EUVIMEDEUVIMED entdecken
Systematische ÜbersichtOpen Access · CC BY 4.0

Machine learning in prospective LCA: insights from a systematic literature review

Nils Pauliks, Franco Donati, Jana Marie Weber, Bernhard Steubing

Frontiers in Sustainability · 13.08.2026 · 7

DOI: 10.3389/frsus.2026.1854730 ↗

Kurz gefasst

Systematische Übersicht in der Zeitschrift Frontiers in Sustainability, online erschienen am 13.08.2026. Verfasst von Nils Pauliks und 3 weiteren Autor:innen. Die Arbeit wurde bisher 1-mal zitiert (Crossref) und nennt 72 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: German Academic Exchange Service, Nederlandse Organisatie voor Wetenschappelijk Onderzoek.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Machine learning (ML) offers considerable opportunities for advancing life cycle assessment (LCA), yet a consolidated overview of ML models with potential for prospective LCA (pLCA) is still missing. This systematic review identifies ML models suited to pLCA's data needs and assesses the transferability of retrospective LCA models to pLCA. We apply keyword-based and automated forward/backward searches to identify studies based on algorithm type, training datasets, input/output variables, and performance metrics. We identify 50 publications and classify them into four ML application groups: (1) streamlined LCA, (2) life cycle inventory data prediction, (3) impact category prediction, and (4) characterization factor prediction. Among these, predicting life cycle inventory (LCI) data and characterization factor estimation are directly applicable to pLCA. Life cycle inventory data prediction and streamlined LCA represent the largest body of literature, offering benefits such as estimating missing unit process data, leveraging large external datasets, advanced simulation of process parameters, identifying molecular substitutes, or the employment of rapid screening tools. While characterization factor prediction shows potential, it requires further research for practical implementation. Overall, ML is rarely applied directly to pLCA. However, many ML models developed for retrospective LCA have the potential to be transferred to pLCA, either directly or with adaptations to data structures. This is most evident for predicting missing or technology-specific life cycle inventory data, where similarity-based learning, simulation-enhanced models, and large datasets offer scalable ways to estimate future inventory parameters. Across all groups, uncertainty is insufficiently assessed, with only a minority of studies addressing model, parameter, or scenario uncertainty. Overall, ML holds potential for pLCA when applied with methodological care and attention to underlying uncertainties.

Förderung

  • German Academic Exchange Service
  • Nederlandse Organisatie voor Wetenschappelijk Onderzoek

Zitieren

APA

Pauliks, N., Donati, F., Weber, J. M., & Steubing, B. (2026). Machine learning in prospective LCA: insights from a systematic literature review. Frontiers in Sustainability, 7. https://doi.org/10.3389/frsus.2026.1854730

Vancouver

Pauliks N, Donati F, Weber JM, Steubing B. Machine learning in prospective LCA: insights from a systematic literature review. Frontiers in Sustainability. 2026;7. doi:10.3389/frsus.2026.1854730

BibTeX
@article{Pauliks2026,
  author = {Pauliks, Nils and Donati, Franco and Weber, Jana Marie and Steubing, Bernhard},
  title = {Machine learning in prospective LCA: insights from a systematic literature review},
  journal = {Frontiers in Sustainability},
  year = {2026},
  volume = {7},
  doi = {10.3389/frsus.2026.1854730}
}
Kostenlos starten →