Towards consensus in chemical characterization modeling for LCA: comparison and harmonization of models for human exposure and toxicity

A comprehensive LCIA characterization model comparison is being undertaken in the UNEP/SETAC Life Cycle Initiative, focusing on toxicity impacts and directly involving the developers of all models included. The main objective is to identify where differences come from, what indispensable model components are and if there can be a consensus model built from them, leading towards recommended practice in chemical characterization for LCIA. The models were selected in an open process inviting all models identified to be capable of characterizing a chemical in terms of environmental fate, human exposure, human toxicity and ecotoxicity. The invitation was accepted by the developers of CalTOX, IMPACT 2002, USES-LCA, EDIP, WATSON, and EcoSense. A consistent chemical test set comprising 66 organic (generic, amphiphilic and dissociating) and inorganic (metals, salts) compounds was selected representing a wide range of substance property combinations. All compared models showed correlation for human health endpoints for generic organics, with high variations on individual chemicals, typically with high Kow. For the other organics and inorganics, less agreement was observed. Influential processes and assumptions were identified and agreed upon to implement in all models for harmonization. These were, e.g., an urban box nested in a continental box with fixed surfaces and populations, consistent biotransfer and –concentration factors from experiments or one source/model, vegetation as an exposure pathway is determined by air-plant and soil-plant BCF correlations. For human toxicity, safety factors are avoided, directly using the TD50 benchmark dose with an applied slope on the dose response curve. Human data are preferred and animal-human extrapolation is done using allometrically based factors. Route-to-route extrapolation options were also explored.