A review of inventory modeling methods for missing data in life cycle assessment

Missing data is the key challenge facing life cycle inventory (LCI) modeling. The collection of missing data can be cost-prohibitive and infeasible in many circumstances. Major strategies to address this issue include proxy selection (i.e., selecting a surrogate dataset to represent the missing data) and data creation (e.g., through empirical equations or mechanistic models). Within these two strategies, we identified three approaches that are widely used for LCI modeling: Data-driven, mechanistic, and future (e.g., 2050) inventory modeling. We critically reviewed the 12 common methods of these three approaches by focusing on their features, scope of application, underlying assumptions, and limitations. These methods were characterized based on the following criteria: domain knowledge requirement (both as a method developer and a user), post-treatment requirement, challenge in assessing data quality uncertainty, challenge in generalizability, and challenge in automation. These criteria can be used by LCA practitioners to select the suitable method(s) to bridge the data gap in LCI modeling, based on the goal and scope of the intended study. We also identified several aspects for future improvement for these reviewed methods.

Automated summary

Missing data is a key challenge in life cycle inventory (LCI) modeling. This study critically reviews 12 common methods for addressing missing data in LCI modeling, analyzes their features, scope, assumptions, and limitations, and identifies areas for future improvement.