Improving the reliability of chemical manufacturing life cycle inventory constructed using secondary data

This study proposes methods to improve data mining workflows for modeling chemical manufacturing life cycle inventory. Secondary data sources can provide valuable information about environmental releases during chemical manufacturing. However, the often facility-level nature of the data challenges their utility for modeling specific processes and can impact the quality of the resulting inventory. First, a thorough data source analysis is performed to establish data quality scoring and create filtering rules to resolve data selection issues when source and species overlaps arise. A method is then introduced to develop context-based filter rules that leverage process metadata within data sources to improve how facility air releases are attributed to specific processes and increase the technological correlation and completeness of the inventory. Finally, a sanitization method is demonstrated to improve data quality by minimizing the exclusion of confidential business information (CBI). The viability of the methods is explored using case studies of cumene and sodium hydroxide production in the United States. The attribution of air releases using process context enables more sophisticated filtering to remove unnecessary flows from the inventory. The ability to sanitize and incorporate CBI is promising because it increases the sample size, and therefore representativeness, when constructing geographically averaged inventories. Future work will focus on expanding the application of context-based data filtering to other types and sources of environmental data.