Towards a consistent management of uncertainty in Life Cycle Assessment

 

Special session coordinator: Olivier Jolliet, University of Michigan

Currently, the quantification and communication of uncertainty in Life Cycle Assessment (LCA), in both inventory and impact assessment, is often omitted. Addressing these issues, this session focuses on the integration of uncertainty management into daily LCA practice. It explores approaches to estimate, visualize, interpret and communicate any kind of uncertainty information in LCA and its inclusion in decision-making. Identification and quantification of main sources of uncertainty in all stages of an LCA is a special point of interest; all with the ultimate goal of minimizing resources needed to perform uncertainty analysis in LCA. LCA case studies, where uncertainty was consistently considered, are welcome. Innovative method for uncertainty calculations are of high interest. Uncertainty information routinely reported for LCA results will improve trust and confidence in the method as users and decision makers will be provided with measures of confidence in the result, enabling for example to differentiate between compared options with scores that are essentially equal (revealed by overlapping uncertainty ranges) or well distinguishable. It will also provide a measure of confidence for impact indicators. This session explicitly invites all stakeholders from developers to practitioners and decision-makers to present and discuss their views, experiences and ideas related to practical uncertainty management in LCA.

Presenters:

Sampling and analytical approaches toward propagating uncertainties in LCA
Reinout Heijungs, Leiden University
Olivier Jolliet, University of Michigan
Ralph Rosenbaum, CIRAIG
Andreas Ciroth, GreenDeltaTC GmbH
Thomas McKone, Lawrence Berkeley National Laboratory
Manfred Lenzen, University of Sydney
Jinglan Hong, University of Michigan
presentation

The analysis of uncertainty in LCA studies has been a topic for more than ten years, and many commercial LCA programs now feature Monte Carlo analysis. Yet, a full Monte Carlo analysis of a large LCA system, for instance containing the 4000 unit processes of ecoinvent v2.0, is hardly carried out by LCA-practitioners. One important reason for this is the computation time involved. A promising alternative for the Monte Carlo method is the analytical error propagation, by means of a Taylor series expansion. This p aper will explore three different implementations of the idea behind the Taylor series expansion. It compares the theoretical background and mathematical formulas involved. A case study on fossil diesel versus biodiesel approaches these techniques from a practical angle, and moreover allows us to contrast their results with those from a Monte Carlo analysis.

Uncertainty and scenario analysis in the life cycle of biofuel systems: modelling issues and applications
Fausto Freire, ADAI. Dep. of Mechanical Engineering. Universisty of Coimbra
João Malça, ADAI, University of Coimbra, and ISEC
presentation

Recently there has been a significant growth in the number of published studies addressing the life cycle of biofuel systems. However, several aspects have been found to affect the life cycle calculations: land use change, data quality, modeling assumptions, and so on. Therefore, a comprehensive evaluation of uncertainty in the life cycle of biofuels is needed. This presentation evaluates the implications of uncertainty and scenario analysis in the life cycle energy efficiency and greenhouse gas (GHG) emissions of rapeseed oil and biodiesel (rapeseed methyl ester) displacing fossil diesel. Several sources of uncertainty have been inves tigated: i) uncertainty related to parameters; ii) uncertainty concerning how co-product credits are accounted for, namely in terms of the multiple options associated with the potential uses of the co-products (rape meal and glycerin); and iii) uncertainty in terms of temporal scenarios of the impact model, which include an assessment of the implications of different time horizons for GHG emissions (20, 100 and 500 years). Energy requirements and GHG emissions have been calculated in terms of probability distributions using system expansion and three allocation approaches. Concerning how co-products are accounted for, it can be observed that results strongly depend on the substitution scheme considered. Avoided GHG emissions show considerably higher uncertainty than energy savings, mainly due to land use (nitrous oxide emissions from soil) and land use conversion (carbon stock changes). Thus, the large degree of uncertainty is mainly associated with the cultivation stage. Re sults demonstrate the relevance of applying uncertainty approaches, emphasize the need to reduce uncertainty in the environmental life cycle modeling, particularly GHG emissions calculation, and show the importance of integrating uncertainty into the interpretation of results.

Analytical Uncertainty Propagation in Life Cycle Inventory and Impact Assessment: high-efficiency versus conventional electric hand dryer and paper towel systems
Olivier Jolliet, University of Michigan
Jinglan Hong, Shandong University
Shanna Shaked, University of Michigan
Ralph Rosenbaum, CIRAIG, École Polytechnique de Montréal
Jon Dettling, Quantis
presentation

Uncertainty information is essential for the proper use of Life Cycle Assessment (LCA) and environmental assessments in decision making. So far, parameter uncertainty propagation has mainly been studied using Monte Carlo techniques that are relatively computationally heavy to conduct, especially for the comparison of multiple scenarios, often limiting its use to research or to inventory only. The present paper aims to develop and apply to both inventory and impact assessment an explicit and transparent analytical approach to uncertainty. This approach applies Taylor series expansions to the uncertainty propagation of lognormally distributed parameters and discusses its validity that is linked to the degree of lognormality of the output result. It estimates the degree of confidence in the prediction that the impact of scenario A is lower than B, accounting for correlations between input variables in different scenarios.

The approach is tested on a case study comparing a high-efficiency electric hand dryer (XLERATOR) with a conventional hand dryer and paper towels. The study indicates that the high-efficiency electric hand dryer, provides significant environmental benefits over the course of its life in comparison to the other options considered. The major cause of its advantage in comparison to conventional electric hand dryers is the reduction of the electricity consumption during the use of the dryer by nearly 4-fold. The uncertainty in the results shows that the confidence in the benefit of the high efficiency in comparison to the other systems drier is very high, with less than a one in a million chance that the opposite case is true. Under the baseline study assumptions, the paper towels show similar environmental performance to the conventional electric dryer: resulting in a slight increase or decrease, the direction of which will depend on variations in the product, its use and the assumptions of the study.

To obtain accurate uncertainty estimates, the case study shows that it is crucial to account for both common inventory processes and common impact assessment characterization factors among the different scenarios.
Overall, the analytical Taylor series expansion based on lognormal distribution was straightforward to implement in an excel spreadsheet and easily provided the explicit contributions of each parameter to the overall uncertainty.

Confronting the Uncertainties in Life-Cycle Impact Assessment for Highway Transportation Fuels
Thomas McKone, Lawrence Berkeley National Laboratory
Agnes Lobscheid, Lawrence Berkeley National Laboratory
presentation

This presentation provides a detailed case study using life-cycle emissions from highway transportation fuels and vehicles in order to explore and evaluate how uncertainty can impact decisions and how uncertainty can be addressed in life-cycle comparisons. Life-cycle impact assessment (LCIA) strives to compare one or more impacts in order to inform product choices. The reliability and consistency of decisions and policies based on LCIA are diminished by a failure to confront and communicate the inherent uncertainties. Multiple sources of uncertainty arise in any impact assessment, but are particularly problematic for comparisons. These sources include lack of knowledge in defining the problem, variation in measured data, disagreement among alternate sources of information, natural heterogeneity, the selection of one model form over another, simplifications of model structure, extrapolation errors, and value judgments. This presentation considers comparisons for the life-cycle impacts among a range of transportation fuel/vehicle alternatives—petroleum-based gasoline, tar-sands gasoline, corn-ethanol, cellulosic ethanol from a number of sources, low-sulfur diesel, and hybrid vehicles. For all major components of the systems that provide highway transportation, both the magnitude and uncertainty associated with emissions impacts are provided on a vehicle-kilometer-traveled (VKT) basis. These transportation components include

• fuel feedstock recovery/production,
• fuel production,
• vehicle operation, and
• vehicle manufacture.

The emissions considered include greenhouse-gas emissions, direct particulate matter (PM) emissions; pollutant emissions that indirectly increase PM levels, and ozone emissions from each life-stage component. These emissions are used to characterize human health impacts and compare overall impacts on human health among fuel/vehicle alternatives. Factors that are key sources of uncertainty include assumptions about the source of energy inputs for non-operational life-stages, choices about the spatial resolution applied to emissions impacts, exposure response factors for air pollutants, the health damage factor for greenhouse-gas emissions, and the choice on how to allocate impacts in time.

Incorporating Variation and Uncertainty in Strategic Life Cycle Decisions
Jeffrey Dahmus, Massachusetts Institute of Technology
Elsa Olivetti, Massachusetts Institute of Technology
Jeremy Gregory, Massachusetts Institute of Technology
Randolph Kirchain, Massachusetts Institute of Technology
presentation

As environmental sustainability rises in prominence among both consumers and firms, the role of life cycle assessment in strategic business decisions has grown. With this growth has come an increased awareness of the considerable data and resource requirements that most life cycle assessments require. The work presented here focuses on streamlining typical life cycle assessment methods, while still yielding actionable results. These results can in turn be used to guide strategic decisions that can move firms towards more environmentally sustainable products and operations.

The streamlined quantitative life cycle assessment approach explored here, comprehends both variation and uncertainty. While these issues are not new to the field of life cycle assessment, the inclusion of such factors in streamlining the life cycle assessment, namely in terms of determining what data – and what level detail of data – is collected, is unique. In addition to streamlining the life cycle assessment process, the inclusion of variation and uncertainty can also play a critical role in determining the range of possible strategic approaches available to a firm. Thus, comprehending these factors is an important component of streamlining existing life cycle assessment methods.

This work will present a streamlined quantitative life cycle assessment method, data on variation and uncertainty in a sampling of industrial processes, and a case study showing how such streamlined life cycle assessment methods, complete with variation and uncertainty data, can guide strategic firm decisions around environmental sustainability.

A method to combine simulation and approximation formulas for uncertainty calculation revisited
Andreas Ciroth, GreenDeltaTC GmbH
presentation

A combination of simulation and approximation for uncertainty calculation in LCA has been developed in a thesis 1998-2001 (Ciroth, A., Fehlerrechnung in Ökobilanzen, doctoral thesis, TU Berlin 2001; english: Ciroth, A., Fleischer, G., Steinbach, J.: Uncertainty Calculation in Life Cycle Assessments - A Combined Model of Simulation and Approximation, Int J LCA 9 (4) 216 – 226 (2004)). Back then, the method was applied for a "virtual case study", based on randomly generated data. The method claims to be faster than simulation alone and yet to be able to calculate the uncertainty as accurately, even in a looped product system. Simulation was only needed for those parts that could not be well reflected by approximation, and the method developed measures to indicate how far approximation is applicable.

The approach is now applied on a large product system from the ecoinvent database. Results are compared to those obtained by approximation formula alone and to those obtained by simulation alone, in terms of time demands and result. Specifically, thresholds proposed in the thesis will be checked. Finally, an implementation in LCA software for the combined method will be shown.