Actionable insights with less data: guiding early building design decisions with streamlined probabilistic life cycle assessment

Abstract

Two obstacles that impede wider use of life cycle assessment (LCA) are its time- and data-intensiveness and the credibility surrounding its results—challenges that grow with the complexity of the product being analyzed. To guide the critical early-design stages of a complicated product like a building, it is important to be able to rapidly estimate environmental impacts with limited information, quantify the resulting uncertainty, and identify critical parameters where more detail is needed. The authors have developed the Building Attribute to Impact Algorithm (BAIA) to demonstrate the use of streamlined (not scope-limiting), probabilistic LCA for guiding the design of a building from early stages of the design process when many aspects of the design are unknown or undecided. Early-design uncertainty is accommodated through under-specification—characterizing the design using the available level of detail—and capturing the resulting variability in predicted impacts through Monte Carlo simulations. Probabilistic triage with sensitivity analyses identifies which uncertain attributes should be specified further to increase the precision of the results. The speed of the analyses allows for sequentially refining key attributes and re-running the analyses until the predicted impacts are precise enough to inform decision-making, such as choosing a preferable design alternative. Twelve design variants for a hypothetical single-family residential building are analyzed. As information is sequentially added to each variant, the significance of the difference in performance between each variant pair is calculated to determine when enough information has been added to resolve the designs (identify which design is preferable) with high confidence. At the sixth step in the analysis, all variant pairs whose mean impacts differ by at least 4% are resolvable with 90% confidence, even though only six attributes are specified and dozens of attributes remain under-specified. Furthermore, the comparative results for each variant pair are validated against a set of conventional LCA results, showing that BAIA identifies the correct preferable design among each resolvable pair at this step. Iterative specification guided by probabilistic triage can help identify promising early-design alternatives even when details are only provided for key attributes. The analysis of hypothetical design variants demonstrates that BAIA is both efficient (arrives at statistically defensible conclusions from design variant comparisons based on few pieces of information) and effective (identifies the same preferable design variants as conventional LCAs).