landscape and exposure
parameters was investigated together with the param-
eter uncertainty using Monte Carlo analysis, as de-
scribed above (Hertwich et al. 1998). For most chemi-
                                                     34

-------
    cals, the variation in landscape and exposure param-
    eters is insignificant compared to the uncertainty in
    chemical specific parameters. In some cases differ-
    ences in exposure parameters have a significant ef-
    fect. What about uncertainty in the exposure param-
    eters?

3.   True model uncertainty captures the uncertainty that
    is introduced through making models, the idealization
    of reality. Model uncertainty includes the uncertainties
    introduced by surrogates (the uses of rodents as a model
    for toxic effects in humans); through the exclusion of
    specific mechanisms; through the inability of models
    to capture abnormal conditions which may have a large
    impact on the outcome; and through incorrect model
    form (errors in the model formulation, or insufficient
    knowledge about the true nature of the investigated
    model processes). An example for model uncertainty is
    that the shape of the dose-response curve in cancer
    risk assessment is usually not known, and the selec-
    tion of any specific model has a large impact on the
    predicted cancer risk at ambient environmental expo-
    sures (low doses) from the extrapolation of rodent test
    results at high doses (Chiu et al. 1999).

    Two examples of model uncertainty were investigated:

    1) Two alternative model formulations for the air-plant
    interactions in the multimedia model CalTOX were com-
    pared. For half of the chemicals, the concentration in
    the plant compartment differed significantly. Significant
    differences in human exposure were observed only for
    20% of the chemicals, and the differences were always
    below a factor of 5.

    2) The effect of the steady state assumption for wet
    deposition was investigated. For 10% of the chemicals,
    the continuous rain led to a significantly lower air con-
    centration than a situation without rain. This led to dif-
    ferences in exposure of up to 2 orders of magnitude.
    The steady state assumption for wet deposition needs
    to be questioned for chemicals with low Henry's Law
    constants. While the rainfall rate has little difference on
    the steady-state results, the total absence of rain does.
    This problem was addressed by combining the model
    results with and without rain.

 4.  Decision rule uncertainty arises whenever there is
    ambiguity or controversy about how to quantify or com-
    pare social objectives. Examples include choosing an
    appropriate measure to describe risk, choosing a sum-
    mary statistic to characterize uncertain risk, choosing
    parameters that define acceptable risk (such a one in
    million chance) or the monetary value of risk (the value
    of a life saved or a disease avoided), choosing a sum-
    mary statistics, choosing a social welfare function, and
    choosing a method for trading off immediate versus
    delayed consequences.

     In LCA a decision rule specifies how impacts should be
    assessed (selection of assessment method, choice of
   a specific endpoint,  etc.). Decision rule uncertainty
   arises when there is ambiguity about the goals of the
   analysis. For example, HTPs are based on the product
   of toxic potency and potential dose. Potential dose in
   models like  CalTOX can be assessed using different
   specifications of removal of the  pollutant. In a closed
   system, no advection is allowed and the entire life cycle
   of the pollutant is tracked by the  model. In an open
   system, advection is allowed and  the assessment is
   closer to a region-specific assessment. This difference
   in boundary conditions can have a large influence on
   the HTP values, especially for persistent pollutants such
   asTCDD.
References
Baird, S. J., J. T. Cohen, J. D. Graham, A. I. Shlyakheter and
    J. S. Evans (1996). "Noncancer Risk Assessment: A
    Probabilistic Alternative to Current Practice." Hum. Ecol.
    Risk Assess. 2(1): 79-102.

Chiu, W., D. Hassenzahl and D. Kammen (1999). "A compari-
    son of regulatory implications of traditional and exact two-
    stage dose-response models." Risk Anal. 19(1): 15-22.

Finkel, A. M. (1990). Confronting Uncertainty in Risk Manage-
    ment-A  Guide for Decision-Makers. Washington, Re-
    sources for the Future.

Gaylor, D. W., J. J. Chen and D. M. Sheehan (1993). "Uncer-
    tainty in Cancer Risk Estimates." Risk Anal. 13(2): 149-
    154.

Hertwich, E. G. (1999). Toxic Equivalency: Accounting for
    Human Health in Life-Cycle Impact Assessment. Ph.D.
    thesis, Energy and Resources Group, University of Cali-
    fornia, Berkeley. 237 p.

Hertwich, E. G., T. E.  McKone and W. S. Pease (in press).
    "Parameter uncertainty and variability in evaluative fate
    and exposure models." Risk Anal.

Hertwich, E. G., W. S. Pease andT. E. McKone (1998). "Evalu-
    ating toxic impact assessment methods: What works
    best?" Environ. Sci.Technol. 32(5): A138-A144.

Maddalena,  R. L, T. E. McKone,  D. W. Layton and D. P. H.
    Hsieh (1995). "Comparison Of Multi-Media Transport and
    Transformation Models - Regional Fugacity Model Vs
    CalTOX." Chemosphere 30(5): 869-889.

McKone, T. (1993). CalTOX, A Multimedia Total Exposure
    Model for Hazardous-Waste  Sites.  Livermore, CA,
    Lawrence  Livermore National Laboratory http://
    www.cwo.com/~herd1 /caltox.htm

Hertwich's Biography
  Hertwich is a Ph.D. candidate in the Energy and Resources
Group (ERG) of the University of California, Berkeley. He
holds a BA in physics from Princeton University and an MS
in Energy and Resources. His work on the application of
                                                      35

-------
multimedia, multiple pathway fete and exposure models lead
to the development of toxic equivalency potentials for can-
cer and noncancer effects. This weighting system is being
used by the Environmental Defense Fund's Scorecard project
[www.scorecard.org] to weight toxic emissions from all major
US manufacturing facilities.
Hertwich's Overheads
     A Framework for the Uncertainty
      Analysis of the Human Toxicity
                      Potential

                   Edgar G. Hertwich
                  Thomas E. McKone
                   William S. Pease

             University of California, Berkeley

             http://greenmfg.me.berkeley.edu/~edgar
                   http://www.scorecard.org
   Finkel's Framework for Uncertainty
                    Analysis

   Parameter uncertainty: random, systematic, and
     measurement errors
   Model uncertainty: due to simplifications , the use
     of surrogates, errors in model form etc.
   Decision rule uncertainty: ambiguity or
     controversy about social objectives
   Variability: natural variation of
     input parameters
                  The Purpose of

               Uncertainty Analysis
     Uncertainly analyiis it a tool for recognizing how prudent or imprudent proposed
     actions may be fn trie face of incomplete knowledge, and for modifying control
     and research decisions accordingly.   (Finkel, p. 39)

    Uncertainty Analysis is required :
    •  to develop confidence in an analytical result
    •  as an input to formal decision analysis techniques,
       and
    •  as a tool to refine impact
       assessment methods.
         Decision Rule Uncertainty
  In order to take any actions using the outputs of a risk assessment, including the
  decision not to take action, one must be prepared to make a series of potentially
  controversial value judgements.
                    (Finkel 1990, p. 16)
           r
           I"
                                                              Decision: do we allow advection of the pollutant out
                                                                          of the control volume?
                Model Uncertainty
                  Case: Plant Compartment
              Model Uncertainty
                  Case: Continuous Rain?
     Alternative models for the interaction between plant
     and air, and their effect on human exposure, as well
     as on the concentration in the plants.
                                                                 1 E-1 4
                                                                                   >*>
                                                                 ,Mj     ',/   s v-  * ^   V
        1 E-4 -1-^-
          1 E-2             1 E-1
                   Tres_air [da for 'no rain'
                                                      36

-------
     Model Uncertainty
         Case: Continuous Rain?
  IE-12   1E-10   1E-8   1E-6   1E-4   1E-2   1E+0
        Parameter Uncertainty
          Chemical Specific Parameters

1 E-2 •
jg 1 E-4 .
j> 1 E-S • *
1 E-7 •


Alremlssic
* 1 1
'1
".
T* *
Iff
i = s
•s
ns
ฃ
1

0
1


=


it
f 1
*.
Surface water emission:
1 1 i I H
• s 1
' *',*ป Mean
' SOXiie •- _ ,,
L j'h
r* * lOSIIe
V'
— --SSIIfr
fill!
' ^ * ฃ
         Variability
   Landscape-Specific Parameters
           Total Adult Exposure
            US vs. California
     y = 09457x 18618    '     !,.>
       Ra = 09845         -ffli^
                  - Jป>T
                    J-Jg.
                      calculated exposure
                      resulting Item air
                      eraisslonj of 1 fnoi/fl
                      Into closed systems
                      based on contiguous
                      U.S.andCaltrotnla—
                      Landscape data from
                      PATRIOT, I
    -12   -10
           Exposure, California
               Log units
Error bars indicate chemical-specific parameter uncertainty
              Conclusions

•  Finkel's framework is useful for organizing
  uncertainty

•  Uncertainty Analysis is exploratory and
  never complete           l

•  Some uncertainties are significant, others
  are negligible

•  This type of analysis provides useful insight
  into the power and limitations of HIP

•  Uncertainty analysis can help to develop
  and refine analytical decision-support tools.
                                            37

-------
              The Different  Levels of Uncertainty Assessment in LCIA;
                             The  Case of Carcinogenic Effects
                                     Patrick Hofstetter, Env. Sciences:
                              Natural and Social Science Interface, ETH Zurich
  Life Cycle Impact Assessment (LCIA) aims to assess
the changes in the environment which are judged to be dis-
advantageous, i.e., damages to safeguard subjects (repre-
sentatives of the environment). The environmental system
as such can be best described as overcomplex (Berg et al.
1994). The identification of disadvantageous changes and
of safeguard/subjects are inherently value-laden steps. So-
phistication in LCIA methods means therefore not just to be
more detailed or comprehensive in modeling but as well to
pay attention to the different levels of uncertainties by the
use of appropriate methods.

  Van Asselt ef al. (1996) identified a similar situation in
integrated assessment and chose to use a typology of un-
certainties based on Funtowicz et al. (1989) (see typology
in the annex). In LCIA we can distinguish three levels of
uncertainties which ask for sophisticated methods to deal
with them:

1. The environmental system is overcomplex, only some
   causal relationships between environmental interven-
   tions  and damages to safeguard systems can be es-
   tablished. Lack of understanding and disagreement
   among experts lead to epistemological uncertainties.

2.  Many steps in the assessment need value-laden judge-
    ments, e.g., the perception of the ecosphere or the time
    preference are dependent on the individual world-view.
    Subjectivityis therefore a further source of uncertainty.

3. The quantitative modeling of the ecosphere is partly
   based on approximations using data with inherent vari-
   ability and statistical variation. This leads to technical
    and some methodological uncertainties.

  Suggestions to deal with these three levels of uncertainty
will be given here. However, the focus will be on  level 3
using the example  of carcinogenic effects in humans. A
detailed description of these steps including data for respi-
ratory and carcinogenic effects is available in Hofstetter
(1998). A more concise description of the damage assess-
ment approach was published in Goedkoop et al. (1998).

1) Epistemological uncertainty
  It is suggested to model the overcomplex ecosphere by
three submodels. The known damage is modeled by the
available knowledge on causal relationships between envi-
ronmental interventions and damages to safeguard subjects.
This includes fate, exposure, effect, and damage analyses
(impact-pathway analysis) and requires that safeguard sub-
ject and damaging changes to safeguard subjects are de-
fined beforehand. The unknown damage is a proxy for all
interactions in the ecosphere we do not know (yet). For the
time being it is suggested to use the ratio between the po-
tential to accumulate and the present geogenic plus anthro-
pogenic flows as characterization factors to construct the
proxy indicator. The third submodel addresses the manage-
ability of damages that is operationalized by three indica-
tors: Ease of damage reduction, excess of long-term policy
targets, success of regulation.

  These three submodels permit inclusion not only of a vast
amount of information on causation but also of precaution-
ary principles and dynamic aspects.

2) Subjectivity

  It is suggested to model and manage the subjectivity rather
than to hide or separate it.Therefore, Cultural Theory (Th-
ompson et al. 1990) has been chosen to model society.
Three of the five archetypes described by Cultural Theory
have been chosen for a sophisticated scenario generation
because of their active role in decision-making: Individual-
ists, hierarchists, and egalitarians. The aim was to model
all value-laden sub-steps compatible with these cultural
perspectives,  i.e., to come up with three instead of one
LCIA, each one being based on a consistent set of value
judgements.

  In the case of the three submodels for the ecosphere this
means that individualists will place most importance on the
submodel known damages while egalitarians will follow a
more precautionary approach and judge the submodel on
unknown damages to be the most important.

  Tumors at different cancer sites give rise to individual
histories of disease and  lead - to some extent - to prema-
                                                    38

-------
ture death at different ages. The years life lost (YLL) per
cancer site and the average duration and severity of dis-
eases (leading to the years lived disabled (YLD)) can be
aggregated to disability adjusted life years (DALYs) (Murray
et al. 1996).To do so a number of value choices are neces-
sary on equity between races, age, and social status, on
time discounting, and on the appropriate procedures to es-
timate the disability weights. This being a typical example
where value choices are needed. World-view dependent
judgements on these factors are summarized in the appen-
dix.

3) Technical and some methodological uncertainties

  The damage assessment of substances causing carci-
nogenic effects in human beings is an illustration of how to
deal with methodological and technical uncertainties in quan-
titative modeling (see appendices). First of all it is assumed
that point estimates have to be supplemented by informa-
tion on the probable distribution. For reasons of simplicity it
is assumed that all parameters are lognormally distributed
and that the 95% confidence interval is described by the
squared geometric standard deviation o2. The tables and
figures in the appendix indicate the impact pathways for
damages to human health and illustrate how the uncertain-
ties have been dealt with.

  Most of the fate factors (F) have been modeled using a
multi-media steady-state model.The accuracy of the model
outcome depends on the quality of the input parameters,
the suitability of the model for the substance to be mod-
eled, and by the ways substances are distributed and parti-
tioned. Using these factors, uncertainty classes for the sub-
stances at hand have been defined and geometric standard
deviations estimated. In the case of heavy metals dry and
wet deposition velocities from literature were used to make
more accurate estimates for the air to air fate factors, which
leads to much smaller geometric standard deviations than
with the fugacity model. The fate factors for heavy metals
and some other substances that do not conform to the limi-
tations of fugacity models should be further improved by
using other models for intake by drinking water and food.

  The exposure of  human  beings is dependent on many
factors and shows high variability. One of the more decisive
factors is the population density. Assuming that an  emis-
sion occurs in Europe and an average wind speed of 5 to 10
m/s, one can assume that substances with high atmospheric
residence times will be distributed over a large area includ-
ing less densely populated areas in Asia and the sea. Tak-
ing into account the geographical situation, the population
densities in the concerned countries and a maximum popu-
lation density at the emission source results  in the graph
included in the appendix. A substance-specific population
density has been attributed to each substance.This proce-
dure significantly reduces the uncertainty in the exposure
assessment.

  Effect data for carcinogenic substances is available from
epidemiological studies,  bioassays and other toxicology
studies. The International Agency for Research on Cancer
(IARC) groups the substances according to the strength of
evidence for their causing of cancer in human beings (Group
1: highest evidence). The selection of the lARC-Groups to
be included in the analysis can not be based on scientific
reasoning and has nothing to do with the quality of the dose-
response factors. Cultural Theory is used here to select the
groups that fit best with each of the three cultural perspec-
tives, i.e., it is treated like a question of value choices (level
2). In addition to this epidemiological uncertainty there is
the uncertainty  due to variability  and approximation. The
extrapolation to low doses and the transferability from ani-
mals to humans are major uncertainty factors. Thanks to
some epidemiological studies (mostly for inhalation) for
substances subsumed under Group 1 and 2, one can as-
sume that the geometric standard deviations is lower for
substances with higher evidence and for the inhalation.

  The following tables show  the aggregation of all the
substeps of the fate, exposure, effect and damage analy-
ses for some carcinogenic substances of lARC-Group 1. It
can  be seen that the geometric standard deviations are
especially large for the heavy metals which poorly fit with
the fate model used. The figure with error bars in the appen-
dix represents the results of the damage assessment for a
set of 55 carcinogenic substances that are emitted to air,
including all the exposure routes. The 95% confidence in-
terval spans 1.5 to 5 orders of magnitude. However, the
highest and the lowest mean for a substance show a differ-
ence of eleven orders of magnitude, i.e., the uncertainty is
reduced from eleven to at least five and in many cases to
two or three orders of magnitude.

  An additional reduction of uncertainty is possible thanks
to the three distinct LCIA models for each of the three cul-
tural perspectives. In the case of a one-score eco-index for
a specified product system, the probability distribution for
just one cultural perspective will be much narrower than the
value-independent probability distribution for all combina-
tions of value choices.

  The three levels of uncertainty discussed each asks for a
different tool to deal with it. The solutions presented here to
deal with these  uncertainties will enhance the sophistica-
tion of LCIA, on the one hand, and  make LCIA results more
valuable as decision support, on the other hand.This com-
prehensive modeling approach may well lead to a robust
LCA that is simple to use in practice.

References
Berg M., M. Scheringer, Problems in Environmental Risk As-
    sessment and the Need for Proxy Measures, Fresenius
    Environmental Bulletin 3 (8): 487-92 (1994)

Funtowicz S.O., J. Ravetz, in:  Les Experts sont Formels:
    Controverse Scientifiques et Decisions Politiques dans
    le Domaine de I'Environnement, Arc et Sanas, France
    1989

Goedkoop M., P. Hofstetter, R. Miiller-Wenk, R. Spriensma,
   The Eco-Indicator 98 Explained, Int. J. LCA 3 (6) 352-
   360(1998)
                                                     39

-------
Hofstetter P., Perspectives in Life Cycle Impact Assess-
    ment; A Structured Approach to Combine Models of
    theTechnosphere, Ecosphere, and Valuesphere, Kluwer
    Academic Publishers, Boston 1998

Murray Ch.J.L, Lopez A.D. (Eds.), The Global Burden of
    Disease, Volume I of Global Burden of Disease and
    Injury Series, WHO / Harvard School of Public Health /
    World Bank, Harvard University Press, Boston 1996

Thompson M., Ellis R.,  Wildavsky A., Cultural Theory,
    Westview Print Boulder 1990

van Asselt M.B.A., A.H.W. Beusen, H.B.M. Hilderink, Un-
    certainty in integrated assessment: A social scientific
    perspective, Environmental Modeling and Assessment
    1(1996)71-90

Hofstetter's Biography
  Hofstetter, 1989,  graduate degree in mechanical engineer-
ing ETH Zurich; 1989 LCA of light bulbs including an evalu-
ation of methods for LCIA; since 1989 own consultancy "Biiro
fur Analyse & Okologie" in the field of industrial ecology,
processes for waste treatment, LCA, and eco-design. 1990-
1994 research for the report and database "Okoinventare
fur Energiesysteme" (Life Cycle Inventories for Energy Sys-
tems), at the Energy Systems Laboratory, ETH Zurich. Since
1994 member of the LCA-Steering Committee of SETAC-
Europe; 3. - 6. 1994 research stay at the Centruum voor
Milieukunde (CML), University of Leiden; since 1994 research
on Life Cycle Impact Assessment. 1998 Ph.D. thesis on
"Perspectives in Life Cycle  Impact Assessment; A Struc-
tured Approach to Combine Models of the Technosphere,
Ecosphere, and Valuesphere" at the Chair of Environmental
Sciences: Natural and Social Science Interface, ETH Zurich.
Since September 1999 he is an ORISE research fellow at
U.S. EPA, Cincinnati and visiting scientist at Harvard School
of Public Health, Boston.
                                                    40

-------
Hofstetter's Overheads
                  EidgetiSssische
                 Technische Hochschale
                ZSrich
 Ecole polylechniquefederale de Zurich
Polilecnico federate di Zurigo
Swiss Federal Institute ofTechnology Zurich
                        An International Workshop on
               Life Cycle Impact Assessment Sophistication
                       Brussels 29-30 November 1998
              Co-organised by UNEP, U.S. EPA, and CMLLeiden

    THE DIFFERENT LEVELS OF UNCERTAINTY ASSESSMENT IN LCIA:
                 THE CASE OF CARCINOGENIC EFFECTS

                              Patrick Hofste tier,
      Environmental Sciences: Natural and Social Science Interface (UNS)
               Swiss Federal Institute of Technology (ETH Zurich)
                        ho6tetter@uns.umnw.ethz.cli
                                                                                  Multifactor interaction network in carcinogenesis based on some of the major
                                                                                   relationships identified (dashed double lines that break an arrow symbolise
                                                                                             inhibition or blockage of the path) (Arcos et al 1995)
      How to deal with main problems in LCIA.?

      Environmental system and its causal relationships are overcomplex

      =>        Assessment has to reflect epistemological uncertainty
                (known damage, unknown damage, manageability)


      Perception of the ecosphere and its behaviour is dependant on the
      perspective

      =>        Assessmentshould cope with this shaped perception
      (subjectivity)
                (different models for different perspectives, Cultural Theory)


      Modelling is hampered by technical and methodological uncertainties

      =>         Use quantitative uncertainty assessment to cope with
                 technical uncertainties and use more than one method to
                 reduce methodological uncertainty
                    (Monte Carlo Simulation, combination of descriptive AND
                             prescriptive approaches)
                                      The modelling framework for the ecosphere (the arrows - the third modelling
                                         level of'manageability1 - symbolise the dynamic aspects of a damage
                                                    reduction towards an acceptable damage).
                                           HtncUlft modcUftfie ci
                                                                                   Technosphere
                                                                                                   Inventory Analysis
                                                                                                 A\\\\\\\\l  "
                                                                                                I   Fate and Exposure
                                                                                                ฅ  Analyses
                                                                                             Exposure

                                                                                                I   Effect Analysis

                                                                                      Effects on Single Endpoints
                                                                                        Damage to Safeguard
                                                                                              Subjects	
                                                                                                                   Models*^
                                                                                                                          Single compartment
                                                                                                                                          Steady state
                                                                                             ' Steady state
                                                                                             „ Dynamic
           Multi-media
   Empirically derived fate factors
 jf Toxicological tests, laboratory experiments
•C                  jf Ecological studies
 ^ Epidemiologica!studies ^-Cohortstudies
                   ^- Cross-sectional studies
                                                                                                                             —

                                                                                                                 > Revealed preferences
                hdlvldmilht'i modcUriht ci
                                                 'i ปoikl offtirf ojpher
             Fig. 4.11:  The relevance of the three submodels for the ecosphere judged
               from the viewpoint of the three active perspectives. The question mark
             indicates that fatalists need not to be considered here because they are no
                                active decision makers.
                                        General set-up of the impact pathway analysis within the ecosphere. The
                                       upperarrows for each level of analysis give examples for the cause-oriented
                                            modelling and the lower arrows specify the descriptive modelling
                                                                               41

-------
Empirical data to reduce uncertainty of fate factors for heavy metals



wtocity\fr votocilyVp "
(FfeeynalDSS) (Pacynaefa/.
19S5)
™* rn'o/irf cnVs m'a/m" nftlnf
(1) 100(((1).36S-24. (2) 100/((2)-36 ' 10tt/(((2)40.2)-
36001 5-24- 36001 365-21-3SO01
WMI* ttl-0.4 32E.6-OE-8 0.2 16E-6 8E-6 4
ซ*•! 04-2 8E-8-1.6E-8 0.5 6E-6 4.5 E-6 4
Cปซ0 is BB a
Calculation of the fate factors F *>,*• from the deposition velocity. The
wet deposition velocity is calculated by the ratio of the assumed dilution
height of 800 m with the residence time due to wet deposition (Tซซ = 5
days) resulting at about 0.2 cm/s.
•mat
•ace
o .
Substance-specific population density
~\^
6M1 Q001 001 01 l '
AtmMptmto mUvMM Ura bl air [ylln]
Population density as a function of the atmospheric residence time of substances emitted
in Western Europe according to Equation (6.12).
Variability and approximation of unit risk factors
IARC-Group O2 for inhalation Og for drinking water and oral intake
1
2A
21!
3

2 4
3 6
S 10
10 20
Estimated geometric standard deviations Qognormal
distribution) for the unit risk factors given in
Appendix (6.7).
Main questions: Extrapolation to low doses
Transferability from animals to
humans
Epidemiological studies are important for validation
Uncertainty classes for different substance classes
CMteria Substances concerned 222
our off af
(for emission tซ air) P'air. F'air P'uir.
•More than 95% of the total Methyl chloride, 1.1.1.2-Telrachtoroethane, 1.2- 248
intake is inhaled, i.ev. the Dichloroethane. Chloroform. Dicbtoromelhane, 1.2-
substance shows only minor Dibromoethune, Hexachloroelhune, 1,1,2-TridiIoroethane,
partitioning Perchloroelhylene, Brornodichloromelhane. Ethybne oxide.
• The properties are well known Benzene- B^yl-chloride. Trichlorcthyfcne, Vinyl chloride.
Acrylonitrile, Epichlorhydrin, Diesel soot particles, 1,1
• The substance is well suited Dichtorethylene. Carbon- telmchloride, Styrene. DCEE,
for the chosen model BCME Acetaldehyde, 1,3- Butadiene
• Tlie substance parnt^ns Trifluralin. Hexachlorobenzene. alpha-HCH, Dichbrvos. 3 f, 12
• The properties arc well known DEHP' ^^^^^^ l^-HCH. R>rmaldehyde.
Aklrm. gammu-HCH. l,4-Dbxane,PCBs.
• The subsume K well suiled Benzfajanlhracene. BaP. Dieldrm. 3-MKthyk;hloanthrene,
for the chosen model TCDD, Dibenz[aJi]anUiracene
• The properlks are uncertain or Arsenic. Cadmium. Nickel. Nickel-ref.-dust. Nickel- 20 4() ffl)
"nt™™- w sulKiilfide, airomium(VI). Propylene-oxide. 1 .1.2.2- (ง)
• The substance Is not well Telrachloroelhme, Benzotrichloride. Hexachlorobutadiene.
suiled for the chosen model 2.4.fi-TrichJorphenol. PenUu;h!orphenol

IARC - Groups of evidence: Epistemological and methodological
uncertainties
Cultural Attitude and Argumentation Choice
Perspective
Individualist The individualist prefers an adaptive management style which means that resources Group 1
are spent fur pniven harm tu human health at Uie moment of decision. The decision
will immediately be adjusted if new evidence emerges so as to realise an optimal
resource allocation. Risks which have not yet resulted in evident harm are accepted.
Such an attitude makes i( possible to consider Group I substances only because of the
limited evidence for all other substances.
Hierarchist The hierarchic I's management style is to control situations. He is therefore closest to Groups
the authoritative bodies in government and in international organisations. The 1, 2A,
responsibility for the group and towards the leader of the group demands finding a ,23*
balance between evidence and probability. WHO (1987, 1996a) gives in its guidelines
risk estimates fora small selection of Group 1. 2A and 2B substances. Substances
from Group 3 are not included because this could lead to an unjustified cut-back of
needs already expressed by society.
Egalitarian The risk avcrsive and preventive management style asks for an attentive altitude Groups
towards risk management. Egalitarians are well aware that many substances in the 1-3
inclusive Group 3 mighlln the coming decades be proven to be potent human
carcinogens. The responsibility towards the others in the group (and coming
generations) leads to the inclusion ofall risk factors. Egalitarians may be ready to
reduce the satisfaction of their needs in order to spend more resources on risk
prevention than individualists or hierarchists would accept (however this is here not
the important point).

Variability and approximation of the damage analysis by
Disability Adjusted Life Years (DALYs)
Cancersiles YLI,m(0,0) Average Average YLDm(0,0) DALYsm(0,0>
disability duration of CTS
weights Din disease Lm
(Murray el (Murray el
al. 1996a) ui I996b)
Lu"g 15.8 0.146 2 0.29 161 15
Skin melanoma 3.2 0.045 4.4 0.20 3.4 1 5
Nasal, nose, nasopharynx 12.1 0.096 4.3 0.41 12,5 2
Uver 15.8 0.239 1.77 0.42 1&2 1 5
Leukaemia 16.0 0.06 3.8 0.45 165 15
AH HHffl 	 "ft 01ซ 
-------
Discounting and age-weighting in the
DALYs-concept per cultural perspective
                                               Individualist    Egalitarian   Hierarchist
Discount rate derived from Cultural Theory                 5%

Discount rate due to LC A sensu stricto restrictions           0%
                                                                     0%
                                                                                  2%

                                                                                  0%
  Age-weighting derived from Cultural Theory
                                                       yes
                                                                                         Multiplying Lognormal Distributions
                                                                                         ซ?= eP   ซg = geometric mean, i.e., expectation value

                                                                                         ag=ea   ag = geometric standard deviation

                                                                                         The product of n independent lognormally distributed random variables Xj j = l...n
                                                                                         with parameters (/Jgj, agj)j = l-n is lognormally distributed with
                                              expectation value

                                              fig tot = IT(j-j__n tig}  and



                                              geometric standard deviation
               Composition of aU elements in the Impact-Pathway Analysis
                              (Uncertainties in quantities)
                                                                                           Compilation of the geometric standard deviations of aU elements in the Impact-Pathway
                                                                                           Analysis
 Substance-name
                 vlARCX
 Arsenic, inorganic
 Benzene
 Bis(chloromelhyl)etl>er BCME

 Chromium(VI)
 Nickel
 Nickel-refinery-dust
 Nickel-subsulfide
 Vinyldilorid
                      3roup*l
                                  eg2          a?
                            DALYs air.    DALYs air,
                              inhalalion  drinking water
                                                    cV         erg2        ag2       cfg2
                                                 DALYs DALYs water, DALYs water,    DALYs
                                                 air.food    inhalation drinking water  waler.food
                                                                                                                       as2  eg2   og2   of    
-------
 Conclusions

 •    Dealing with overcomplex and ill-defined systems asks for:
     - tools dealing with technical, methodological, and epistemological
      uncertain lies
     - not aiming for world-models but sufficient information for ranking of
      alternatives, i.c., most important impact pathways are mostly sufficient

 •    Present knowledge on causal relationships between interventions and
     damages shall be compiled and harmonised (and supported by proxy
     indicators)

 •   This tremendous work will allow fora feedback to science policy and
     priorisation in the research of the concerned disciplines

 •   The turn-around via increased sophistication will allow for simple to use
    and robust LCIA

•   Quantitative uncertainty assessment allows to identify non-significant
    differences between product alternatives (important finding)
                                                                    44

-------
            Human Toxicity and  Ecotoxicity - Modeling Versus  Scoring


                                               O. Jolliet
           Swiss Federal Institute of Technology Lausanne, Institute of Soil and Water Management
                             Ecosystem management, CH-1015 Lausanne EPFL
                          olivier.jolliet@epfl.ch, http://dgrwww.epfl.ch/GECOS/DD
Introduction
  In order to avoid the present possibility to choose an im-
pact assessment method in order to obtain a desired result,
there is an urgent need for criteria to select consistent im-
pact assessment methods. The present paper aims at the
following objectives:

  - to define criteria for consistent effect, fate and expo-
   sure analysis,

  - to identify key parameters for the fate assessment and
   to analyze different fate models on the basis of these
   parameters,

  - to discuss the possibility of overall experimental checks
   and finally,

  - to conclude on modeling versus scoring approaches.

Framework for Fate and Exposure in
     Toxicity: Consistency  Requirements
  The first SETAC-Europe working group on Life Cycle Im-
pact Assessment (Jolliet et al., 1996) developed a frame-
work for human and ecotoxicity where characterization fac-
tors are expressed as the product of an effect factor times
a fate factor (fig. 3).This framework also provided a consis-
tency principle: the Life Cycle Impact assessment method
performs a full fate and exposure analysis if the fate factor
F effectively relates the inventory emission to the chosen
toxicity effect factor (fig. 4). For instance,  if a Reference
Dose is used as the toxicity reference, then the emission
should be related to the amount of pollutant effectively in-
haled or ingested.

Identification of Key Parameters for
     Exposure Efficiency
  Jolliet and Crettaz (1999) introduced the concept of hu-
man exposure efficiency in LCIA: retaining a dose as toxic-
ity reference, the consistency principle means that the
fate factor for human toxicity can be described as the over-
all fraction of an emission which is taken in (inhaled or
ingested) by all human beings over the lifetime of  a
substance (fig. 3):
      	 r~ I""* 	
      — G ;  —
mass of substance i emitted in media m and
taken in by mankind through pathway n
  mass of substance i emitted in media n
  Evans et al. (1997) used the same concept for risk as-
sessment, calling the ratio of total human intake to total
emissions the "exposure efficiency" (e).

  Jolliet and Crettaz (1999) also developed an analytical
and modular approach to calculate exposure efficiency (fig.
5). The new model is developed on the basis of a modular
approach which offers an alternative way of simplification
than the traditional Mackay I, II and III classification. A sum-
mary of this calculation is provided in the appendix. The
approach allowed characterization of the fate and exposure
in each media using the five following key parameters (bold-
faced):

  a) in a first step, fate and exposure is calculated for each
media, assuming that the concentration in other media is
null. Concentration increases are proportional to resident
times divided by the height of dilution. To obtain expo-
sure, these factors are then multiplied by the total amount
of air, water and food ingested annually by humans and
by bio-concentration factors for food exposure.

  b) In a second step, the different compartments are con-
nected on the basis of inter-media transfer factors (frac-
tion of the emitted substance in medium m that reaches
medium n)

  c) It becomes possible to take into account the exposure
to daughter degradation product of a substance, introduc-
ing an inter-substance transfer factor, the mass of daugh-
ter degradation product j per unit mass of substance i ini-
tially emitted

Analysis of Different Fate Modeling +
     Scoring
   a)  No fate analysis

   Most of the methods applied in life cycle assessment
until recently did  not consider the "fate behavior" of pollut-
                                                    45

-------
 ants, i.e., they assumed that all substances have the same
 fate properties ("critical volume", CML 1992, Ecolndicator
 95, etc.).This is clearly not a valid assumption, since routes
 of exposure and residence times vary widely (up to factors
 of 10,000).

   b)  Partial fate analysis: scoring

   The ED/Pmethod (Hauschild and Wenzel, 1997) performs
 a "partial fate" analysis. Fate is incorporated through a bio-
 degradability factor based on OECD biodegradability tests,
 varying from 0.2 for biodegradable substances to unity for
 persistent substances.This allows coverage of a large num-
 ber of substance, but does not satisfy the consistency re-
 quirements of fig. 4. Ranking methods based on  Persis-
 tence-Bioaccumulation-Toxicity information could be inter-
 esting to screen and select pollutants; however such coef-
 ficients cannot usually be directly multiplied by the mass of
 the environmental interventions, as required for LCA (Udo
 de Haes et at. 1999). It is suggested that this  information
 can be used to extrapolate the key parameters described in
 fig. 5 and appendix 1. Figure 8 presents how half-life in soil
 could be extrapolated based on OECD biodegradability tests.

   c)  Full fate and exposure analysis using the concept of
      exposure efficiency

   So far, determination of fate coefficients fortoxicity char-
 acterization in LCIA has mostly been based on the prin-
 ciple of fugacity modeling (Guineeetal., 1996; Hertwich et
 al., 1997; Goedkoop et al., 1998). Especially Mackay type
 three models have been used in several calculations of char-
 acterization factors for human and ecotoxicity (fig. 9 to 11).
 The applied models are rather comprehensive in the de-
 scription of the exposure pathways, but have the drawback
 of complexity linked to the integrated partitioning between
 the different compartments. Indeed, a bias may be more
 easily hidden in a more sophisticated approach. The num-
 ber of parameters and equations to be solved simultaneously
 make them very difficult to check against measured data or
 even to check their order of magnitude (Berding et al., 1999).

  Based on this concept of exposure efficiency, the com-
 parison between two fugacity multi-media models  is pre-
 sented: errors and default data can be quickly identified and
 corrected. Large variations of up to seven orders of magni-
 tude are observed on the overall exposure efficiency. The
 highest exposure efficiencies are observed for substances
 that are ingested through the food pathway, the exposure by
 inhalation being a factor smaller for these substances.

  Residence times and exposure efficiencies calculated by
 the EUSES and CALTOX models were compared (Jolliet et
 al, 1998, fig.  10 and 11). On the one hand, figures 12 and 13
 show that the use of default data for risk assessment can-
 not be used in LCIA, as best estimates and not safe values
 are required. On the other hand, once errors or default data
are corrected,  there is a good agreement in the exposure
efficiencies calculated by the two models, with  a variation
of up to 8 orders of magnitude (fig. 13). Interestingly, the
highest exposures are observed for exposure by food in-
gestion.
   d) an intermediary approach

   The modular approach described by figure 5 and appen-
 dix 1 (Jolliet and Crettaz, 1999) offers an alternative to fugac-
 ity modeling (fig. 14 to 16). Each compartment is modeled
 in an independent module and then connected to the other
 compartments by inter-media transfer coefficients. This
 enables each module to be validated against measurements
 or even to shortcut some parts using empirical data. Be-
 cause of its modular character, this approach also enables
 to take into account the exposure to daughter degradation
 product of a substance (appendix 1). Hauschild and Jolliet
 (1999) have recently developed the air module that predicts
 the atmospheric residence time and the fraction of the emis-
 sion deposited (fig. 17). Checked against a number of mea-
 surements, there was a good correlation between predicted
 and measured lifetimes in air (fig. 18: R2=0.85).The inter-
 mittent character of rain plays an important role and a simple
 model for intermittent wet deposition has been developed
 for steady state modeling (fig. 19).

 Effect Data for Human- and Ecotoxicity
   For the analysis of effect coefficients, Hofstetter (1998)
 recently proposed a first approach to determine Years of
 Life Lost or Years of Life Disabled for carcinogenics and a
 few other pollutants for which epidemiological data are avail-
 able. Crettaz et al. (1999) proposes to use a slope based on
 ED10, enabling a similar approach for carcinogenics and
 non carcinogenic substances and avoiding the use of con-
 servative estimates such as q1* for carcinogenics (fig. 20
 and 21). Extrapolation procedure are being developed and
 checked for substances with little information. Most of the
 conclusion for human toxicity remains valid and can be
 adapted to ecotoxicological damages, considering that hu-
 man toxicity aims at the protection of the individual whereas
 ecotoxicology aims at the species protection.

 Need for Sophistication
  Spatial differentiation can be introduced as modifiers of
the effect and fate factors (fig. 23 and 24), as a function of
geographical or background information. Rabl and Spadaro
(1998) provides a good example of exposure variations with
population density around the emission source or with stack
height. Differentiation is to be introduced only if significant
changes (e.g. more than a factor 5) are observed. If a fully
site-specific approach is necessary, an environmental im-
pact assessment is preferred to LCIA.

Conclusion and Consistency Criteria
  - Variations on exposure efficiency and fate are very large,
   up to 106 (6 orders of magnitude)

  - Consistency between effect and fate is a must! The
   choice of the level of the effect factor depends on un-
   certainties.

  - Exposure efficiency and overall residence times are
   useful concepts as an intermediary between modeling
   and scoring.
                                                     46

-------
  - Reasonable consistency between results of different
   models is observed, once errors are identified. Model
   comparison needs to be performed  in a systematic
   manner with checks against measurements.

References
Berding, V.; S. Schwartz; M. Matthies (1999). Visualization
   of the Complexity of BUSES. Environmental Science
   and Pollution Research 6:37-43.

Crettaz, P., Brand, K., Rhomberg, L. and Jolliet, O., 1999.
   Human health effects of carcinogenic compounds; an
   application for LCIA. AlChE Presentation Record, Oc-
   tober 99.

Evans, J.S., Thompson, K.M. and Hattis D., 1997. Expo-
   sure efficiency. Journal of the air and waste manage-
   ment association.

Goedkoop, M.J.,  P. Hofstetter, R. Muller-Wenk and R.
   Spriensma, 1998:The Eco-lndicator 98 Explained. Int.
   J. LCA 3 (6): 352-360.

Guinee, J., Heijungs, R., van Oers, L., van de Meent, D.,
   Vermeire, T. and Rikken, M., 1996. LCA impact assess-
   ment of toxic releases. Generic modelling of fate, expo-
   sure, and effect on ecosystems and human beings with
    data for about 100 chemicals. RIVM report, pp. 70.

Haushild, M.andWenzel, H., 1997. Environmental assess-
    ment of  products - Scientific background, Thompson
    science, Chapman & Hall, New York,  pp. 525.

Hauschild, M. and Jolliet, O., 1999. Determination of atmo-
    spheric residence times and transfer factors from air to
    soil and surface water. Prepared for Chemosphere.

Hertwich, E., McKone, T. and Pease W., 1997. Hazard and
    risk-based approaches to comparing toxic emissions.
    IEEE international symposium on Electronics and the
    Environment,  pp. 261-266.

Hofstetter, P., 1998: Perspectives in life  cycle impact as-
    sessment. A structured approach to combine models
    of the technosphere,  ecosphere and valuesphere.
    Kluwer, Boston, Dordrecht, London.

Jolliet, O., 1996, editor. Impact assessment of human and
    eco-toxicity in Life Cycle Assessment, in "Towards a
    methodology for Life Cycle Impact Assessment" (Udo
    de Haes ed.). SETAC, pp. 49-61.

Jolliet, O. and Crettaz, P., 1997: Calculation of fate and ex-
    posure coefficients  for the life cycle toxicity  assess-
    ment of air emissions. International  Journal of LCA 2
    (2): 104-110.

 Jolliet, O., Crettaz,  P.,  Hofstetter, P., Hertwich,  E. and
    Spriensma, R., 1998. Comparison and check of mod-
    els to include fate and exposure in LCIA. Abstracts of
    the SETAC-Europe annual conference, Bordeaux.
Jolliet, O. and Crettaz, P., 1999:Tox Mod: Modular Fate and
    Modelling of exposure efficiency for the characteriza-
    tion of human toxicity in Life Cycle Assessment. Int.
    Journal of Risk Analysis, submitted.

Rabl, A. and Spadaro, J.V., 1998 Estimates of real damage
    from air pollution: site dependence and simple impact
    indices for LCA. Abstracts of the SETAC-Europe con-
    ference Bordeaux, May 98.

Udo de Haes, H.A., O. Jolliet, G. Finnveden, M. Hauschild,
    W. Krewitt and R. Mueller-Wenk, 1999: Best available
    practice regarding impact categories and category indi-
    cators in Life Cycle Impact Assessment. Report of sec-
    ond working group on life cycle impact assessment of
    SETAC-Europe.

Appendix
A Modular Approach to Calculate Fate and
          Exposure Efficiency
  Jolliet and Crettaz (1999) also developed and presented
in detail an analytical and modular approach to calculate
exposure efficiency (fig. 5).The new model is developed on
the basis of  a modular approach that offers an alternative
way of simplification than the traditional Mackay I ,ll and III
classification:

   a) in a first step, exposure is calculated for each media,
     assuming that the concentration in other media is null.
     Three issues are successively addressed:

   a1)  From emission to concentration: assuming steady
       state, the increase in concentration can be calcu-
       lated as the initial emission flow per unit area multi-
       plied by the ratio of two key parameters: the overall
       residence time (T/", yr) divided by the volume of
       dilution per unit area (l/m, m3/m2) of the medium
       that contains the pollutant, per unit area:

   ,„-_*?•_.,ซ                     [kg/m3]      (D
   This ratio corresponds to the intuitive deduction that the
 longer the residence time, the higher the concentration in-
 crease. The overall residence time is a combination of the
 different removal processes: both the degradation of the
 substance and its removal by transfer to another medium
 on the one hand, and its advection on the other hand.

   a2) From concentration to direct intake: Direct intake is
       directly proportional to the time integrated concen-
       tration increase in air (resp. water and soil) multi-
       plied by the overall volume of air inhaled (resp. water
       or soil ingested) annually by humans per unit area.

                                     H          (2)
   Where P is the population density [pers/m2] and ^'"the
 yearly inta&e of air, water or soil particles per person and
                                                     47

-------
 per year [m3/pers-yr]. Depending on local characteristics
 (e.g. air emission in high or low populated area), exposure
 varies typically by a factor of 10 to 30 (Rabl and Spadaro,
 1998). This is relatively restricted compared to the total
 variation in exposure efficiencies of up to 8 orders of mag-
 nitude between different substances (fig. 13). It could there-
 fore be addressed through generic spatial differentiation (e.g.
 scenarios for emissions in high, mean and low populated
 areas).

   a3) Bioconcentration and indirect intake through the diet
      For all media, it is possible that an increase in the
      concentration in the media will lead to an increased
      concentration in food products (plants, fishes), that
      eventually will be ingested by humans. The resulting
      exposure efficiency through the diet is very similar to
      that for direct intake: only here, the daily intake of the
      surrounding medium is replaced by the food intake
      and multiplied by the bioconcentration factor:
           Tm Bf*E>mfo(*d
   ^=^^^PP2/0ฐrf       "           ฎ

  Where BCFimlo
-------
Jolliet's Overheads
       Human toxicity and eeotoxiclty
             Modelling vs scoring
            	mi	mil	i mam mom mm mm mm urn $
  1. Introduction

  2. Framework for fate and exposure in toxicity

  3. Analysis of different fate modelling + scoring

  4. Enabling experimental shortcuts and check
       + robust extrapolations

  5. Conclusion and consistency criteria


               Prof O.Jolliet (olivier.jolliet@epfl.ch)
         EPFL-Swiss Federal Institute of Technology-Lausanne
             Institute of Soil and Water Management
               IATE-HYDRAM, CH-1015 Lausanne
            Tel. +41 21 693 7011, Fax +41 21 693 70 84
                                           Challenge
                                          Tell me your results
                                        I will tell who paid you !
                               Urgent need for recognised and consistent
                                      impact assessment methods

                                   + address ISO 14042 requirements
                              (scientifically valid + environmentally relevant)

                              --> needs for a general framework for toxicity
                                             assessment.

                             No                   an                   !
  2.Framework for toxicity characterisation
          (SETAC working group on toxicity assessment 1996)
                               Fate factors required for a full
                                fate and exposure analysis
                           M
     E:      factor
    Toxicity measure
 F: fate factor
links emission-effect
neglected until 1994!
   Human
   toxicity:
     RFD,, [kg/pers-an]
     Reference dose
8,, [-] "exposure efficiency"
 = fraction of the emission
   taken in by mankind 01M.„
Effect
coefficient
Effect
variable
Fate
factor
required

Concentration
to
concentration
increase
ADI, RfD
Dose
Emission to
intake
dose
Acceptable
deposition rate
Deposition
rate
Emission to
deposition
     Fate: analysis of existing methods
                                    Modelling versus scoring
   Mo     No       only emissions, E=1, F=1


       effect: critical volumes or analog
   (Tellus 92, BUWAL, CML 92,): F = T,/ V = 1
   Compares an emission with a concentration limit!
   Pollutants with short lifetimes are overestimated
        I       EDIP method  (Haushild, Denmark),
   F=Biodegradibility factor equal to 0, 0.5 or 1;
   Problem of consistency
                            - Consistency between fate and effect is a priority

                            - Important to have coefficients multiplicative with
                            mass

                            - Simple ranking is not sufficient: in LCA it matters
                            if it is 5 times or 10'000 times higher

                            - Use Persistence - Bioaccumulation - Toxicity
                            informations to extrapolate key parameters in a
                            consistent fate & effect assessment
                                                   49

-------
       Full fate: Modelised approach
           fugacity modelling
	Inter-media  Complexity Accuracy
 1 Box model    No        low          low
 Mackay 3       yes        high         ???
 CML-RIVM, 1996:USES / Hertwich, 1997:CALTOX

 - inter-media transport included
 - beware default data,
 - little check against measured data
 - height of dilution fixed to 1000 m3/m2
                                   Fate modelisation: intermediary level
                                  Challenge: Simple and robust+accurate approach !

                                  idea: introduce a double level
                                  - Detailed modeiisation, description
                                   of physico-chemichal phenoma,
                                   laboratory experiments (>150 parametres)
                                  - intermediary modelling based on a limited number
                                  of key parameters, according to physics&chemistry,
                                  check at larger scale
   key parameters
   direct exposure
         key parameters
         food exposure
Human Toxicity Potentials (HTP)
     Method comparison
Method
Critical Surface-Time
CST air
CST soil-food
CST air-soil-food
CST 95
CML96VUSES "original
CML96/USES revised
Pollutant

Pb"
Pb"'
Pb"-'-'
NOx"
'NOx"
NOx"
Volume of
dilution
[m3/m2)

3500
0.2
(11% depos.)
890
1000
1000
Residence
time
[day]

11
200 to 2000 yr

1.1
180
1
exposure
efficiency
[-] .

6.8E-7
0.024 to 024
0.0028 to 0.028
2.5E-7
4.0E-5
Z2E-7
Hum.Tox.P.
HTP
[kgpb "/kg,]

1
MOO to 200 00
230 to 2M
0.002


               inter-media transfers
   Full fate: empirical approach

Critical surface-time
(Jolllot and Crettaz, 1994 and 1997):
                                        Modelling atmospheric fate:
                                  a modular approach to fate  modelling
                                 •••"'ffWWilKIIBWIVWrffffBl'M             BHHRtffM MKMM 90SSM JMffll WKS $&& ^ ^ VS J
 F, =
Ii
V,
                       Possible shortcut:
                       C  : measured
                     5 M, : estimated
— > overall response of the environment
At regional, continental or world level


ป%dop



orti'o

into
(Wx



prccipi
-2 10



)
1 	 y~
WBideppg
1
into precipitation

": '. -



.w
1C
v


llhiJ
ceptori



(plant*) •

                                            50

-------
   Intermittent wet deposition:
variations of concentration in air
                      'wet
                                                Residence times:
                                         check  with measured values
                                                           predicted residence time of pollutant in air as a
                                                              function of measured residence time
                                                                                         • tmeas
                                                                                  Mboo.oo
Comparison and check of models to
 include fate and exposure in LCI A
                                       Fugacity fate and exposure model
                                        CALTOX (UC Berkeley): risk assessment
Dr O.Jolliet, P. Crettaz
EPFL-Swiss Federal
Institute of Technology
Lausanne (CH)
P.Hofstetter      E. Hertwich   R.Spriensma
ETHZ-Swiss Federal  University of   Pre consultant
Institute of Technology California     Amersfoort (NL)
Zurich (CH)'-      UC-Berkeley (USA)
SETAC-Europe conference in Bordeaux, May 98



Comparison between CALTOX and EUSES

     - Residence time in air and water
     - Transfer factor between air and water
     - Overall human exposure efficiency
                                                                            Sediment
                                                   > 150 parameters,
                                                   Based on partition coefficient at equilibrium,
               EUSES (CALTOX)
           residence half-lives in air
                                          Exposure efficiency for air emissions
                                          Models comparison UC-Berkeley/ NL
loco).
S ,™
i Phenol(default) •
- t *T
:ฃ < jtf
ฃ aoioi ooioi oil om 'njjK^T
s ^^^

1 ^^-^ 	
o 4^--^ 	 ,
is _j Heavy metals *'
g \jtjt omnm
CALTOX Residence h

* -ff*^
ป3ซ"*r* ป
>^^*'ป
* 10 100 1000 10000

*^-1


tetrachloride: d'1
alfliteinalr(d)
                                                       On the basis of the EPFL intermediary model
                                                    NL
                                                                        Pb air-soil-food
                                                     5*0.00000010.000001 0.00001  0.0001 O.OOT\ 0.01
                                                     t
                                                     tn
                                                                   CALTOX exposure efficiency (-)
                                                                    UC Berkeley (USA)
                                                 51

-------
       Ecotoxicity characterisation
        (SETAC working group on toxlclty assessment:! 996)
            s,=
                        ,
    E: effect factor
  Toxlcity measure
 Eco
 toxicity:
   RFC,,[kg/pers-an]
   Reference concentration
 F: fate factor
links emission-effect
neglected until 1994!
S'=^5C'A
At — ^' * M
NECj '
      5C, , [-] "integrated
    concentration increase"
         Toxicity characterisation
        (SETAC working group on toxicity assessment:!996)
       ••"•••••••••••••••i mmm mmm mam mm mm BB n m iปi
                                        Spatial differentiation
          Toxicity score: S, = Ej Fr M,
           IE; effect factor!
           effect information
                background" information
               geographical Information  ;
                            In case of large variations:

                            F and E can  be spatially  differentiated  as a
                            function    of   geographical   or   background
                            information (water volume, indoor, etc.)

                            Spadaro, 1998: exsposure indices
                            Population density:     0.5 rural    1 avg  6 urban
                            Stack height:         15 (Om)    1 (10m) 0.6 (200m)

                            For a fully site-specific approach: EIA !
               Conclusions
                                             IXII
- Variations on exposure efficiency/fate are very
large, up to 10s (6 orders of magnitude)
- Consistency between effect and fate is a must!
The choice of the level of the effect factor
depends on uncertainties.

- Exposure efficiency and overall residence times
are useful concepts as an intemediary between
modelling and scoring

- Reasonable consistency between results of
different models, once errors are identified.
Models comparison needs to be performed
systematic., with checks against measurements^
                                                 52

-------
                   Priority Assessment of Toxic  Substances  in  LCA
                                  - A Probabilistic  Approach  -


                                      M. Huijbregtsi and L. Reijnders
                                Interfacuity Dept. of Environmental Sciences
                                         University of Amsterdam
                                       Amsterdam,The Netherlands

                                         U.Thissen and A. Ragas
                                      Dept. of Environmental Studies
                                          University of Nijmegen
                                        Nijmegen,The Netherlands

                                       T. Jager and D. van de Meent
                           National Institute of Public Health and the Environment,
                                        Bilthoven,The Netherlands


'Address correspondence to: Drs. M.A.J. Huijbregts, Interfaculty Department of Environmental Science, Faculty of Environmental Science, University of Amsterdam, Nieuwe
Prinsengracht 130, NL-1018 VZ, Amsterdam, The Netherlands, tel. ++31 525 62 63, e-mail: m.huijbregts@frw.uva.nl
1.   Model development
  In the LCA impact assessment, toxicity potentials are
used for a weighted aggregation of the potential impacts of
a large number of toxic pollutants into one toxicity score.
Up to now, toxicity potentials were calculated with multi-
media fate models, using unrealistic system settings to meet
LCA requirements. Some of these unrealistic system set-
tings can be overcome by modelling a larger part of the
world, i.e. by using the nested multi-media fate and expo-
sure model USES 2.0. This model has four scales, which
are a local, a regional, a continental and a global scale (Fig-
ure 1).The latter consists of three parts, reflecting arctic,
moderate and tropical geographical zones of the Northern
Hemisphere.

  Using this model, toxicity potentials were calculated for
the impact categories human toxicity, aquatic ecotoxicity,
terrestrial ecotoxicity and sediment ecotoxicity after initial
emission to the compartments air, fresh water, seawater,
industrial soil  and  agricultural soil,  respectively. Because
the model was originally developed for risk assessment
Figure 1: Schematic representation of the scales in USES 2.0.
                                                     53

-------
purposes, a number of LCA-specific model changes were
implemented in USES 2.0. The most important model
changes are:

  (1) Discarding the local and regional scales. In current
life-cycle inventories, emissions are summed up per pollut-
ant type regardless of the spatial  context of these emis-
sions. This results in an inventory outcome lacking any re-
trievable relation with a particular region. As a consequence,
the local and regional scales in USES 2.0 are inappropriate
for the computation of toxicity potentials;

  (2) Excluding 'economic' processes. In LCA there is a
strict distinction between economic processes, which should
be taken into account in  the inventory analysis, and the
impact assessment in which the potential impact of these
emissions is estimated. Therefore,  economic processes in
USES 2.0, such as waste water treatment and emission
estimates, are switched off in the model calculations;

  (3) Standard emissions at the continental scale.The built-
in emission estimation procedures of USES 2.0 were not
used, but standard emissions to the different environmen-
tal compartments (air, fresh water, seawater, industrial soil
and agricultural soil, respectively) were implemented at the
continental scale. When using continental standard emis-
sions in the computation of toxicity potentials, it is implic-
itly assumed that the accumulated emissions related to the
complete life-cycle of a product system, i.e. the outcome of
a life-cycle inventory,  (i) all take place in Western Europe
and (ii) are uniformly distributed over the initial emission
compartment;

  (4) Aggregation of risk characterization ratios (RCRs) at
the continental and global scales. RCRs, also known as
PEC/NEC-ratios, are calculated by dividing the estimated
exposure level with a no-effect parameter for that particular
substance. RCRs form the basis of the calculation of toxic-
ity potentials. USES 2.0 gives the possibility to compute
RCRs on a local and regional scale. However, these scales
are not taken into account in  the computation of toxicity
potentials. Consequently, RCRs are computed on the conti-
nental scale, and the three global scales for every impact
category. But one is still left with the problem that several
RCRs are computed for the same impact category. Compu-
tation of one toxicity potential per initial emission compart-
ment for the four impact  categories can only take place
after aggregating the RCRs of the different environmental
scales.This aggregation was performed with weighting fac-
tors resulting in a weighted RCR for each toxicity category.
The RCRs of the aquatic compartments were aggregated
on the basis of the compartment's volume, and the RCRs
of the sediment and terrestrial compartments were both
aggregated on the basis of the compartment's mass. For
humans, the human population present at a certain scale
has been used as a weighting factor. Thus, per impact cat-
egory the  weighted RCR of a substance after a standard
emission to a certain compartment is calculated by
  Weighted RCR = S
          (PECxW)

             NEC
                                                 (1)
in which W is the compartment-specific weighting potential;

  (5) Introduction of a reference substance and reference
compartments. It is chosen to use a reference substance
in the calculation of toxicity potentials which follows the
established use of carbon dioxide (CO2), ethylene (C2H4),
and chlorofluorocarbon (CFC11) for evaluating global warm-
ing, photochemical ozone formation and stratospheric ozone
depletion, respectively. 1,4-Dichlorobenzene (1,4-DCB) is
taken as a reference substance in the calculation of toxicity
potentials. Furthermore, a reference emission compartment
per toxic  impact category is chosen: the air compartment
for human toxicity, the fresh water compartment for aquatic
ecotoxicity and sediment ecotoxicity, and the industrial soil
compartment for terrestrial ecotoxicity. Toxicity  potentials
(TP) are calculated by dividing the aggregated RCRs of a
substance after emission to a certain compartment with
the aggregated RCRs of the reference substance after emis-
sion to the reference compartment:
  TPx=
 WeightedRCRx

Weighted RCR ref
                                                (2)
resulting in 20 toxicity potentials per substance (see Table
1).

2.   Probabilistic Approach
  Apart from the possibility to calculate toxicity potentials
in a deterministic way, it is also possible to follow a probabi-
listic approach in which the operational uncertainty in the
toxicity potentials is quantified by specifying the uncertain-
ties in the input parameters and performing Latin Hypercube
sampling. To save time and money in collecting data for
Table 1. Type of Toxicity Potentials per Substance
Emission compartment
Impact category Air
Aquatic ecosystems AETPoir
Terrestrial ecosystem TETPปir
Sediment ecosystem SETP^
Humans HTP..,
Fresh
Water
AETP|W
TETP,,
SETP|W
HTR,
Sea
Water
AETPsaa
TETPaea
SETPSBป
HTP
Industrial
Soil
AETp,3
TETPfe
SETPr3
HTP.
Agricultural
Soil
AETPป
TETPM
S^P.3
HTP
                                                     54

-------
defining density distributions, only the most important pa-
rameters may be considered in detail. Those parameters
can be identified by a preliminary uncertainty importance
analysis. In the preliminary analysis, the uncertainty distri-
butions of all parameters are estimated only roughly, with
the aim to determine what input parameters contribute most
to the uncertainty of the output parameters.The parameters
that appeared to contribute insignificantly to the output un-
certainty can be replaced by their most likely values. From
our preliminary analysis, it appeared that in general uncer-
tainty in substance-specific input parameters from the ref-
erence substance and environmental characteristics will not
contribute significantly to  the uncertainty of LCA toxicity
potentials. Consequently, only substance-specific input pa-
rameters of the substances under study and human char-
acteristics should be specified in more detail. This can be
done in  a second more detailed uncertainty analysis,  re-
sulting in frequency plots of the output parameters repre-
senting the operational uncertainty of the toxicity poten-
tials.

  To simplify the implementation  of the uncertain toxicity
potentials in LCA case studies, the sampled distributions
may be fitted to standard uncertainty distributions, such as
Lognormal and Gamma distributions. The uncertainty in the
toxicity potentials may be represented by an uncertainty
factor k which is defined such that 95% of the values of a
stochastic variable (X) is within a factor k from the median
M(X) of a skewed distribution:
   V.
                       =0.95
                                                 (3)
  As these uncertainty distributions are to be used in LCA
case studies, correlations between the toxicity potentials
should also be known. The toxicity potentials of one sub-
stance may be correlated, because they all stem from the
same uncertain substance-specific parameters.
3.   Discussion
  Although both method and input data were improved in
comparison to former research, still a number of further
improvements should be addressed. First, model improve-
ments in USES 2.0 itself are needed. For instance, the
groundwater compartment is not implemented in the model
as a full compartment. LCA-specific changes in the model
structure may also have an effect on the validity of the model
results. One item for improvement relates to the geographi-
cal choice. An important assumption in the calculation of
toxicity potentials is that all life cycle emissions take place
in Western Europe. Of course, this is a simplification of
reality because emissions in many product life cycles take
place partly outside Western Europe. Another problem of
using Western Europe as the initial emission compartment
is that, in this way, all emissions are assumed to be homo-
geneously distributed  over Western Europe, not allowing
computation of specific toxicity potentials per West Euro-
pean  country. Finally, the aggregation of RCRs related to
aquatic and terrestrial toxicity take place on the basis of
volumes and weights of the related compartments, respec-
tively. Other weighting methods, for instance, on the basis
of species density  per compartment may result in other
toxicity potentials for aquatic, terrestrial and sediment eco-
systems.

Huijbregts' Biography
  Huijbregts studied  Environmental Science and  Health
Science at the University of Nijmegen from 1990 to 1996. In
1996, he started his Ph.D. research at the Interfaculty De-
partment of Environmental Sciences from the University of
Amsterdam.The main goal of this research is to perform an
uncertainty analysis for the Life Cycle Assessment of low-
energy options applied in Dutch one-family buildings. Re-
cent publications are: Huijbregts, M.A.J., 1998. Application
of uncertainty and variability in LCA. Part I: A general frame-
work for the analysis of  uncertainty and variability in life
cycle  assessment.  Int. J. LCA 3 (5): 273-280, and Part II:
Dealing with parameter uncertainty and uncertainty due to
choices in life cycle assessments. Int. J.  LCA 3 (6): 343-
351
                                                      55

-------
Huijbregts' Overheads
  Priority Assessment of Toxic Substances in LCA

               A Probabilistic Approach
              M. Huijbregts & L Reijnders (IDES-UvA)
         U. Thisssn & A. Ragas (Dep. Env. Studies, Nijmegen)
              T. Jager & D. van de Meent (RIVM-ECO)
                  Brussels, 30 November 1998
       1.  Introduction




       Why Toxicity Potentials in LCA?

       •   Weighted summation of toxic emissions

       .   Toxic Impact categories
           (1) Humans
           (2) Aquatic ecosystems
           (3) Terrestrial ecosystems
     How are current Toxicity Potentials calculated?

     •    Fate & Effects are taken into account
          (Guinea et a/., 1996; Hertwich ef a/., 1998)

     •    Use of Reference Substance
 Improvements needed in current approach?

 •   Avoiding the artificial change of system parameters

 •   Taking into account all relevant compartments

 •   Usage of realistic values instead of defaults
       Goal
           Improvement of model structure

           Show effect of data uncertainty and variability
                                                            2.  Methods
Uniform System for the Evaluation of Substances
2.0
                                                                Global nested multi-media fate model (Simplebox 2.0)

                                                                Human exposure model

                                                                Risk Characterisation Ratios
                                                        56

-------
Adaptations needed for LCA purposes

•    Regional and local scale are turned off

•    Economic processes are turned off

•    Standard emission at continental scale

•    Weighted summation of RCRs per impact category

•    Use of reference substance and compartment
                                                                              Calculation of Weighted Risk Characterisation Ratios
                    Terrestrial PEC/PNEC
                    • Western Europe
                      ,- Natural soil
                      -Agricultural soil
                      -Industrial soil
                    Moderate zone
                    Arctic zone
                    • Tropic zone
Human PDI/HLV
Western Europe
Moderate zone
Arctic zone
Trope zone
   Probabilistic simulation


   •    Sampling method: Latin Hypercube

        - Substance-specific input data

        - Human characteristics

        - No system parameters


   •    Importance analysis: Spearman rank correlation


   •    Covariance matrix: Spearman rank correlation


   •    Separate treatment of Reference Substance
 Weighted Aquatic RCR after emission to fresh water

                       2,3,7,8-TCDD
Proba
.229
.171
.114
.057
.000
bllity fce^
1

1
IlliB,1iA,w.i-.— 	 	
ency
1143
857
571
285
O.OE+0 6.3E-3 1 ,3E-2 1 .9E-2 2.5E-2
  Median = 1E-3
                        Mean = 2.4E-2
                                               = 2.3E+3
                     1,4-Dichlorobenzene
Proba
.118
.088
.059
.029
.000
Weighted Aquatic RCR after emission to fresh water
biiity FreclL
1
1

IL^...,
ency
588
441
294
147
0
O.OEtO 2.3E-7 4.5E-7 6.8E-7 9.0E-7
Weighted Terrestrial RCR after emission to indust soil

                      2,3,7,8-TCDD
                                                                      Probability

                                                                       .033f
                                                                                                               Frequency
                                                                          O.OE+0      1.3E+1      2.5E+1      3.8E+1     S.OE+1
  Median = 4E-8
                         Mean ='
                                                 = 4.9E+1
                                                                     Median = 7.0
                                                                                            Mean = 13.7
                                                                                                                   k = 1
                                                              57

-------
Weighted Terrestrial RCR after emission to indust soil

                     1,4-Dichlorobenzene
Probability Frequency
(•
i'
i!
i "ป
*
! .072
035
000



t 	
IBKiUbu*...... . 	
, , ...'•. -.-^-, 	 , 	 j_L
0.0&0 3.8E-3 7.5E-3 1.2E-2 1.5E-
717
537
358
179
0
2
                        Mean = 2.7E-2
                                              k=1.7E+6
   Weighted Human RCR after emission to air

                    1.4-Dichlorobenzene
Prol
.138
,103
.069
.034
.000
wbilily Frequency



L 	 	
t 	
S^Hi&'iM.ซ4...... j . t, 	
r ' • ' ™~TJ" 	 ^
O.OE+0 2.5E-5 5.0&5 7.5E-5 1.0E-
688
516
344
172
0
4
                       Mean = 1.5E-4
                                               = 7.8E+3
  Covarlance matrix

A
AIR T
H
A
FW T
H
SEA A
T
H
A
AGR T
H
A
IND T
H
AIR
ATM
1
1
1
1.0
0.7
0.9
1.0
0.7
0.9
0.9
1.0
0.9
0.9
1.0
0.8
FW
A T H

1
1
1
1.0
1.0
1.0
0.9 0.8
0.8
0.9 0.8
0.9
SEA
ATM


1
1
1
0.9
0.8
0.9
0.9
AGR
A T H



1
1
1
1.0
1.0
0.8
    Weighted Human RCR after emission to air


                     2,3,7,8-TCDD
Potability R-equency
.171
.128
.035
.043
000



L
•IIIIISl.l.u,,,i.,. 	 „..„„„.„ „
lUUU > ' 	 '' ' 4 '
O.C&0 6.3E& 1.3E+7 1.9E+7 25E+
854
640
427
213
0
7
                         Mean = 1.2E+7
                                                 1.3E+3
Uncertainty importance
Aquatic ecotoxicity
Fresh water emission
Parameter Importance
Koc 63%
AFIab-field 24 %
Kdeg_sed 6 %
Kdeg_water 3 %
VP 3 %
TOTAL: 99 %
Terrestrial ecotoxicity
Industrial soil emission
Parameter Importance
Kdeg_soil 75 %
AFIab-field 24 %



TOTAL: 99 %
Human toxicity
Air emission
Parameter Importance
AFJnter 72%
BAF_meat 12 %
Kdeg_soil 5 %
VP 4 o/0
AFJntra 3%
TOTAL: 96 %
                                                                   Discussion & Conclusions


                                                                   Validity of results



                                                                   •   Model structure is more in line with LCA



                                                                   •   Uncertainties in model structure are still important
                                                                      Standard emission at West European scale is not fully
                                                                      in line with LCA
                                                                      Simple weighting factorsare used within impact
                                                                      categories
                                                                      Defining input uncertainties is uncertain itself
                                                        58

-------
Significance of results

•    Human variability and system parameter uncertainty
     are not of significant importance
     Only a few substance-specific parameters are
     responsible for uncertainty in Toxicity Potentials
     Correlations between Toxicity Potentials are large for
     the same substance and impact category


     Care must be taken with implementation of Toxicity
     Potential uncertainties in LCA case studies

     - Reference substance
     - Correlations
     - Overlap in product comparisons
                                                                 59

-------
                            Implications  of  Inventory Structure for
               Life  Cycle Impact  Assessment and  Uncertainty  Analysis
                                              Gregory A. Morris
                                            Sylvatica, Maine, USA.
   Site-specificity in characterization within Life Cycle Im-
 pact Assessment (LCIA) is a highly active area of current
 research and development.This level of activity may largely
 be in  response to two factors: (1) past criticisms of the
 simplicity of generic or "global" methods for characteriza-
 tion analysis, and (2) empirical information which indicates
 that location can contribute to multiple orders of magnitude
 differences among appropriate characterization factors, for
 emissions emanating from a given site.

   However, the emissions within an LCA inventory ema-
 nate from many sites, not one. The important question  is
 how many. If only a few sites contribute the bulk of emis-
 sions for most products' life cycles, then site-specificity  is
 essential and feasible. If on the  other hand, scores of sites
 contribute significantly to inventory emissions totals for most
 products, then  site-specificity is  less feasible and much
 less necessary. Thus, the need  for location-based charac-
 terization factors, and the uncertainty reduction benefits
 which  result from applying locational information in LCIA,
 depend critically upon an  unexamined issue: the structure
 of the emissions inventory.

   By "structure" we mean a few related attributes of the
 inventory emissions. First, of importance is the minimum
 number of sites whose emissions must be taken into ac-
 count in order for an analysis (such as an LCIA) to address
 high percentages (e.g., 80% or 90% or 95%) of the total life
 cycle emissions. More generally, the shape of the ranked
 cumulative distribution of emissions is important and of in-
 terest.

  The above factors influence the spaf/a/distribution of the
 emissions, which, together with the regional variability in
 appropriate characterization factors, ultimately determines
 the uncertainty reductions which accrue from site- or re-
 gion-specific LCIA. How geographically dispersed are  the
 life cycle emissions? On the one hand, we want to be able
to address this question for specific life cycle inventories.
 At the same time, it would be highly valuable to gain some
 insight about the answers to these questions about inven-
 tory structure for life cycle inventories in general, to help
 guide the development of LCIA in general.

   This presentation summarized results of empirical inves-
 tigations into the issues  outlined above. First, a method
 and database were presented for addressing the dependence
 of terrestrial (non-oceanic) deposition of sulfur and nitrogen
 upon the geographic location of SO2 and NOx emissions
 sources. A modified characterization factor that accounts
 for predicted terrestrial deposition was introduced. Next, the
 influence of the number of emissions sites in an inventory
 upon the expected inventory characterization error was simu-
 lated and presented, under the simplifying assumption of
 equal emissions from all inventory sites, and random geo-
 graphic distribution of sites. Following this discussion, an
 input/output LCA system for the USA was briefly described,
 as a means for characterizing aspects of the inventory struc-
 ture in general, across products. Finally, the presentation
 summarized the implications of the concepts for desired/
 required levels of specificity in life cycle  impact assess-
 ment, and for uncertainty analysis and decision making in
 LCA.

 Morris' Biography

   Norris has been active  in Life Cycle Assessment for 5
 years. He founded Sylvatica, an LCA-related consultancy
 located in Maine, USA. He conducts research in LCA and
 Industrial Ecology-related topics for several US Federal
 Agencies, and applies LCA  on behalf of private corpora-
 tions and non-profit research institutes. Norris is adjunct
 professor with the Complex Systems Research Center at
 the University of New Hampshire, and director of US  Op-
 erations for the international Athena Institute, which devel-
 ops LCA databases and decision support software for build-
 ing designers and architects. He has degrees in Engineer-
ing and Natural Resources,  from MIT, Purdue University,
and the University of New Hampshire.
                                                     60

-------
Morris' Overheads
             Specificity in LCIA:
     Towards a Value-of-lnformation Approach
                  Gregory A. Norris
         Sylvatioa/ Univ. of New Hampshire / Harvard Univ.
                       USA
 Our Conference Organizers are to be
 Thanked and Congratulated for Posing
 the Question:
  "What levels of sophistication are
   needed, desirable, and useful
   in Life Cycle Impact Assessment?"
      To answer this question:

      B Requires a quantitative assessment of the
        uncertainty reduction payoff in results
        achieved by increasing sophistication
        :  requires LCA with uncertainty analysis as a
          function of modeling/methods choices
          what matters is clarity of decision support as a
          function of choices - reliability or confidence in
          conclusion that "A < B"
      m If done well will also point to places in
        analysis where uncertainty reductions yield
        biggest gains in clarity of decision support
A main message
   LCI
           Char.
                    Valuation >—*>•  Decision
  Most guidance to our search for
  appropriate levels of sophistication must
  come from context of full analysis
  Can't make much headway when LCIA
  is isolated from LCI and Valuation
                                                     Specif really...
     "you've got to look at the whole system... "


     m not this claim again!

     ซ"Prove it."
     • "And Be Specific."
     ซ"With real examples, real data."
             FOR ONE THING:
m Clearly, site-specific factors may dramatically
  increase precision when characterizing site-
  specific emissions.
B But do they matter for the LCIA results as a
  whole?
m Will they increase the clarity of decision
  support?
a Will they appreciably narrow confidence
  intervals on [A-B]?
                                                 61

-------
 Specifically...
    Suggest that at least these factors matter
   when searching for the desired level of soph.:
 i The structure of the inventory system
 2 Uncertainty in the LCI results
 3 The appropriate modeling, and our ability to
   model, regionally of inventory emissions
 4 Decisions about valuation
   Examples drawn from: Regionalized CFs
   taking account of SOx and NOx transport,
   deposition, and concentration effects in USA
  Example Starting Point
  • Tracking and Assessment Framework
  B US DOE Integrated Assessment, 1996
  • Valuation of impacts of reducing sulfur
   and nitrogen pollution emissions
  m Human respiratory, visibility, and limited
   ecological effects (aquatic toxicity in a
   few example lakes --> recreation)
 Source/Receptor Matrices

 • Basis: Lagrangian ASTRAP model, validated
   against Eulerian RADM model
 • Translates: seasonal emissions of NOx and
   SO2 per US (source) state (KT/season)
 • Into:
     Seasonal wet & dry dep/ (KT S, N per (receiving)
     US state, 10 Canadian prov., northern Mexico;
     Delta avg. seasonal ambient concentrations of
     nitrogen and sulfur species (mg/m3)
 • Also characterizes climate variability
    (...to model of uncertainty* resulting from
     using generic characterization factors
          when emissions occur at
    randomly selected (i.e., unknown) sites)
     * Expected characterization prediction error
 The Multiplicity Question(s)

• How many sites,
  with how much geographic dispersion,
  contribute significantly to inventory totals?
• What are the expected shapes of these
  distributions?
  We would like to know these answers
  for LCI's in general - that is,
  for products' inventories in general.
 The Multiplicity Questions)


• CHALLENGE 1:
  The results will be strongly influenced by
  the boundaries of the LCI - need very
  comprehensive LCIs.
• CHALLENGE 2:
  There are a lot of products.
                                            62

-------
US Input/Output-Based Upstream LCI:
 500 sectors consuming goods and services from other
  sectors, emitting pollution, and consuming resources
US Input/Output-Based Upstream LCI:

ซ Consumed inputs are measured in $:
  include materials, fuels, intermediate
  products, equipment, services
m Allocation among products on $ basis
n Pollution emissions in kg/$ of output
ซInput/output data result from census
m Annual sectoral emissions inventories
 Elements of the
 Input/Output View of Multiplicity
 • No boundary truncation
 s Multiple product outputs, modeled as 1
   * not all inputs may be used for given product
   .- many inputs used for more than one product
 ** Each input flow modeled as 1 flow
   *< net result: likely to be undercounting sites
 *ป Model each input as single-supplier
 • Can re-do, supressing inputs of services, etc.
       (to model of tierwise emissions)
                                a
      What does the distribution of
  emissions within each tier look like?
      (to model of site-by-site pollution
            within each tier -

                next time!)
   Two final, very brief, topics

   • Correlation and making distinctions
   • A "Value Space" heuristic
                                           63

-------
Correlation, Canceling, and
Confidence about Differences
Cannot draw conclusions by comparing
  PDFs or CDFs!  Below could imply...
       B
...either of the following:
                                                           A-B
                         0
                                                                              A-B
                                               Must Simulate the Difference!
                                               Requires simulated model to contain
                                               information on correlations and/or their
                                               structural causes.
Value of Information Analysis

• How might we apply uncertainty in LCA
  decision making
• How do we treat valuation and
  uncertainty jointly for decision support
• Which uncertainties matter most
• What level of sophistication is needed
The "Value Space" as a heuristic for VOI
and distinguishing uncertainties from
decisions/values
  V2/V1
             0.1  1     10   V3/V1

    95% CI's for S's lower bound
The "Value Space" as a heuristic for VOI
and distinguishing uncertainties from
decisions/values
Example with higher uncertainty
  V2/V1
                    90%
                                               V2/V1
                 r2  o
             0.1  1    10

    Confidence isoclines for S>0
                          V3/V1
             0.1   1    10

   95% CI's for S's lower bound
                                                                      V3/V1
                                        64

-------
Example with higher uncertainty
         "Confidence isoclines" for S>0
  V2/V1
                50%  80%
              0.1   1    10
                             V3/V1
Example with higher uncertainty
        "Confidence isoclines" for S<0
                                                    V2/V1
                                                                  50%
                                                                0.1   1     10
                                                                              V3/V1
                                             65

-------
                                      Levels  of Sophistication in
                        Life Cycle  Impact  Assessment  of Acidification
                                        Jose Potting and Michael Hauschild
                                         Institute for Product Development,
                                   Building 424,Technical University of Denmark
                                                  DK-2800l_yngby
                                                     Denmark
                                             Telephone: +45-45-254677
                                                Fax:+45-45-935556
                                              Email: JP@IPT.DTU.DK
 introduction
   The lack of spatial differentiation in current life cycle im-
 pact assessment affects the relevance of the  assessed
 impact. A framework has been developed for how to arrive
 at factors that relate the region of emission to the acidifying
 impact on its deposition areas. Next, such factors are es-
 tablished for 44 European regions with help of the RAINS
 model.

   RAINS is an integrated assessment model that combines
 information on regional emission levels with information on
 long range atmospheric transport in order to estimate pat-
 terns of deposition for comparison with critical  loads1 for
 acidification2.

   Application of the acidification factors in life cycle impact
 assessment is straightforward. The only additional data re-
 quired, the geographical site or region where an emission
 takes place, is in general already  provided by current life
 cycle inventory analysis, or by the scoping of the product
 system. The use of the acidification factors add resolving
 power of a factor thousand difference between highest and
•As a matter of fact, RAINS works with critical load functions, giving all combina-
tions ol sulphur and nitrogen deposition above which ecosystems are at risk to be
damaged (see Posch et al. 1995). For the sake of clarity, the term "critical load" is
used in this paper in stead of the correct term "critical load function".
"RAINS contains projections of the actual emissions for 44 regions covering all of
Europe, including the European part of the form Soviet Union. The model first calcu-
lates atmospheric dispersion and depositions from the emissions of all European
over the 612 grid elements (150* 150km) that together form the European grid.
Next, Ihe total deposition in each grid element is compared with the critical loads
for all ecosystems within that grid element (hence that several grid elements distin-
guish more than several thousand ecosystems; Posch et al. 1995). In this way, the
area ol unprotected ecosystem is determined per grid element (i.e., the area of
ecosystems where critical loads are exceeded). Finally, the model accumulates
over all grids resulting in the total area of unprotected ecosystems over the full
European grid. These features of RAINS are used to calculate acidification factors
that quantity the chance in area of unprotected ecosystems caused by the change
in the emission of one region (while background emissions and resulting deposi-
lions from other regions remain the same). See Potting et al. (1998"'b) for further
clarification.
 lowest rating, while the combined uncertainties in RAINS
 are cancelled out to a large extent in the acidification fac-
 tors due to the large area of ecosystems they cover. The
 information gained  by the use of these factors thus com-
 pensates fully for the additional uncertainties they introduce.
 The framework presented  is suitable to  establish similar
 factors for eutrophication and tropospheric ozone creation.
 RAINS also covers these impact categories.

  An extensive description of the developed framework, and
 the achieved acidification factors are given in Potting et al.
 (1998a'b). Since these publications cover also quite well the
 presentation at the  workshop, the presentation is not fur-
 ther summarized here. Rather, some main discussion points
 in relation to site-characterization are raised and commented
 upon here.

 Different Levels of Sophistication
  Due to lack of time, it was impossible to elaborate in the
 presentation on the different levels of sophistication allowed
 by the acidification factors of Potting et al. (1998a-b). Since
 these  publications do not discuss this issue  either,  it is
 addressed in more detail here.

  The level of sophistication in impact assessment can, as
 mentioned a few times during the workshop, be understood
 in two ways (see also Potting et al. 1997b):

  1. The extent to which relevant parameters in the causal-
     ity chain are taken into account in the characterization
     factors (i.e. whether the characterization  factors are
     based  on  no, some or full fate and exposure model-
     ing).

  2. The extent to which spatial (and temporal) variation is
     allowed in each parameter of the modeling underlying
     the characterization factors

  The  acidification factors from Potting et al. (1998a'b) are
sophisticated in both senses. They cover all the relevant
                                                         66

-------
parameters in the causality chain, and they allow a high
degree of spatial variation. Next Sections will discuss other
types of acidification factors with a similar level of sophisti-
cation. This Section will go into the limitations  in spatial
differentiation in the factors from Potting et al. (1998a'b).

  As already mentioned, application of the acidification fac-
tors from Potting et al. (1998a'b) in life cycle impact assess-
ment is quite straightforward. Each emission is multiplied
with the acidification factor for the relevant substance and
region. Next the product from all emissions times acidifica-
tion factors are summed-up to arrive at the total acidifying
impact from the analyzed product.

  Application of the acidification factors from Potting et al.
(19983ib) requires data additional to current impact assess-
ment: The  geographical site or region where an emission
takes place. The requirement of this additional data is often
put forward as an objection against spatial differentiation.
However, the geographical site or region where an emission
takes place is  in general already provided by current life
cycle inventory analysis (since this data is needed to cal-
culate the interventions from transport). Nevertheless, this
spatial differentiation may not always be possible or de-
sired. In that case, the acidification factors from Potting et
al. can be  used also to establish a site- or region-generic
default per substance. Such default can for example be
based on the European average acidification factor. Though
these defaults avoid spatial differentiation, they do cover all
relevant parameters in the causality chain (in contrast with
acidification factors based on the potential of substances
to release  hydrogen ions). An additional advantage is that
at least the uncertainties posed by refraining from spatial
differentiation are known.

  The acidification factors of Potting et al. (1998a>b) do ac-
count quite accurately for regional dispersion, but local dis-
persion is modeled rather poorly in RAINS3. There are mod-
els available that estimate regional dispersion similar as
RAINS, but account in a more sophisticated way for local
dispersion. Better modeling of local dispersion probably
contributes little to the accuracy of the acidification factors,
however, since the first few kilometers  only have a small
share in the total impact from long-lived substances as sul-
fur and nitrogen. In addition, such additional local modeling
would increase considerably the data demands from inven-
tory analyses. Nevertheless some life cycle assessments
do allow (and also desire) such high level of spatial differen-
tiation as discussed in Potting et al. (1999). In those cases,
the more sophisticated modeling of local dispersion can be
used in addition4.

  It would be interesting to investigate what additional infor-
mation different degrees of spatial differentiation provide,
and for which application they are relevant (see also the
presentation of Norris elsewhere in these proceedings).

Basis for Spatial Characterization
  The RAINS model also contains the data to establish
impact factors for regional ozone formation and for terres-
trial eutrophication from atmospheric depositions. A similar
framework as for the acidification factors can be followed
for these impact categories. Results will be published in
Potting and Hauschild (forthcoming). These three impact
categories have some general features in common that
actually provide the basis for spatial differentiation.

  As we emphasized in Potting et al. (1998a'b), Potting and
Hauschild (1997a'b), and several presentations, many present
environmental problems are characterized  by the fact that
the exposure of any receptor is the result of the long dis-
tance transport of emissions from very many sources. The
background exposure of one single receptor results thus
from multiple sources (far more than hundred thousand5).
The other way around, one single source contributes to ex-
posure of very many ecosystems (also far more than hun-
dred thousand6). This means that a single source (thus in-
herently also an emission per functional unit) contributes
only very little, in most cases marginally to the background
exposure of any receptor. This means also that the time
behavior of the emission from a single source (i.e. whether
it is a flux or a pulse) is hardly important. The temporal
variation of the very small contribution from our single source
emission will namely be cancelled out anyway against the
very high background exposure from all sources together.

  These are the characteristics that provide the possibility
to establish site-dependent impact factors that with consid-
erable accuracy estimate the given impact over the whole
European grid posed by a single source located in a given
region. To establish these impact factors,  one needs an
assessment model like RAINS that integrates information
on regional emission levels with atmospheric transfer-ma-
trices and threshold evaluation.

Uncertainties in Threshold Information
  A hot item at the workshop was the use of threshold infor-
mation. Finnveden showed in his presentation (see else-
3The dispersion estimates in RAINS are based on trajectory modelling. A trajectory
model assumes emissions to disperse homogeneously within an air parcel and next
follows this air parcel on its way of atmosphere motion. While the assumption of
homogenous dispersion not affects the accuracy in the medium to long range, it
does lead to an underestimate of concentrations within the first few kilometres from
a source. That is because the course of the concentrations from an emission will
rapidly increase and decrease again within these first kilometres, after which it
continues to decrease more gradually. More accuracy in the first kilometres can be
obtained by Gaussian plume modelling. There are models available that integrate
Gaussian plume modelling with trajectory modelling.
••Hence that critical loads are available on rather detailed scale (see previous foot-
note), which facilitates evaluation of deposition patterns with high spatial resolu-
tion.
5This follows logically from the breakdown of total deposition in one grid element
into the separate contributions from each region (RAINS is able to provide this
information). The emission of each region (usually similar with countries) is on its
turn the total from all small and big sources individually in that region.
6Relative long-lived substances as sulphur and nitrogen distribute over several
thousand kilometres. Depositions from a particular source occur thus basically
over the whole European grid. The number of ecosystems distinguished over the
full European grid amounts to almost 700,000 in Posch et al. (1997).
                                                          67

-------
 where in these proceedings) that uncertainty in threshold
 information could be:

   • A factor two due to parameter uncertainty (see his ex-
     ample about critical loads for acidification), and

   • Up to a factor million due to differences in the chosen
     endpoint and/or way of effect modeling (see his example
     about no-effect-levels for human toxicity).

   These large uncertainties raised during the workshop the
 justified question  whether characterization factors in life
 cycle assessment should be based on threshold informa-
 tion. We like to  add to a few additional elements to this
 discussion.

   The impact assessment phase emerged from the wish to
 aggregate the large amount of data from inventory analysis
 to a manageable amount of impact data. For most impact
 categories, initially rather simple modeling was used to es-
 tablish characterization factors. These characterization fac-
 tors were limited to equivalency assessment on the basis
 of intrinsic substance characteristics like the potential to
 release hydrogen  ions (acidification assessment) or (no-)
 effect-levels  (toxicity  assessment)7. No assessment of
 threshold exceedance (i.e. PEC/PNEC >1)8 was performed
 since the available data did not allowsuch evaluation.Thresh-
 old information, usually in the form of a no-effect-level, was
 used in toxicity assessment only to express the emission
 of a given substance in the equivalent emission of a refer-
 ence substance, or as a dilution volume of the receiving
 environment. The basis of equivalency was taken in the
 toxicity potential of each substance9. The impact from an
 emission quantity equal to the  no-effect-level was put on
 one, and the impact from any deviating quantity as the ratio
 of the emission quantity divided by the no-effect-level. This
 provided the possibility to aggregate substances with dif-
 ferent toxic effects9.

   Present toxicity  factors cover now also fate and expo-
 sure modeling, but aggregation  of the calculated exposure
 increases from different substances is still often based on
 no-effect-levels10 (  A PEC/PNEC; see elsewhere in these
 proceedings the  presentation of Hertwich, Jolliet and
 Huijbregts).

 Threshold Information in Life Cycle
      Assessment
   Though evaluation of threshold exceedance was initially
 not performed due to lack of data, it has meanwhile turned
•No-effect-levels are based on experiments on test-species under laboratory condi-
tions and therefore say something about the intrinsic potential of a substance to
cause toxic effect (rather than something about the sensitivity of a species in real-
life for this toxic substance).
•PEC la the acronym of Predicted Environmental Concentration. PNEC is the acro-
nym of Predicted No Effect Concentration.
*The underlying assumption is that the toxicity impact from a quantity at the no-
elfect-tevel of one substance has the same importance as the toxicity impact from
a quantity at the no-effect-level of another substance. To put it more clearly: If the
quantities of both substances are at their no-effect-level, the impacts from a neuro-
toxic substance and an irritating substance are regarded as equally important. The
adding together o( completely different effects is one of the more serious problems
In LCA, but not further addressed here.
 for many practitioners into a principle in itself that is justi-
 fied by the reasoning that "less pollution is better". The thresh-
 old discussion is topical for already quite some years now.
 Unfortunately, the discussion seems to get stuck in a con-
 troversy about "less is  better" versus "only above thresh-
 old".

   As mentioned in the previous section, the question was
 raised at the workshop  whether characterization factors in
 life cycle assessment should  make use of threshold infor-
 mation because of their uncertainty. Reluctance was ex-
 pressed with regard to  acidification and eutrophication in
 particular. Refraining from the use of threshold information
 in toxicity assessment was on the other hand not a point of
 discussion10. This  is surprising, since the presentation of
 Finnveden illustrated that uncertainties in threshold infor-
 mation about toxicity of substances are far larger than for
 acidification and eutrophication.

   Present toxicity  factors use threshold information to fa-
 cilitate aggregation of exposure increases of different sub-
 stances (see also  previous section). The use of threshold
 information inherently requires making a threshold evalua-
 tion. As the  next section will show, this applies as well for
 "less is better" methods!

   We like to stress here that the uncertainties in toxic no-
 effect-levels are as a matter of fact far  more problematic
 than the uncertainties in the threshold information of other
 impact categories11. On the other hand, the modeling of fate
 and exposure are sources for considerable uncertainty as
 well in toxicity factors particularly.  It would thus not solve
 any problem to exclude threshold information as basis for
 characterization factors in life cycle assessment.

   We fully agree with Finnveden that the large uncertainties
 in threshold  information (together with other uncertainties)
 are of major concern. However, it seems far more relevant
 to discuss how to deal with those uncertainties, than sim-
 ply to abandon the use of uncertain  information  from life
 cycle assessment. Such  discussion  does  also justice to
 the valuable points brought forward by Finnveden.

   As we will discuss in the next sections, the issue of how
 to deal with  uncertainties in threshold information is quite
 another one than the issue of how to perform threshold evalu-
 ation.
'"Threshold information is still needed to aggregate different substances. If it were
no longer acceptable to base toxicity factors on no-effect-levels because of their
uncertainty, we would almost be back to square one. There would be little possibili-
ties left than to start again adding together toxic substances on mass basis, or to
report them as separate impact categories.
"The critical loads for acidification for instance are based on the capability of
target-systems to cope with system pressure, but not on the effect of the individual
substance as such  (as is the case for toxic no:effect-levels). The capability of
target-systems to cope with acidifying substances depends mainly on the charac-
teristics of the exposed target system to neutralize and/or remove acidifying sub-
stances and the sensitivity of the target species for acidifying exposure. As a
result, the uncertainties in critical loads are mainly due to uncertainties in target
characteristics, but hardly resulting from uncertainties in substance characteris-
tics. Since critical loads are based on system pressure, also uncertainties posed
by differences in the chosen endpoint and/or way of modelling effect are hardly at
stake (in contrast with toxic no-effect-levels).
                                                           68

-------
"Less is Better" Also Performs Threshold
Evaluation

  We would like to elaborate a little bit further on the use of
threshold information as basis for characterization factors.
Although up to now implicit, also "less is better" methods
do perform a threshold evaluation by basing their character-
ization factors on threshold information. We like to illustrate
and clarify with an example about acidification. We will use
this example to illustrate other forms of threshold evalua-
tion as well.

  Let's assume that we have three ecosystems  receiving
similar quantities of acidifying substance from our functional
unit12. Let's further assume that these ecosystems have a
priori equal tolerance for acidifying loading (i.e. similar criti-
cal  loads), but that they are now at different levels of back-
ground loading:

  • A Scottish ecosystem with a background load at that
    critical  load

  • A Scandinavian ecosystem with a background load far
    below that critical load

  • A German ecosystem with a background load far above
    that critical load

  We want to prioritize for which ecosystem we like to re-
duce the full acidifying load (background load + marginal
increase from our functional unit).

  All ecosystems receive an equal contribution  from our
functional unit and have also similar critical loads. "Less is
better" acidification factors consistent with the existing ones
for toxicity can easily be established from the increase in
the acidifying load divided by critical load for the given eco-
systems ( A PAL/CL)13. Since all ecosystems have the same
critical load, this would result in a similar acidification factor
for each of the above ecosystem. The background concen-
tration does not play a role here.

  The practical implication is that the ecosystem exposed
around the critical  load is regarded equally important as the
ecosystem  exposed far above  the critical load (and thus
difficult to rescue), and as equally important as the ecosys-
tem exposed far below the critical load (and thus hardly in
danger). Whether these ecosystems are exposed  far below
or at or far above their critical loads, they are all character-
ized as being equally vulnerable in such a "less  is  better"
approach. That might be or might not be a justified choice,
but  it is at  least important to realize that such choice is
made by following a "less is better" approach!
12We would like to recall here the important feature of many present environmental
problems that a single source (thus also an emission/f.u.) contributes only very
little, in most cases marginally to the background exposure of any receptor.
""PAL!' stands for predicted acidifying load (copied from predicted environmental
concentration in toxicity assessment).
 "Only Around Threshold" An Alternative?
  We make another choice than the above "less is better"
 approach. We would like to rescue that nice ecosystem in
 Scotland that is still unspoiled but about to be damaged.
 We therefore would like to prioritize the Scottish ecosys-
 tem. We would thus like to reduce the acidifying load on the
 Scottish ecosystem rather than the load on that Scandina-
 vian ecosystem or the load on that German ecosystem.

  One source (i.e. one process) contributes to very many
 ecosystems (far more than hundred thousand).Those eco-
 systems may differ in their acidifying background loads (as
 in the above example), but they can also differ in sensitivity
 (i.e. the moors in Scotland, and the rather insensitive eco-
 systems/soils in South Europe). In Potting et al. (1998a'b),
 we based our characterization factors on the slope of the
 sigmoid dose-effect curve that is defined by the critical loads
 of all ecosystems to which one source contributes. We es-
 tablished the slope of this dose-effect curve in the working
 point that is determined by the operative background loads
 and thus accounted for both differences in ecosystem sen-
 sitivity and differences in background load!

  The slope of the sigmoid dose-effect curve used by Pot-
 ting et al. (1998a'b) is one way to  account for both differ-
 ences in ecosystem sensitivity and differences in  back-
 ground load. The characterization factors based on this slope
 tell us what area of ecosystem becomes unprotected as a
 result of an emission.The practical implication of using the
 change in unprotected ecosystem area is, as we have al-
 ways emphasized explicitly, that ecosystems with deposi-
 tions far above or far below their critical loads are not ex-
 pressed in the characterization factor (it might inspire one
 to name "our" approach as "only around threshold"). That
 may be or may not be regarded as a disadvantage. How-
 ever, several other possibilities are available to deal with
 both differences in ecosystem sensitivity and differences
 in background load.

 Other Ways of Threshold Evaluation
  In Potting et al. (1998a'b), we propose to transfer "our" risk-
 based dose-effect curve into a damage-based curve to over-
 come this problem. A critical load value can tell if there is a
 risk of ecosystem damage (i.e. the acidifying load is at or
 above critical load), but it doesn't tell whether this risk actu-
 ally results in damage or how large this damage will be.
 Nevertheless, one expects the damage to become asymp-
totically larger with increasing exceeding of critical load
values. This can be used to establish acidification factors
 in which all ecosystems exposed above ecosystem are re-
flected, but in which ecosystems exposed around critical
 load have higher importance than ecosystems exposed far
above critical load. The damage oriented eco-indicator of
 Goedkoop follows basically a quite similar approach, al-
though his acidification factors are based on a different end-
 point and lacks for the time being fate and exposure model-
 ing (see elsewhere in these proceedings).

  Lindfors et al. (1998) established acidification factors
based on models very similar to those used by Potting et
                                                      69

-------
al. (1998a'b).Their acidification factors are calculated as
the share of an emission that deposits on ecosystems
with already exceeded critical loads multiplied with the
potential of the given substance to release hydrogen ions.
These characterization factors tell us what change an
emission evokes in the exceedance of already exceeded
critical loads (A PAL/CLI PAL>CL). Huijbregts (1998)
proposes acidification factors almost similar to the ones
of Lindfors etal. (1998).

  A disadvantage of the acidification factors of Lindfors et
al. and Huijbregts is that they do not account for the extent
to which the critical loads of ecosystems are exceeded (i.e.
differences in background loads above threshold do not play
a role). Practical implication of their acidification factors is
that ecosystems exposed around critical load are regarded
equally  important as ecosystems exposed far above criti-
cal load. Another, more sophisticated way to account  for
background load is presaged by Pleijel et al. (1997). The
characterization factors of Lindfors and Huijbregts can eas-
ily be adjusted by  multiplying the change in exceedance
with a severity factor given by the existing background load
divided  by the critical load for the given ecosystem (Posch
1998). Such acidification factors do, similar as the  damage
based ones from Potting et al. (1998a'b), account for both
differences in ecosystem sensitivity and differences in back-
ground loads above threshold.

  Huijbregts (1998) also proposes another way to calculate
acidification factors. These factors are basically similar to
those mentioned in the previous section about "Less is bet-
ter". He wants to calculate the relative contribution of a source
to exceedance of the critical load by calculating the share
of emission that deposits on ecosystems with already ex-
ceeded critical loads divided by the critical load of the given
substance. Such acidification factors do account for differ-
ences in ecosystem sensitivity, but not for differences in
background loads above threshold (see above).
                               "Less Is Better" Versus "Only Above
                               Threshold" Too Simple
                                 The threshold discussion has been topical for already quite
                               some years now but seems to get stuck in a controversy
                               about "less is better" or "only above threshold" principle.

                                 As shown in the previous sections, the "only above thresh-
                               old" principle  covers  very different ways of dealing with
                               threshold. An  important objection from the "less is better"
                               advocates towards the "only above threshold" and the "only
                               around threshold" principles, is that ecosystems exposed
                               below their critical loads are not accounted for in the char-
                               acterization factors. That might be a justified criticism,  but
                               it is the question whether it is a good alternative to consider
                               all ecosystems as equally vulnerable by disregarding their
                               background loads. However, most of the characterization
                               factors discussed above can be adjusted easily to cover
                               also ecosystems exposed below their critical load. It would
                               therefore be interesting to calculate each of the above-dis-
                               cussed acidification factors with the same models (for in-
                               stance with the RAINS model), but to establish each of
                               these factors for a below, and above, and below & above
                               threshold situation (Table 1).

                                 Within each principal framework of the acidification fac-
                               tor, the "below & above threshold factors" should logically
                               be a summation of the "below threshold factor" with the
                               "above threshold factor". Furthermore, we expect the spa-
                               tial differentiated "below & above threshold factors" to show
                               the same, but smoothed trend as the similar "above thresh-
                               old factor". However, it would be interesting to see whether
                               the acidification factors based on different principal frame-
                               works also show the same trends.

                                 The results of above comparisons, completed with a thor-
                               ough and fair evaluation of the implications of each type of
Table 1.
Name
acidification factors
Excoodance of CL
Severity of exceedance
Exceedance relative to CL
Principal
framework
AF=PAL
AF=ฃAPAL*PAL/CL
AF=SAPAL/CL
AF = "principal
PAUCL
X
X
X
X
X
X
framework"
PAL>CL
X
X
X
X
X
X
Authors

Huijbregts 1998, Lindfors et al. 1998
Present document
Huijbregts 1998
Pleijel etal. 1997
Present document, Posch
Present document, Posch
Huijbregts 1998
Present document
Huijbregts 1998
1998
1998

Unprotected area


Ecosystem damage
      AF=SA
(forPAL+APAL)/CL=1)

 AF=2D(PAL+ APAL)
                                                  X
                                                  X
   X
(PAL=CL)

   X

   X
Potting etal. 1998
Potting etal. 1998
Present document
Goedkoop in present proceedings
CL stands for Critical Load, PAL stands for Predicted Acidifying Load, A stands for area of ecosystem, D stands for Damage
                                                       70

-------
acidification factor may give fresh input to the threshold
discussion. It will for sure provide a sound basis to make an
explicit choice for either one or another type of acidification
factor.

Closing Remark
  It might be clear that the whole discussion above is ex-
emplified with acidification but is valid as well for the other
impact categories!

Acknowledgment
  We like to thank Goran Finnveden for the stimulating dis-
cussions in relation to this presentation.

References
Huijbregts, M. Operational uncertainties in life-cycle impact
    assessment of acidification and eutrophication; a pro-
    posal. Amsterdam (the Netherlands), Interdisciplinary
    dept. of Environmental Science (IVAM) at the Univer-
    sity of Amsterdam, not published.

ILSI. Human health impact assessment in life-cycle assess-
    ment: analysis by an expert panel. Washington DC (United
    States of America), International Life Science Institute,
    1996.

Lindfors, LG., M. Alkemark, C. Oscarsson and C. Spannar. A
    manual for the calculation of ecoprofiles intended for third
    party certified environmental product performance decla-
    rations. Stockholm (Sweden), Swedish Environmental
    Institute (IVL), 1998.

Pleijel, K., J. Altenstedt, H. Pleijel, G. Lovblad, P. Grennfelt, L.
    Zetterberg, J. Fejes and L-G. Lindfors. A tentative meth-
    odology for the calculation of global and regional impact
    indicators in type-Ill ecolabels used in Swedish case stud-
    ies. Stockholm (Sweden), Swedish Environmental Insti-
    tute (IVL), not published.

Posch, M., P.A.M. de  Smet,  J.P. Hettelingh, R.J. Downing.
    Calculation and mapping of critical thresholds in Europe:
    Status report 1995. Bilthoven (the Netherlands), Co-ordi-
    nation Center of Effects located at the National Institute
    of Public Health and the Environment, 1995.

Posch, M., J.P. Hettelingh, P.A.M. de Smet, R.J. Downing.
    Calculation and mapping of critical thresholds in Europe:
    Status report 1995. Bilthoven (the Netherlands), Co-ordi-
    nation Center of Effects located at the National Institute
    of Public Health and the Environment, 1997.

Posch, M. Personal communication from M. Posch. Groningen
    (the Netherlands), 1998.

Potting, J., M. Hauschild, H. Wenzel. "Less is better" and "only
    above threshold";Two incompatible paradigms for human
    toxicity in life-cycle assessment? Journal of Life Cycle
    Assessment, Vol. 4 (1999), Issue 1, pp 16-24.
 Potting, J. and M. Hauschild. Predicted environmental im-
    pact and expected occurrence of actual environmental
    impact. Part 1: The linear nature of environmental im-
    pact from emissions in life-cycle assessment. Interna-
    tional Journal of Life Cycle Assessment, Vol.2 (1997a),
    Issue 3, pp 171-177.

 Potting, J. and M. Hauschild. Predicted environmental im-
    pact and expected occurrence of actual environmental
    impact. Part 2: Spatial  differentiation in life-cycle as-
    sessment by site-dependent assessment of environ-
    mental impact from emissions. International Journal
    of Life Cycle Assessment, Vol. 2 (1997&), Issue 4, pp
    209-216.

 Potting, J. and M. Hauschild. Spatial differentiation in life-
    cycle impact assessment; technical backgrounds (pro-
    visional title). Copenhagen (Denmark), Danish Environ-
    mental Protection Agency, forthcoming.

 Potting, J., W. Schopp, K. Blok and M. Hauschild. Com-
    parison of the acidifying impact from emissions with
    different regional origin  in life-cycle assessment. Jour-
    nal of Hazardous Materials, Vol. 1998a, Issue 61, pp
    155-162.

 Potting, J., W. Schopp, K.  Blok and M. Hauschild. Site-
    dependent life-cycle impact assessment of acidifica-
    tion.  Journal of Industrial Ecology, Vol. 2 (1998b), Is-
    sue 2, pp 63-87.

 Potting's Biography
  In 1991, Potting started her research in the field of life-
 cycle assessment and energy analysis within the Dept. of
 Science, Technology and Society at Utrecht University. Her
 research was case study oriented, but she made also sev-
 eral outstanding methodological contributions. This resulted
 in an  invitation  by the Life-Cycle Centre at the Technical
 University in Denmark to elaborate her thoughts on spatial
 differentiation in life-cycle assessment. In 1996, she started
 as a guest researcher at the Technical University of Den-
 mark. She was offered here in 1997, on  return from a three
 month fellowship at the International Institute of Applied
 System Analysis in Laxenburg in Austria, a permanent po-
 sition  at the Institute of Product Development that is affili-
 ated to the Technical  University of Denmark. Since 1996,
the research activities from Potting are focussed  fully on
 method development for spatial differentiation in life-cycle
assessment. During all her appointments, she actively par-
ticipated  in several international work groups (NNI/ISO,
SETAC, LCANET a/o.). At present, she is chair of the sci-
entific task group on acidification, eutrophication and nutri-
ent enrichment (WIA2/STG-AENE)  established under the
 European branch of SETAC. In all her university appoint-
ments, Potting has been actively involved in  educational
matters. She initiated and elaborated several MSc-courses
and she tutored in several subjects.
                                                      71

-------
Polling's Overheads
      Levels of sophistication in life
        cycle impact assessment of
                acidification

               November 1998

            Jose Potting (EPU/DTU)
           Wolfgang Schopp (EASA)
          Komelis Blok (NWenS-UU)
         Michael Hauschild (IPU/DTU)
       Outline
       • Problem setting
       • Methods and means
       • Results
       • Levels of sophistication
       • Conclusions
                 An example:
     Location alternatives for copper
                 production

                     in
          Albania & Belgium & Finland

                 Assumptions
          Similar production technologies
                    with
     Similar emission quantities per kton copper
       Inventory analysis

   Acidifying emissions per kg copper

            SO2 * 100 gram
            NOX ซ 10 gram
   Assessment of "hazard potential"
              (per kton copper)

   SO2 ~ 100 * 1ซ 100 gram SO2-equivalents
    NOX ~ 10 * 0.7 ~ 7 gram SO2-equivalents

       total * 100 + 7 ซ107 gram SO2-
                  equivalents
         Poor accordance between
 Hazard potential and actual impact


           Due to disregard of:
  Emission dispersion and deposition patterns
Background depositions on receiving ecosystems
     Sensitivity of receiving ecosystems
                                           72

-------
Traditional impact assessment tools
        (like gaussian plume modelling)


High accuracy to predict acidification local
            to the source (r < 50 km)

  Unable to predict acidification over large
             distances (r >1000 km)
                                                                       RAINS

                                                    Regional Air pollution INformation System


                                                                 Emission inventories
                                                        Atmospheric dispersion and deposition
                                                                      Critical loads
                                                         (Optimization of emission reduction)
Emissions of SO
  Albania Belgium Finland
   11
   35
    2
   35
   13
    4
    0
   100
   120
 0
 26
 27
 32
 10
  5
  0
100
317
  3
 35
 24
 30
  6
  3
  0
100
         Sector
%   Fuel conversion
%   Industrial combustion
%   Industrial processes
%   Power plants/district heating
%   Domestic, agriculture
%   Transportation
%   Otherwise
%   Total
260   kton Total
SOa depositions from
Finland
Transfer-matrices from EMEP
MSC-W at the Norwegian
Meteorological Institute
SO2 depositions from
All 44 regions
Transfer-matrices from EMEP
MSC-W at tiie Norwegian
Meteorological Institute
o
o
i
 Deposition of SO2 on grid x=20, Y=26 of Finland
 Region
 Austria
 Baltic sea
 Belarus
 Belgium
 Bohemia
 Bulgaria
 Croatia
 Czech republik
 Denmark
 Estonia
 Rnland
 France
 Germany-exBRD
 Germany-exDDR
 Hungaria
 Ireland
 Italy
 Latvia
 Contribution
     0.01%
     1.24%
     3.13%
     0.38%
     0.11%
     0.14%
     0.04%
     3.81%
     0.97%
      8.6 %
    25.95 %
     0.67 %
     13.4%
     1.64%
     1.21 %
     0.02%
     0.14%
     1.52%
       Region
       Lithuania
       North sea
       Norway
       Poland
       Romania
       Russia - Kalingrad
       Russia - Kola, Karelia
       Russia - Remaining areas
       Russia - St. Petrusburg
       Slovakia
       Slovenia
       Sweden
       The Netherlands
       Ukraine
       United Kindom
       Yugoslavia
       Natural emissions
       Non-attributable sources
       Total
                 Contribution
                     1.77%
                     0.24%
                     0.23%
                     9.68%
                     0.53%
                     2.84%
                     3.69%
                     0.16%
                     3.15%
                     0.61 %
                     0.01 %
                     1 .77 %
                     0.24%
                     2.41%
                     3.89 %
                     0.13%
                     0.36%
                     5.33%
                    100.00%
                                                                          0    500    1000   1500   2000   2500

                                                                            Critical Loads for actual acidity in eq/ha.yr
                                                          73

-------
r
                                                Calculations based on:

                                                Transfer-matrices from EMEP
                                                MSC-W at the Norwegian
                                                Meteorological Institute
                                                Critical loads provided by
                                                CCEatthe National Institute
                                                for Effects of the Public Health
                                                and the Environment in The
                                                Netherlands
                           Area (in %) of unprotected
                           ecosystem (depositions
                           exceed critical loads)
                                                                                             Logical deductions
                                                               Total emissions from a region contribute little to the
                                                                    total depositions on receiving grid elements

                                                               The full emission from a separate source contributes
                                                                     little to the total emissions from a region

                                                               The full emission from a separate source contributes
                                                                marginally to the total depositions on receiving grid
                                                                                    elements
Mathematical framework for the acidification factors
                 m
  ( UES(E(ref)) - ZUESsJ((l-5)E(ref))) / 5Es4(ref) =
                 •fi
                       Example for an SO2 emission from Finland
                                   ( 81671-81278 )/26 =
                                        394/26 =
                                        15 ha/ton
                                      0.15 mVgram
                                                                                 Site-dependent assessment of the acidifying impact
                                                                                                  (per kton copper)
                                                                                     n    z
                                                                                    E   Z   Emission;  * Acidification Factor;
                                                                                                      '                      '
                                                                          SO2          NOX
                                                                Albania  100*0.00  +   10*0.00  =    Om2
                                                                Belgium  100*0.01  +   10*0.01  =   1.4m2
                                                                Finland  100*0.15  +   10*0.02  =  15.2m2
                     Acidification factors (in ha/ton)
                     Region
                     Albania
                     Austria
                     Batons
                     Belgium
                     BosniaHerzogovina
                     Bulgaria
                     Croatia
                     CRZF
                     Denmark
                     Estonia
                     Finland
                     France
                     Gomanynow
                     Gomany old
                     Grooco
                     Be	
                           S02
                           0.02
                           1.31
                           4.6S
                           1.28
                           0.15
                           0.07
                           0.30
                           1.91
                           5.56
                          12.43
                          15.14
                           0.79
                           2.17
                           1.94
                           0.01
NO.
0.00
0.42
4.54
0.82
0.04
0.02
0.12
0.69
2.02
1.54
2.42
0.47
0.90
1.42
0.00
 NHj
 0.01
 3.44
 5.72
 1.10
 0.06
 0.05
 0.17
 1.26
 5.28
 3.92
13.40
 0.74
 1.89
 3.31
 0.01
Interim-conclusions
• Acidification factors have been established
  for 44 European regions
• The application of these factors is very simple
• The resolving power is large, while the
  uncertainties are small
• Acidification factors show reasonable good
  stability for changes in the reference situation
                                                                         74

-------
Recent developments/
                forthcoming work

• Hogan et al. (Int. J. of LCA 1996)
• Tolle et al. ant. J. of LCA 1997)
• Seppalla (FBI 1997)
• Potting et al. (JIE 1998)
• Norris (proposal; US-EPA 1998)
• Rabl et al. (1998)
• Pleijel et al. (proposal; not published)
• Huijbregts (proposal; not published)
Levels of sophistication

• Detail of assessment?
• Uncertainties?
• Marginal or average impacts?
• Threshold evaluation?
• Anticipated indicators/endpoints?
• Type of modelling?
Final conclusions

• A considerable level of spatial detail in LCA is
  possible with limited additional data
  requirement from inventory analysis
• Lack of spatial differentiation results in large
  uncertainties in the assessed impact
• The level of sophistication in impact
  assessment may differ depending on the type
  of application
• All recent developments make use of (similar)
  integrated assessment models
                                           75

-------
                         Eutrophication - Aquatic and Terrestrial
                                           State-of-the-Art
                                             Goran Finnveden
                                    fms (Env. Strategies Research Group)
                               National Defence Research Establishment (FOA)

                                               Jose Potting
                                      Institute of Product Development
                                      Technical University of Denmark
 Abstract

  State of the art and research needs for the impact cat-
 egory eutrophication are discussed. Eutrophication is a dif-
 ficult impact category because it includes emissions to both
 air and water, both subject to different environmental mecha-
 nisms, and  impacts in  different types of terrestrial and
 aquatic ecosystems.The possible fate processes are com-
 plex and include transportation between different ecosys-
 tems. In some recent approaches modeling of transporta-
 tion of air emissions has been included. However, in gen-
 eral the characterization methods used lack fate modeling
 and  this is a limitation.  Also the definition  of the impact
 indicator may need further research.The inclusion of other
 nutrients than those typically considered should also be
 investigated.

 On "Thresholds'VNo Effect Levels"/
 "Critical Loads" etc.

  Based on two example, human toxicity and acidification,
 it is suggested that "thresholds" are always based on a choice
of what is "acceptable", or "significant" or something similar
that in turn is based on  value judgements. It is also sug-
gested that the determination of "thresholds" is based on
models that often are highly uncertain. It is thus clear that
LCA is not the only science where there is an uncertainty
concerning the appropriate methods and where different
methods can give significantly different results. It is also
suggested that since the determination of "thresholds" ulti-
 mately is based on values, the "only above thresholds" ap-
 proach should be avoided if we want to avoid values in the
 characterization. It is also important to realize that impacts
 can occur below "thresholds".

 References
 Finnveden, G. and Potting, J. (1999): Eutrophication as an
    Impact Category - State of the Art and Research Needs.
    Int. J. LCA. In press.

 Finnveden, G.and Potting, J. (1999): On the nature of "thresh-
    olds" and some implications. In preparation. Provisional
    title.

 Finnveden's Biography
  Finnveden has a M.Sc in Chemical Engineering and PhD
 in Natural Resources Management. Currently working at fms
 (Environmental Strategies Research Group), which is a co-
 operation between the Department of Systems Ecology at
 the Stockholm University and the National Defence Research
 Establishment (FOA) in Stockholm, Sweden. Has been
 working with Life Cycle Assessment since 1990, both on
 case studies and development of methodology, e.g. the
 development of the Nordic Guidelines on Life Cycle Assess-
 ment. Active in international working groups, e.g. the "Inter-
 national Forum on Life  Cycle Assessment and Integrated
Solid Waste Management" and the SETAC-Europe working
group on Life Cycle Impact Assessment where he currently
is chairing the task group on normalization and weighting.
                                                   76

-------
Finnveden's Overheads
                 Eutrophication -
           Aquatic and Terrestrial -
                 State of the Art
     Goran Finnveden
     fms and Stockholm University
     Jose Potting
     Institute of Product Development
     (affiliated to the Technical University of Denmark)
Eutrophication is a difficult category


•  Emissions to air and water
•  Impacts in different types of terrestrial and aquatic systems
•  Fate processes are site dependent
•  Impacts are site dependent
   - different background loads
   - different nutrients may be limiting
   - different sensitivities of ecosystems
         Inventory parameters normally
          assigned to "Eutrophication"/
                "Oxygen depletion"

          N-emissions to air
          Macro-nutrients (N and P) to water
          Organic material to water

          Other nutrients which locally could be
          limiting are normally not considered
           Limiting nutrients
   Terrestrial systems, often N
   (at least for terrestrial ecosystems in Europe and North
   America)

   Aquatic systems
    - freshwater often P-limited
    — seawater often N-limited
    - coastal and brackish water, P or N or both
                       ...but...
        The picture of "limiting nutrient" is often
        simplified because
         - the limiting nutrient may change over the year
         - the limiting nutrient may change over the years
         — the balance between nutrients is often important
         - nutrients may be transported
        Therefore, the contribution from an emission to
        eutrophication is always(?) larger than 0
    Examples of possible fate of N
               emitted to air

   It can be deposited on surface water
   It can be deposited on (natural) soil or vegetation
   - it can be leached out to surface water
      • the fraction being leached out depends on site specific aspects
       (Varies between approx 5-80 % in Europe).
   - it can be taken up by growing vegetation, contributing
     to terrestrial eutrophication
      • after degradation it can be taken up again, leached out,
       exported with harvested timber, immobilised in the soil or
       denitrified
                                                       77

-------
                   ...fate...
  It can reach strictly P-limited waters
   — it can be lost from the system (e.g. by denitrification)
   - it can be transported to other waters
  It can reach waters where N contributes to
  eutrophication
   — after degradation it can be taken up again, transported
     to other waters, exported with fish, or denitrified
Three methods were suggested in
                  1992-93

        Method 1 (Jensen et al, 1992)
        • No aggregation at all
        • Four subcategories
          - emissions of N to air
          - emissions of N to water
          - emissions of P to water
          - emissions of BOD
            ...three methods...

 Method II (Heijungs et al, 1992)
 • Complete aggregation into one category
 • Definition of the effect: Biomass production.
 • All emissions of N (air and water), P and organic
   material is assumed to contribute once and to the
   same impact.
 • Weighting factors based on Redfield ratio:
   C:N:P= 106:16:1
           ...three methods...

Method III (Finnveden et al, 1992 and Samuelsson, 1993).
• Scenario based approach
• Terrestrial and aquatic systems are separate
• Terrrestrial systems: N-emissions to air.
• Aquatic systems:
   - Definition of the effect: Oxygen consumption
   - Weighting factors based on Refield ratio
   - Emissions can only contribute once, but sometimes
     not at all
            ...three methods...

Method in - continued
• Four different scenarios for aquatic systems:
   — P-limited: Aggregation of P and org. mat.
   - N-Iimited: Aggregation of N to water and org. mat.
   - N-limited+- air: Aggregaton of N to water and air and
     org. mat.
   - Max: Aggregation of P and N to water and air and org.
     mat. (Identical to Heijungs et al, 1992)
       Developments  1993-1996

 Demands for site dependent factors
 Nice student work from ETH suggesting site dependent
 factors reflecting sensitivity and limiting nutrient.
 Method IV (Hauschild and Wenzel, 1996).
  - Similar to methods II and III
  - Organic material excluded
  — Either N and P separate or combined
 Method V (Tolle, 1997)
  — Regional scaling factor (1-9) based on present loadings
                                                     78

-------
 SETAC-Europe working group
           onLOA(1996)

Asking for adaption of earlier approaches:
 - Several subcategories: terrestrial systems, smaller
   aquatic systems, larger aquatic systems?
 -Other nutrients?
 — Sensitivity of receiving environments
 - Transportation
    • For emissions to air, similar models as for acidification
    • For water emissions: ?
      Recent work by Seppala

  Two impact categories for aquatic ecosystems:
  Oxygen depletion: BOD
  Aquatic eutrophication
  - Definition of the effect: Increased production based on
    Redfield ratio
  - For water emissions of N and P, site specific transport
    and effect factors were determined by experts.
  — For air emissions of N, transport factors were based on
    EMEP-data (in this case, 6-7% of emitted N are
    deposited in N-sensitive areas)
 Recent work by lindf ors et al and
                Pleijel et al
 Aquatic oxygen depletion for aquatic ecosystems
  - Definition of the effect: oxygen depletion
  - Weighting factors based on Redfield ratio
  - Site specific assessment in which emissions are only
    considered for receiving environment where pollutants
    have an effect as determined by expert judgement
   Recent work by Potting et al
  So far, only emissions of N to air for terrestrial
  ecosystems
  - (research on water emissions is ongoing)
  Transport from EMEP-models
  Definition of the effect: A marginal approach:
  The area of ecosystem that becomes unprotected as a
  result of that emission
     Discussion issues - effects

   Distinction between terrestrial and aquatic systems?
   Definition of the effect
   — Aquatic systems: Oxygen consumption/biomass
     production?
     * 7
   - Terrestrial systems:
     * Area "unprotected" which receives emission?
     * Amount emitted into "unprotected" area?
     *9
   Only above thresholds, or both below and above, perhaps
   in separate subcategories?
   Marginal or average approach?
Research issues - fate and target modelling

 • Emissions to air
       Integrated assessment models?
       "Background" concentrations
       Target sensitivity
       Leaching of nutrients?
          N deposited on land can be leached out and reach N-
           sensitive area (possibly after transport)
 • Emissions to water?
       Research project? Study some site specific cases in
        order to get an understanding of the variability.
       Role for expert judgements?
                                                     79

-------
   Research issues - spatial info

  Spatial differentiation?
  How do we make optimal use of spatial
  information?
  Can we develop spatially differentiated and
  not differentiated characterisation factors
  which are compatible?
  Requirements towards inventory	
  On "thresholds'/no effect
  levels'Vcritical loads" etc


     Goran Finnveden
     fms and Stockholm University

     with contributions from
     Jose Potting
     Technical University of Denmark
              Suggestion
  "Thresholds" are always based on a choice of
  what is "acceptable" or "significant" or similar.
  The choice of an "acceptable" impact is based on
  values.
  Determination of "thresholds" are based on
  models which are uncertain

  Two examples:
  acidification and human toxicity
           Acidification
Critical loads often defined as

"A quantitative estimate of an exposure to one or
more pollutants below which significant harmful
effects on specified sensitive elements of the
environment do not occur according to present
knowledge".
                       (Nilsson and Grennfelt, 1988)
           Acidification....

Results from "Lake Gardsjon", Sweden,
  Andersson et al (1998):
Comparison of 8 more or less "official"
  methods for calculating CL:
CL for S varies between 1-20 kg S/ha yr
CL for N varies between 8-15 kg N/ha yr
          Human toxicity
 Dose-response curves for individuals may
 or may not have a threshold
 Dose-response curves for populations do
 not have a threshold.
 "Thresholds" are always based on a choice
 of a "acceptable" risk, such as 1:106
                                              80

-------
             Human toxicity...
Differences in model-derived estimates of virtually safe doses
      at 10-6 risk level, relative to the one-hit model
 (Casarett and Doull's Toxicology, based on other sources)
Carcino-   One-hit
gen
Linearized Weibull
multistage
                                         Multihit
Vinyl      1
chloride

Aflatoxin   1
1
20
                               1000
                                         2*10~8
8000
                                                Conclusions/discussion
                      LCA is not the only science where there is
                      an uncertainty concerning the appropriate
                      methods and where different methods can
                      give significantly different results
         Conclusions/Discussion...
   If we want to avoid values in the characterisation, we
   should perhaps avoid the use of "only above thresholds"
   Impacts can occur below "thresholds"
   - especially if the threshold is for a single substance while it
     contributes to a cocktail of pollutants for which no threshold has
     been defined
   Suggestion: if thresholds are used, include both above and
   below thresholds in the impact assessment, but make a
   distinction
   We can use thresholds as a basis for characterisation
   factors and use the science behind to define the effect
   Strive for a good balance between the additional
   information gained and the new uncertainties introduced
                                                       81

-------
                                             Appendix
                                       List  of Participants
Guntra Aistars
University of Michigan
Dana Building 430 E. University
Ann Arbor, Ml 48109-1115
tele:734-764-1412
fax:734-647-5841
email: guntra@umich.edu

Jane Anderson
BRE
Watford, Herts, WD2 7JR United Kingdom
fax:44-1923-664084
email: andersonj@bre.co.uk

Jane Bare
U.S. EPA (MS-466)
Cincinnati, OH 45268
tel: 513-569-7513
fax:513-569-7111
email: bare.jane@epa.gov

Angeline de Beaufort-Langeveld
FEFCO-GO-KI
Groeterweg13
6071 ND Swalmen
tel. 0475-501 014
fax. 0475-504 478
e-mail: beulang@wxs.nl

Mark Buridard
International Iron & Steel
Rue Colonnel Boury 120
1140 Brussels Belgium
email: lcaletter@iisi.be

Paul Butterworth
International Iron & Steel Institute
email: butterworth@IISI.be

Andreas Ciroth
Technische Universitat Berlin
Institut fQrTechnischen Umweltschutz
Abfallvermeidung und Sekundarrohstoffwirtschaft
Sekt. KF 6
Strasse des 17. Juni 135
D-10623 Berlin
tel.: 49-30-314 79564
fax:49-30-31421720
email: ciroth@itu302.ut.tu-berlin.de

Garrette Clark
UNEP
Tour Mirabeau -39-43,
Qua! Andre" Citroen
75739 Paris Cedex 15 France
Email: aarrette.clark@unep.fr
Pierre Crettaz
Ecole Polytechnique Federale de Lausanne
DGR-IATE-HYDRAM
CH-1015 Lausanne Switerland
tel.:41-21-6933729
fax:41-21-6933739
email: pierre.crettaz @ epf l.ch

Mary Ann Curran
U.S. EPA (MS-466)
Cincinnati, OH 45268
tel: 513-569-7782
fax:513-569-7111
email: curran.maryann@epamail.epa.gov

Suzy Edwards
BRE
Watford, Herts, WD27JR United Kingdom
fax:44-1923-664084
email: edwardss@bre.co.uk

Tomas Ekvall
Chalmers Industriteknik
Chalmers Teknikpark
S-41288G6teborg Sweden
tel. 46-31-772 4336
fax. 46-31 -827 421
e-mail: tomas.ekvall@cit.chalmers.se

Goran Finnveden
Fmw and Department of Systems Ecology
Box2142
SE-10314 Stockholm Sweden
tel.: 46-8-402 3827
fax:46-8-4023801
email: finnveden ฉfms.ecology.su.se

Marina Franke
Proctor & Gamble
Global Technical Policy/Environmental Management
Postfach 5825, Industriestrafe 30-34
D-65733 Eschborn, Tanus
tele:(06196) 89-4723
fax:(06196)89-6648
email: f ran ke. m @ pg .com

Jean-Paul Fretiere
European Product Responsibility Manager
3M  France
Boulevard de I'Oise
95006 Cergy Pontoise Cedex
tele:33/130318107
fax:33/130318192
email: ifretiere@mmm.com
                                                   82

-------
Bernard Gonzales
3M
U.S.A.
Email: bagonzales@ mmm.com

MarkGoedkoop
PRe Consultants B.V.
Plotterweg12
NL-3821 BBAmersfoort Netherlands
tel.: 31-33-455 5022
fax:31-33-4555024
email: goedkoop@pre.nl

Jeroen Guinee
Centre of Environmental Science
P.O. Box 9518
NL-2300 RA Leiden Netherlands
tel.: 31-71-527 7477
fax:31-71-5277434
email: guinea @ rulcml.leidenuniv.nl
Michael Hauschild
Department of Manufacturing Engineering
Technical University of Denmark
Building 424
DK-2800 Lyngby Denmark
tel.: 45-45-254 664
fax:45-45-935556
email: mic@ipt.dtu.dk

David Heath
ICI
email: david heath@ici.co.uk

Edgar Hertwich
University of California, Berkeley
310 Barrows Hall #3050
Berkeley, CA 94720-3050
tele:510-642-8853
fax:510-642-5815
email: hertwich @ socrates.berkelev.edu
Patrick Hofstetter
Umweltnatur- und Umweltsozialwissenschaften (UNS)
ETH-Zentrum HAD F1
CH-8092 Zuerich Switzerland
tel: 41-1-632 49 78
fax:41-1-6321029
email: hofstetterฎ uns.umnw.ethz.ch

Maarten ten Houten
TNO-lnstitute of Industrial Technology
P.O. Box 5073
NL-2600 GB Delft  Netherlands
tel.: 31-15-260 8828
fax:31-15-2608756
email: m.tenhouten@ind.tno.nl
Mark Huijbregts
IVAM, University of Amsterdam
Nieuwe Prinsengracht 130
NL-1018 VZ Amsterdam Netherlands
tel.: 31-20-525 6263
fax:31-20-5256272
email: m.huijbregts@frw.uva.nl

Norihiro Itsubo
Japan Environmental Management Association for Industry
Hirokohji Bldg. Ueno 1 -17-6, Taitoh-ku
Tokyo Japan 110-8535
tel.: 81-3-3832 0515
fax:81-3-38322774
email: itsubo @ jemai.or.jp

Olivier Jolliet
Institute of Soil and Water Management
EPFL-IATE-HYDRAM
CH-1015 Lausanne Switzerland
tel..: 41-21-693 7011
fax: 41-21-693 7084
email: olivier.jollietฎ epfl.ch

Henry King
Unilever
email: henry.king@unilever.com

Walter Klopffer
C.A.U.GmbH
DaimlerstraBe 23
D-63303 Dreieich Germany
tel.: 49-6103-983 28
fax:49-6103-98310
email: C.A.U.@t-online.de

Helena Malkki
VTT Chemical Technology
P.O. Box 14031
FIN-02044VTTEspoo Finland
tel.: 358-9-4566442
fax:358-9-4567043
email: Helena.Malkki@vtt.fi

Lloreng Mila i Canals
Universitat Autonomade Barcelona
Unitat de Qufmica Ffsica, Edifici Cn,
Bellaterra
Barcelona 08193 Spain
tel.: 34-3-581 21 64
fax:34-3-5812920
email: llorenc@klingon.uab.es

Karl-Michaesl Nigge
Rheinweg 27
D-53113 Bonn Germane
tel. + fax: 49-228-233 763
email: Karl-Michael.Niaae@dir.de
                                                    83

-------
Leo De Nocker
VITO
Boeratang 200
B-2400 Mol Belgium
32-14-334885
fax:32-14-321 185
email: denockel@vito.be

Gregory Morris
Sylvatica
147 Bauneg Beg Hill Rd.
Suite 200
North Berwick, ME 03906
Tel: 207-676-7640
Fax:207-676-7647
email: norris @ svlvatica.com

Stig Irving Olsen
Institute for Product  Development
Technical University of Denmark
Building 424
DK-2800 Lyngby Denmark
tel.: 45-45-254 611
fax:45-45-935556
email: sio@ipt.dtu.dk

Willie Owens
Procters Gamble
5299 Spring Grove Ave.
Cincinnati, Ohio 4521 4
Tel: 51 3-627-81 83
Fax:513-627-8198
email: owens.jw@pg.com

David Pennington
ORISE Post Doctoral Researcher
U.S. EPA (MS-466)
Cincinnati, OH 45268
tel: 51 3-569-761 8
fax:513-569-7111
email: pennington.david@epamail.epa.gov

Jose" Potting
Technical University of Denmark
c/o Bo Pedersen Weidema
IPU, Life Cycle Engineering Group
Building 40311
DK-2800 Lyngby Denmark
tele:4545254677
fax:4545935556
email: j
Steven Rolander
SAIC
11251 Roger Bacon Dr.
Reston,VA20190
Tel: 703-31 8-4605
Fax:703-736-0826
email: rolanders@saic.com
En/van Saouter
Procter & Gamble, Environmental Science Department
lOOTemselaan
B-1853 Strombeek Bever Belgium
tel.: 32-2-4562491
fax:32-2-4562845
email: saouter.e@pg.com

Guido Sonnemann
Chemical Engineering Department
Universitat Rovira i Virgili
Carretera de Salou s/n
E-43006 Tarragona Spain
tel.: 34-977-559 618
fax:34-977-559667
email: gsonnema@etseq.urv.es

Bengt Steen
CPM
Chalmers University of Technology
SE-412 96 Gothenburg Sweden
tel.: 46-31-772 2177
fax:46-31-7722172
email: bengts@vsect.chalmers.se

DuaneTolle
Battelle Columbus Division
Columbus, OH 43201
Tel.:614-424-7591
Fax:614-424-3404
email: tolled@battelle.org

Arnold Tukker
TNO-STB
P.O. Box 541
NL-7300AMApeldoorn Netherlands
tel.: 31-55-5493907
fax:31-55-5421458
email: tukker@stb.tno.nl

Helias A. Udo de Haes
Centre of Environmental Science
P.O. Box 9518
NL-2300 RA Leiden Netherlands
Tel.: 31-71-527 7488
Fax:31-71-5275587
email: udodehaes@rulcml.leidenuniv.nl

Bo Weidema
Technical University of Denmark
Inst. for Product Development
Building 424
2800 Lyngby Denmark
email: bow@ipt.dtu.dk
                                                   84

-------
Keith Weitz
Research Triangle Institute
3040 Cornwallis Rd.
RO. Box 12194
RTR NC 27709
Tel: 919-541-6973
Fax:919-541-7155
email: kaw@rti.org
HenrikWenzel
IPU-DTU
Technical University Denmark
Building 424
D-2800 Denmark Denmark
tel.: 45-45-254 663
fax:45-45-935556
email: hwc@ipt.dtu.dk
                                   ซU.S. GOVERNMENT PRINTING OFFICE:  2000-650-101-40006
                                                    85

-------

-------

-------

-------

-------
United States
Environmental Protection Agency
National Risk Management
   Research Laboratory, G-72
Cincinnati, OH 45268
Please make all necessary changes on the below label,
detach or copy, and return to the address In the upper
left-hand corner.

If you do not wish to receive these reports CHECK HEREQ ;
detach, or copy this cover, and return to the address in the
upper left-hand corner.
PRESORTED STANDARD
 POSTAGE & FEES PAID
          EPA
    PERMIT No. G-35
Official Business
Penalty for Private Use
$300
EPA/600/R-00/023

-------