es501847p_si_001.pdf (1.74 MB)
Reproducibility of LCA Models of Crude Oil Production
journal contribution
posted on 2014-11-04, 00:00 authored by Kourosh Vafi, Adam R. BrandtScientific
models are ideally reproducible, with results that converge
despite varying methods. In practice, divergence between models often
remains due to varied assumptions, incompleteness, or simply because
of avoidable flaws. We examine LCA greenhouse gas (GHG) emissions
models to test the reproducibility of their estimates for well-to-refinery
inlet gate (WTR) GHG emissions. We use the Oil Production Greenhouse
gas Emissions Estimator (OPGEE), an open source engineering-based
life cycle assessment (LCA) model, as the reference model for this
analysis. We study seven previous studies based on six models. We
examine the reproducibility of prior results by successive experiments
that align model assumptions and boundaries. The root-mean-square
error (RMSE) between results varies between ∼1 and 8 g CO2 eq/MJ LHV when model inputs are not aligned. After model
alignment, RMSE generally decreases only slightly. The proprietary
nature of some of the models hinders explanations for divergence between
the results. Because verification of the results of LCA GHG emissions
is often not possible by direct measurement, we recommend the development
of open source models for use in energy policy. Such practice will
lead to iterative scientific review, improvement of models, and more
reliable understanding of emissions.