Reducing bias through process inventory dataset normalization

This paper explores a computational method to resolve some of the problems of external normalization in the life cycle impact assessment (LCIA) process of midpoint characterized impacts. Problems inherent to external normalization (per capita per year for a defined region) that reduce the ability to accurately calculate the most significant impact categories include a) Bias created by a range of measurement disparities b) Inverse proportion of the scale of the reference system impacts to the normalized product system impacts c) Measurement and methodological uncertainties This paper demonstrates a method called Process Inventory Dataset (PID) normalization. PID normalization modifies the normalized impact value by a normalizing factor which puts a probability distribution on average normalized impact categories for an entire process inventory dataset. PID normalization allows for significant variation of normalized impact ratio impact values among impact categories and among materials and processes. PID normalization works with incomplete process inventory and normalization data to deliver normalized impact ratio values that more accurately identify the impact categories with the most significant impacts in the LCIA process. Although PID normalization does not eliminate all of the bias that can occur from midpoint characterization and external normalization and may not reduce all uncertainties, it substantially trims the effects of normalization bias and eliminates inverse proportionality within one normalization dataset. It allows for a more accurate interpretation of normalized and weighted life cycle assessment results.