(732e) Data-Driven Life Cycle Assessment Accounting Uncertainties | AIChE

(732e) Data-Driven Life Cycle Assessment Accounting Uncertainties


Li, Y. - Presenter, National University of Singapore
Sun, Z., National University of Singapore
Wang, X., National University of Singapore
Ideally, primary data collection is recommended for every life cycle assessment (LCA) study. However, due to limited availability or accessibility to first-hand data, related sources of secondary data can be a good alternative in practice. In this work, various sources of uncertainty in using secondary data from the ecoinvent Life Cycle Inventory (LCI) database [1] are illustrated with an LCA case study on global air travel.

Among the three linked system models ecoinvent provides [2], the cut-off model is used in this work with raw data retrieved from the database to build technology and intervention matrices [3]. Inside the database, the basic uncertainty associated with parameters’ intrinsic variability can be modeled with one of the seven different statistical distributions provided by the database while the additional uncertainty associated with imperfect data quality is quantified with the five criteria in pedigree matrix. The two types of uncertainty can be combined via the pedigree approach [4] and then assessed with a Monte Carlo simulation. Given the recent effort of ecoinvent to derive uncertainty factors for the pedigree matrix based on empirical studies [5], the effect of updated pedigree matrix coefficients is also evaluated.

Furthermore, since unit processes in the database provide information on gate-to-gate LCIs, analysis on emission hotspots along the supply chain of air travel can be conducted by recursively expanding upstream processes until desired level of details is achieved. Such analysis also quantifies the sensitivity with respect to the choice of system boundary and helps to prune away insignificant processes during subsequent data collection and/or refinery efforts.

When using ecoinvent as a source of secondary data for LCA studies on real world processes, uncertainty arises not only from the data and parameters inside the database, it can also come from the imperfect mapping between processes available in the database and the real-world process of interest. Concretely, for the LCA case study on air travel, flight specific parameters, e.g., plant type and occupancy level, are not explicitly modeled in the database. Instead, global averages or fixed-value assumptions are used, which could introduce additional uncertainty when using ecoinvent to calculate CO2 emissions for a real flight. To combat this issue, the International Civil Aviation Organizations (ICAO) carbon emissions calculator [6] can be adopted, but it also suffers from the following limitations. Firstly, the ICAO calculator only provides direct CO2 emissions during air travel. Other types of environmental flows and indirect emissions are not considered. Secondly, only a single value is estimated by the calculator without any uncertainty quantification. Therefore, a two-step method is proposed to leverage on the advantages of both sources. In particular, the more accurate data from ICAO calculator is used for fuel consumption while the more comprehensive ecoinvent database is used for calculating LCIs (including but not limited to fossil CO2) per unit fuel consumed with uncertainty evaluated from ratio estimators using the technique of Monte Carlo simulation [7]. The developed framework can be generally used for any process when conducting LCA, with specific results in this work informing global air travel related life cycle impact.


[1] Wernet G, Bauer C, Steubing B, Reinhard J, Moreno-Ruiz E, Weidema B. The ecoinvent database version 3 (part I): overview and methodology. Int J Life Cycle Assess 2016;21:1218–30. https://doi.org/10.1007/s11367-016-1087-8.

[2] ecoinvent. System Models in ecoinvent 3 2013. https://www.ecoinvent.org/database/system-models-in-ecoinvent-3/system-m... (accessed April 10, 2020).

[3] Guillaume B. Building matrices from spold files. 2013.

[4] Muller S, Lesage P, Ciroth A, Mutel C, Weidema BP, Samson R. The application of the pedigree approach to the distributions foreseen in ecoinvent v3. Int J Life Cycle Assess 2016;21:1327–37. https://doi.org/10.1007/s11367-014-0759-5.

[5] Ciroth A, Muller S, Weidema B, Lesage P. Empirically based uncertainty factors for the pedigree matrix in ecoinvent. Int J Life Cycle Assess 2016;21:1338–48. https://doi.org/10.1007/s11367-013-0670-5.

[6] International Civil Aviation Organizations (ICAO). ICAO Carbon Emissions Calculator 2016. https://www.icao.int/environmental-protection/Carbonoffset/Pages/default... (accessed November 30, 2020).

[7] Art B. Owen. Monte Carlo theory, methods and examples. 2013.