Supporting Information: USEtox - The UNEP-SETAC toxicity model: recommended characterisation factors for human toxicity and freshwater ecotoxicity in Life Cycle Impact Assessment
Ralph K. Rosenbaum, Till M. Bachmann, Lois Swirsky Gold, Mark A.J. Huijbregts, Olivier Jolliet, Ronnie Juraske, Annette Koehler, Henrik F. Larsen, Matthew MacLeod, Manuele Margni, Thomas E. McKone, Jérôme Payet, Marta Schuhmacher, Dik van de Meent, Michael Z. Hauschild
Organic chemical test set
Below is the list of chemicals with CAS numbers that are part of the organic chemical test set applied to all compared models.
Table 1: Organic chemical test set
CAS# / Retained Name127-18-4 / Tetrachloroethylene
56-23-5 / Carbon tetrachloride
106-99-0 / 1,3-Butadiene
16752-77-5 / Methomyl
30560-19-1 / Acephate
50-00-0 / Formaldehyde
1336-36-3 / PCBS
117-84-0 / Di(n-octyl) phthalate
87-82-1 / Benzene, hexabromo-
52315-07-8 / Cypermethrin
2385-85-5 / Mirex
1582-09-8 / Trifluralin
115-32-2 / Dicofol
106-46-7 / p-Dichlorobenzene
309-00-2 / Aldrin
79-34-5 / 1,1,2,2-Tetrachloroethane
133-06-2 / Captan
23950-58-5 / Pronamide
120-12-7 / Anthracene
58-89-9 / gamma-Hexachlorocyclohexane
131-11-3 / Dimethyl phthalate
67-56-1 / Methanol
107-06-2 / 1,2-Dichloroethane
141-78-6 / Ethyl acetate
55-18-5 / N-Nitrosodiethylamine
137-26-8 / Thioperoxydicarbonic diamide, tetramethyl-
114-26-1 / Propoxur
133-07-3 / 1H-Isoindole-1,3(2H)-dione, 2- (trichloromethyl)thio -
17804-35-2 / Benomyl
87-68-3 / Hexachlorobutadiene
77-47-4 / Hexachlorocyclopentadiene
1024-57-3 / Heptachlor epoxide
118-74-1 / Hexachlorobenzene
76-44-8 / Heptachlor
1746-01-6 / 2,3,7,8-TCDD
3546-10-9 / Phenesterin
1163-19-5 / Decabromophenyl ether
22966-79-6 / Estradiol mustard
50-14-6 / Vitamin D2
104653-34-1 / Difethialone
1461-25-2 / Tetrabutyltin
4151-50-2 / Sulfluramid
56073-10-0 / Brodifacoum
2227-17-0 / Dienochlor
19408-74-3 / 1,2,3,7,8,9-HxCDD
USEtox model parameters
The following tables provide an overview of the landscape parameters used to characterise the different geographical scales (Table 2) as well as the minimum chemical-specific input data needed to run the model (Table 3).
Table 2: Landscape and exposure parameters
Parameter / Unit / Urban air / Continental scale / Global scaleFate parameters
Area land / km2 / 10000 / 9013369.37 / 1.41E+08
Area sea / km2 / - / 986630.63 / 3.29E+08
Fraction of area covered by fresh water / [-] / 0.03
Fraction of area covered by natural soil / [-] / 0.647 / 0.485
Fraction of area covered by agricultural soil / [-] / - / 0.485
Fraction of area paved / [-] / 0.323 / -
Mixing height air / m / 254 / 1000
Mixing depth fresh water / m / - / 2.5
Mixing depth sea water / m / - / 100
Mixing depth soil (all) / m / 0.1
Air temperature / oC / 12
Wind speed / m/s / 2.4 / 3
Rain rate / mm/yr / 700
Fraction of rain run off from soil / [-] / 0.25
Fraction of rain infiltrating soil / [-] / 0.25
Exposure parameters
Human population / - / 1.00E+07 / 3.00E+08 / 5.69E+09
Human breathing rate / m3/(person*day) / 13.3
Water ingestion / l/(person*day) / - / 1.4
Production-based intake rates
Exposed produce / kg/(day*capita) / - / 1.964
Unexposed produce / kg/(day*capita) / - / 0.193
Meat / kg/(day*capita) / - / 0.099
Dairy products / kg/(day*capita) / - / 0.273
Fish freshwater / kg/(day*capita) / - / 0.003
Fish coastal marine water / kg/(day*capita) / - / 0.050
Dairy cattle intake rates
Dairy cattle vegetation / kgFM/day / - / 59.21
Dairy cattle air / m3/day / - / 122
Dairy cattle water / kg/day / - / 60
Dairy cattle soil / kg/day / - / 0.8
Weighted intake rates for meat producing farm animals
Meat cattle vegetation / kgFM/day / - / 10.55
Meat cattle air / m3/day / - / 52.16
Meat cattle water / kg/day / - / 10.27
Meat cattle soil / kg/day / - / 0.20
Cattle meat fat content / [-] / - / 0.20
Table 3: Chemical specific input data needed to run USEtox
Minimum substance input data setOctanol-water partitioning coefficient / KOW / -
Henry's law constant at 25 oC / KH25C / Pa.m3.mol-1
Degradation rate in air / kdega / 1/s
Degradation rate in water / kdegw / 1/s
Degradation rate in soil / kdegs / 1/s
Average of log EC50 for aquatic species / avlogEC50 / mg/l
ED50 via inhalation for non-cancer effects in humans / ED50inh,noncanc / kg/lifetime
ED50 via ingestion for non-cancer effects in humans / ED50ing,noncanc / kg/lifetime
ED50 via inhalation for cancer effects in humans / ED50inh,canc / kg/lifetime
ED50 via ingestion for cancer effects in humans / ED50ing,canc / kg/lifetime
Complementary parameters used for minimum input parameter estimation
Molecular weight / MW / g/mol
Melting point / Tm / oC
Particle-water partition coefficient normalized to organic carbon / KOC / l/kg
Vapour pressure at 25 oC / Pvap25 / Pa
Solubility in water at 25 oC / Sol25 / mg/l
Biotransfer factor for meat / BTFmeat / d/kgmeat
Biotransfer factor for milk / BTFmilk / d/kgmilk
Bioaccumulation factor for fish / BAFfish / l/kgfish
Qualitative model comparison
For the qualitative model evaluation, the findings of the model framework analysis (Guinée & Hauschild 2005), of previous model comparison exercises from Fenner et al. (2005) or the OMNIITOX project (Pant et al. 2004) and the recommendations of three toxicity-related workshops previously performed under the umbrella of the Life Cycle Initiative in Lausanne (Jolliet et al. 2006), Apeldoorn (Ligthart et al. 2004), and Portland (McKone et al. 2006) were taken into account for deriving evaluation criteria. Further, the evaluation procedures and standards for defining recommended practice in LCIA as proposed by Task Force 1 of the LCIA programme were implemented into the assessment strategy for the models under study (Margni et al. 2006). A set of five leading evaluation categories was developed comprising the following general principles and indicators:
• Comprehensiveness,
• Environmental relevance, reproducibility, and model up-to-dateness,
• Scientific validity and reliability,
• Transparency, ease of understanding, and applicability,
• Compatibility with weighting and normalization.
These five major assessment categories include supplementary sub-criteria specifying particular evaluation aspects such as the question of input-data quality, model uncertainty and accuracy, and peer review under the criteria domain of scientific validity and reliability. Furthermore, criteria were partly directly related to the model matrix elements (e.g., fate and exposure model parts). Each criterion was characterized by an evaluation statement reflecting the recommendations of the above listed studies and expert workshops as well as the recommendations of the Toxic Impact Task Force members.
The final criteria list (see Box 1) was applied to and tested for all seven models under comparison and therefore represents a qualitative assessment scheme of consensus.
Box 1: Final list of qualitative evaluation criteria
ComprehensivenessThe model covers significant environmental impacts on human health, ecosystems, natural resources and man-made environment.
Environmental relevance, reproducibility and model "up-to-dateness"
General
The respective model components (i.e., fate, human exposure, human and ecosystem effects) reflect latest state of knowledge and are appropriate for their domain of validity.
Structure and intermediary results
The model considers fate, exposure and effect separately and in a quantitative way.
Mass (or concentration) in the environment, intake fraction and dose-response information are given as intermediary results.
Fate
Advection out of a region or of a continent is not considered a final loss.
Marine environment and coastal zone are differentiated and adequately treated.
Influential fate processes are considered (see main article and further specific publications)
Human exposure
Main exposure pathways are covered (inhalation, ingestion of meat, dairy products, fish, eggs, dermal uptake).
Biomagnification is included, carry over rates do comply with mass balance principles even at high Kow..
A production based in contrast to a subsistence based approach is retained, using best estimate or upper estimate of food consumption.
Human effect
Regarding potency and dose-response, benchmark doses are used avoiding safety factors, if not available extrapolation from NOAEL to LOAEL is performed, human data are preferably used.
Regarding severity and aggregation, value judgments are transparent and intermediary results are kept separately.
Ecotoxicological exposure and effect
The characterisation factor for ecotoxic effects is based on the Hazardous Concentration 50% (HC50) level rather than the HC5 or the No Observed Effect Concentration (NOEC) level, thereby relying on the most representative, not the most sensitive species.
Chronic data are preferably used over acute data as a basis for toxicity, extrapolation from acute data to chronic data is used in the absence of chronic data.
Effects are also available for sediment, terrestrial and ocean compartments.
Scientific validity and reliability
Quality of input data
Chemical input data have been carefully peer reviewed
Landscape/environment including exposure input data have been carefully verified
Uncertainty and model accuracy
Uncertainty estimates are made and reported in statistical terms.
Model results have been checked against experimental data
Peer review and documentation
The model has been published in peer reviewed journals
The model is well documented
Transparency, ease of understanding and applicability
Spatial applicability
Geographical validity (define the spatial applicability of the model)
Continental average characterization factors are available for all continents.
Data availability & applicability
In case of spare data, there is a strategy to deal with data poor chemicals in a compatible and consistent way.
The model is applicable to and covers conventional LCI data (please define the domain of application and number of chemicals covered).
Applicability to specific classes of chemicals
The model applies to metals, thereby considering aspects of speciation, essentiality, persistence, bioavailability.
The model is applicable to classical air pollutants (e.g. primary and secondary particles, NOx, SO2, CO).
The model applies to direct applications of pesticides.
Characterization factors are also available for the indoor air environment.
Compatibility with weighting and normalization
The approach is compatible with a damage approach.
Normalization method and data are available.
Weighting method (monetary, expert panel, distance to target) and data are available.
The compliance of the original models with the proposed qualitative criteria is classified as ambiguous. While for some criteria there appears to be an ‘implemented consensus’ on the state-of-the-art (e.g., the use of multimedia models of the Mackay level III type, no consideration of biomagnification, production-based exposure assessments), for other model aspects the models differed substantially. The most obvious positive and negative deviations from the proposed evaluation statements expressed as strengths and weaknesses are summarized in Table 4.
Table 4: Ambiguous compliance of the models with the qualitative evaluation criteria for human toxicity and ecotoxicity characterization models
Model / Strong point(s) / Weak point(s)CalTOX / Most encompassing in terms of exposure pathways
Advanced modelling of soil (several layers)
Monte Carlo uncertainty estimation / No severity measure for human toxicity, only partly compatible with damage approach
Ecosystem toxicity not assessed (e.g., marine environment and coastal zone not included for fate modelling)
IMPACT 2002 / Continental average characterization factors available for different global regions
Considering indoor air exposure
Direct application of pesticides considered
HC50 approach for effect modelling / Marine environment poorly represented so far
USES-LCA / Marine environment included
Global coverage, however, not spatially resolved
HC50 approach for effect modelling
One-dimensional uncertainty factors available
BETR / Flexible structure allows for spatially explicit chemical fate assessment at a variety of scales. / Chemical fate model only.
No integrated ecotoxicity assessment.
No multi-pathway human exposure assessment
EDIP / Key property based
Normalization and weighting methods provided / Mainly representative for Europe
No explicit fate results available
No severity measure for human toxicity
WATSON / European-wide spatially resolved fate, exposure and impact assessment (bottom-up analysis)
Monetary valuation as weighting method / Confined to Europe
Open system boundaries
Ecosystem toxicity not assessed
At present, confined to non-volatile compounds
EcoSense / Most reliable modelling of classical air pollutants amongst the chosen models
Bottom-up, i.e. spatially resolved, assessment capabilities for Europe, Russia, China/Asia, and Brazil/South America
Monetary valuation as weighting method / Open system boundaries
Organic chemicals mostly not considered
Only inhalation exposures with respect to toxic impacts (additionally impacts on crops and building materials)
Ecosystem toxicity not assessed
Quantitative model comparison
Below are the comparison graphs for the characterisation factors from USEtox against those from the other participating models, from the Bilthoven workshop before harmonisation (left graphs) and from the Montreal workshop after harmonisation (right graph). For human health impacts the air emission scenario is given in the main paper, as well as the freshwater emission for the aquatic ecotoxicity impacts, for the Montreal workshop results respectively. These graphs show the evolution of comparison results via harmonisation of the models.
Human health impacts
Figure 1: Comparison of characterisation factors for organics of the consensus model against the other models for human health impacts due to an emission to air. The left graph shows the first comparison from the Bilthoven workshop without any harmonisation and the right graph depicts the final result after several rounds of harmonisation from the Montreal workshop.
Figure 2: Comparison of characterisation factors for organics of the consensus model against the other models for human health impacts due to an emission to water. The left graph shows the first comparison from the Bilthoven workshop without any harmonisation and the right graph depicts the final result after several rounds of harmonisation from the Montreal workshop.
Figure 3: Comparison of characterisation factors for organics of the consensus model against the other models for human health impacts due to an emission to soil. The left graph shows the first comparison from the Bilthoven workshop without any harmonisation and the right graph depicts the final result after several rounds of harmonisation from the Montreal workshop.