Variance-based global sensitivity analysis and beyond in life cycle assessment: an application to geothermal heating networks

Abstract Purpose Global sensitivity analysis increasingly replaces manual in life cycle assessment (LCA). Variance-based global identifies influential uncertain model input parameters by estimating so-called Sobol indices that represent each parameter’s contribution to the variance output. However, this technique can potentially be unreliable when analyzing non-normal outputs, and it does not inform analysts about specific values of or output may decision-relevant. We demonstrate three emerging methods build on variance-based provide new insights uncertainty typical LCA applications present distributions, trade-offs between environmental impacts, interactions inputs. Methods To identify inputs, trade-offs, decision-relevant interactions, we implement techniques for distribution-based (PAWN technique), spectral clustering, scenario discovery (patient rule induction method: PRIM). choose these because they are applicable with generic Monte Carlo sampling common software. compare indices, using a previously published case study geothermal heating networks. assess eight impacts under design alternatives, spanning different production temperatures network configurations. Results In application networks, PAWN generally consistently indices. some discrepancies highlight misleading interpretation distributions obtained our analysis, where meaningfully describe uncertainty. Spectral clustering highlights groups results impacts. Compared second-order interaction PRIM then provides more precise information regarding combinations associated calculated have computational advantage yield stable at relatively small sample sizes ( n = 12,000), unlike 100,000 indices). Conclusions recommend adding as give well additional regardless distribution outputs. computationally efficient sensitivities compared analysis. The combination enables precisely uncertainties outcomes