Background, aim, and scope
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. Furthermore, Monte Carlo simulations do not automatically assess the sensitivity and contribution to overall uncertainty of individual parameters. 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.Materials and methods
We first apply the Taylor series expansion method to analyze the uncertainty propagation of a single scenario, in which case the squared geometric standard deviation of the final output is determined as a function of the model sensitivity to each input parameter and the squared geometric standard deviation of each parameter. We then extend this approach to the comparison of two or more LCA scenarios. Since in LCA it is crucial to account for both common inventory processes and common impact assessment characterization factors among the different scenarios, we further develop the approach to address this dependency. We provide a method to easily determine a range and a best estimate of (a) the squared geometric standard deviation on the ratio of the two scenario scores, “A/B”, and (b) the degree of confidence in the prediction that the impact of scenario A is lower than B (i.e., the probability that A/B<1). The approach is tested on an automobile case study and resulting probability distributions of climate change impacts are compared to classical Monte Carlo distributions.Results
The probability distributions obtained with the Taylor series expansion lead to results similar to the classical Monte Carlo distributions, while being substantially simpler; the Taylor series method tends to underestimate the 2.5% confidence limit by 1-11% and the 97.5% limit by less than 5%. The analytical Taylor series expansion easily provides the explicit contributions of each parameter to the overall uncertainty. For the steel front end panel, the factor contributing most to the climate change score uncertainty is the gasoline consumption (>75%). For the aluminum panel, the electricity and aluminum primary production, as well as the light oil consumption, are the dominant contributors to the uncertainty. The developed approach for scenario comparisons, differentiating between common and independent parameters, leads to results similar to those of a Monte Carlo analysis; for all tested cases, we obtained a good concordance between the Monte Carlo and the Taylor series expansion methods regarding the probability that one scenario is better than the other.Discussion
The Taylor series expansion method addresses the crucial need of accounting for dependencies in LCA, both for common LCI processes and common LCIA characterization factors. The developed approach in Eq. 8, which differentiates between common and independent parameters, estimates the degree of confidence in the prediction that scenario A is better than B, yielding results similar to those found with Monte Carlo simulations.Conclusions
The probability distributions obtained with the Taylor series expansion are virtually equivalent to those from a classical Monte Carlo simulation, while being significantly easier to obtain. An automobile case study on an aluminum front end panel demonstrated the feasibility of this method and illustrated its simultaneous and consistent application to both inventory and impact assessment. The explicit and innovative analytical approach, based on Taylor series expansions of lognormal distributions, provides the contribution to the uncertainty from each parameter and strongly reduces calculation time. 相似文献Objective uncertainty quantification (UQ) of a product life-cycle assessment (LCA) is a critical step for decision-making. Environmental impacts can be measured directly or by using models. Underlying mathematical functions describe a model that approximate the environmental impacts during various LCA stages. In this study, three possible uncertainty sources of a mathematical model, i.e., input variability, model parameter (differentiate from input in this study), and model-form uncertainties, were investigated. A simple and easy to implement method is proposed to quantify each source.
MethodsVarious data analytics methods were used to conduct a thorough model uncertainty analysis; (1) Interval analysis was used for input uncertainty quantification. A direct sampling using Monte Carlo (MC) simulation was used for interval analysis, and results were compared to that of indirect nonlinear optimization as an alternative approach. A machine learning surrogate model was developed to perform direct MC sampling as well as indirect nonlinear optimization. (2) A Bayesian inference was adopted to quantify parameter uncertainty. (3) A recently introduced model correction method based on orthogonal polynomial basis functions was used to evaluate the model-form uncertainty. The methods are applied to a pavement LCA to propagate uncertainties throughout an energy and global warming potential (GWP) estimation model; a case of a pavement section in Chicago metropolitan area was used.
Results and discussionResults indicate that each uncertainty source contributes to the overall energy and GWP output of the LCA. Input uncertainty was shown to have significant impact on overall GWP output; for the example case study, GWP interval was around 50%. Parameter uncertainty results showed that an assumption of ±?10% uniform variation in the model parameter priors resulted in 28% variation in the GWP output. Model-form uncertainty had the lowest impact (less than 10% variation in the GWP). This is because the original energy model is relatively accurate in estimating the energy. However, sensitivity of the model-form uncertainty showed that even up to 180% variation in the results can be achieved due to lower original model accuracies.
ConclusionsInvestigating each uncertainty source of the model indicated the importance of the accurate characterization, propagation, and quantification of uncertainty. The outcome of this study proposed independent and relatively easy to implement methods that provide robust grounds for objective model uncertainty analysis for LCA applications. Assumptions on inputs, parameter distributions, and model form need to be justified. Input uncertainty plays a key role in overall pavement LCA output. The proposed model correction method as well as interval analysis were relatively easy to implement. Research is still needed to develop a more generic and simplified MCMC simulation procedure that is fast to implement.
相似文献Several models are available in the literature to estimate agricultural emissions. From life cycle assessment (LCA) perspective, there is no standardized procedure for estimating emissions of nitrogen or other nutrients. This article aims to compare four agricultural models (PEF, SALCA, Daisy and Animo) with different complexity levels and test their suitability and sensitivity in LCA.
MethodsRequired input data, obtained outputs, and main characteristics of the models are presented. Then, the performance of the models was evaluated according to their potential feasibility to be used in estimating nitrogen emissions in LCA using an adapted version of the criteria proposed by the United Nations Framework Convention on Climate Change (UNFCCC), and other relevant studies, to judge their suitability in LCA. Finally, nitrogen emissions from a case study of irrigated maize in Spain were estimated using the selected models and were tested in a full LCA to characterize the impacts.
Results and discussionAccording to the set of criteria, the models scored, from best to worst: Daisy (77%), SALCA (74%), Animo (72%) and PEF (70%), being Daisy the most suitable model to LCA framework. Regarding the case study, the estimated emissions agreed to literature data for the irrigated corn crop in Spain and the Mediterranean, except N2O emissions. The impact characterization showed differences of up to 56% for the most relevant impact categories when considering nitrogen emissions. Additionally, an overview of the models used to estimate nitrogen emissions in LCA studies showed that many models have been used, but not always in a suitable or justified manner.
ConclusionsAlthough mechanistic models are more laborious, mainly due to the amount of input data required, this study shows that Daisy could be a suitable model to estimate emissions when fertilizer application is relevant for the environmental study. In addition, and due to LCA urgently needing a solid methodology to estimate nitrogen emissions, mechanistic models such as Daisy could be used to estimate default values for different archetype scenarios.
相似文献Life Cycle Assessment (LCA) is the process of systematically assessing impacts when there is an interaction between the environment and human activity. Machine learning (ML) with LCA methods can help contribute greatly to reducing impacts. The sheer number of input parameters and their uncertainties that contribute to the full life cycle make a broader application of ML complex and difficult to achieve. Hence a systems engineering approach should be taken to apply ML in isolation to aspects of the LCA. This study addresses the challenge of leveraging ML methods to deliver LCA solutions. The overarching hypothesis is that: LCA underpinned by ML methods and informed by dynamic data paves the way to more accurate LCA while supporting life cycle decision making.
MethodsIn this study, previous research on ML for LCA were considered, and a literature review was undertaken.
ResultsThe results showed that ML can be a useful tool in certain aspects of the LCA. ML methods were shown to be applied efficiently in optimization scenarios in LCA. Finally, ML methods were integrated as part of existing inventory databases to streamline the LCA across many use cases.
ConclusionsThe conclusions of this article summarise the characteristics of existing literature and provide suggestions for future work in limitations and gaps which were found in the literature.
相似文献It is frequently mentioned in literature that LCA is linear, without a proof, or even without a clear definition of the criterion for linearity. Here we study the meaning of the term linear, and in relation to that, the question if LCA is indeed linear.
MethodsWe explore the different meanings of the term linearity in the context of mathematical models. This leads to a distinction between linear functions, homogeneous functions, homogenous linear functions, bilinear functions, and multilinear functions. Each of them is defined in accessible terms and illustrated with examples.
ResultsWe analyze traditional, matrix-based, LCA, and conclude that LCA is not linear in any of the senses defined.
Discussion and conclusionsDespite the negative answer to the research question, there are many respects in which LCA can be regarded to be, at least to some extent, linear. We discuss a few of such cases. We also discuss a few practical implications for practitioners of LCA and for developers of new methods for LCI and LCIA.
相似文献In support of the sustainable development of our societies, future engineers should have elementary knowledge in sustainability assessment and use of life cycle assessment. Publications on pedagogical experience with teaching life cycle assessment (LCA) in high-level education are however scarce. Here, we describe and discuss 20 years of experience in teaching LCA at MSc level in an engineering university with the ambition to share our insights and inspire teaching of LCA as part of a university curriculum.
MethodsWe detail the design of an LCA course taught at the Technical University of Denmark since 1997. The course structure relies on (i) a structured combination of theoretical teaching, practical assignments and hands-on practice on LCA case studies, and (ii) the conduct of real-life LCA case studies in collaboration with companies or other organisations. Through the semester-long duration of the course, students from different engineering backgrounds perform full-fledged LCA studies in groups, passing through two iterations—a screening LCA supporting a more targeted LCA.
Results and discussionThe course design, which relies on a learning-by-doing principle, is transparently described to inspire LCA teachers among the readers. Historical evolution and statistics about the course, including its 192 case studies run in collaboration with 105 companies and institutions, are analysed and serve as basis to discuss the benefits and challenges of its different components, such as the theory acquisition, the assignment work, the LCA software learning, the conduct of case studies, the merits of industrial collaborations and grading approaches.
ConclusionsWe demonstrate the win-win situation created by the setting of the course, in which the students are actively engaged and learn efficiently how to perform an LCA while the collaborating companies often get useful insights into their analysed case studies. The course can also be an eye opener for companies unfamiliar with LCA, who get introduced to life cycle thinking and the potential benefits of LCA. We have no hesitation in recommending industries and LCA teachers to engage into such collaborations even in the fundamental teaching of LCA techniques.
相似文献Purpose
The analysis of uncertainty in life cycle assessment (LCA) studies has been a topic for more than 10 years, and many commercial LCA programs now feature a sampling approach called Monte Carlo analysis. Yet, a full Monte Carlo analysis of a large LCA system, for instance containing the 4,000 unit processes of ecoinvent v2.2, is rarely carried out by LCA practitioners. One reason for this is computation time. An alternative faster than Monte Carlo method is analytical error propagation by means of a Taylor series expansion; however, this approach suffers from being explained in the literature in conflicting ways, hampering implementation in most software packages for LCA. The purpose of this paper is to compare the two different approaches from a theoretical and practical perspective.Methods
In this paper, we compare the analytical and sampling approaches in terms of their theoretical background and their mathematical formulation. Using three case studies—one stylized, one real-sized, and one input–output (IO)-based—we approach these techniques from a practical perspective and compare them in terms of speed and results.Results
Depending on the precise question, a sampling or an analytical approach provides more useful information. Whenever they provide the same indicators, an analytical approach is much faster but less reliable when the uncertainties are large.Conclusions
For a good analysis, analytical and sampling approaches are equally important, and we recommend practitioners to use both whenever available, and we recommend software suppliers to implement both. 相似文献Product systems use the same unit process models to represent distinct but similar activities. This notably applies to activities in cyclic dependency relationships (or “feedback loops”) that are required an infinite number of times in a product system. The study aims to test the sensitivity of uncertainty results on the assumption made concerning these different instances of the same activities. The default assumption assumes homogeneous production, and the same parameter values are sampled for all instances (e.g., there is one truck). The alternative assumption is that every instance is distinct, and parameter values are independently sampled for different instances of unit processes (e.g., there are infinitely many trucks). Intuitively, sampling the same values for each instance of a unit process should result in more uncertain results.
MethodsThe results of uncertainty analyses carried out under either assumption are compared. To simulate models where each instance of a unit process is independent, we convert network models to acyclic LCI models (tree models). This is done three times: (1) for a very simple product system, to explain the methodology; (2) for a sample product system from the ecoinvent database, for illustrative purposes; and (3) for thousands of product systems from ecoinvent databases.
Results and discussionThe uncertainty of network models is indeed greater than that of corresponding tree models. This is shown mathematically for the analytical approximation method to uncertainty propagation and is observed for Monte Carlo simulations with very large numbers of iterations. However, the magnitude of the difference in indicators of dispersion is, for the ecoinvent product systems, often less than a factor of 1.5. In few extreme cases, indicators of dispersion are different by a factor of 4. Monte Carlo simulations with smaller numbers of iterations sometimes give the opposite result.
ConclusionsGiven the small magnitude of the difference, we believe that breaking away from the default approach is generally not warranted. Indeed, (1) the alternative approach is not more robust, (2) the current default approach is conservative, and (3) there are more pressing challenges for the LCA community to meet. This being said, the study focused on ecoinvent, which should normally be used as a background database. The difference in dispersion between the two approaches may be important in some contexts, and calculating the uncertainty of tree models as a sensitivity analysis could be useful.
相似文献