European Risk Model Comparison Study

1
European Risk ModelComparison Study
Sponsored by NICOLE

• Terry Walden, BP
• Lawrence Houlden, Archon
• Matt Gardner, Arcadis

2
Acknowledgements

• Sponsors

• Akzo Nobel
• BNFL
• BP
• Fortum
• ICI
• JM Bostad

• NICOLE
• Powergen
• SecondSite Property
• Shell Global Solutions
• Solvay
• Total

Peer Review Team

• SKB, Netherlands
• Kemakta, Sweden

• RIVM, Netherlands
• VITO, Belgium

Consultant

• Arcadis Geraghty and Miller, UK

3
Reasons for Study

• NICOLE and CLARINET advocate risk-based approach
  to land management, but
• Many member states develop own models
• Differences in model results can be orders of
  magnitude
• Poor understanding of differences may undermine
  credibility of risk assessment

4
Objectives

• Compare human health risk models used in Europe
• Examine model results to explain output
  differences - not to show which is better
• Generic site with standardised inputs
• Real test cases using model defaults
• Limit comparison to dose or receptor point
  concentrations
• Risk comparison avoided because of different ways
  countries handle toxicity (esp. cancer)

5
Countries and Models

• Austria No model
• Belgium (Flanders) Vlier-Humaan
• Denmark JAGG
• Finland No model
• France No model
• Germany UMS SISIM

6
Countries and Models (2)

• Greece No model
• Ireland No model
• Italy Guiditta ROME
• Luxembourg No model
• Netherlands HESP SUS Risc-Human
• Norway SFT 9906

7
Countries and Models (3)

• Portugal No model
• Spain LUR (Basque Country)
• Sweden Report 4639
• Switzerland Transim
• UK Consim RAM P20 CLEA
• Commercial RISC RBCA Toolkit

8
Soil Ingestion (Generic Site)
Cadmium Relative Dose (normalised to Vlier-Humaan)
9
Soil Ingestion Models

• All models have essentially the same soil
  ingestion algorithms
• In Vlier-Humaan, soil ingestion rates are fixed
  (locked cell) at relatively low values
• CLEA uses hard-wired probabilistic exposure at
  95 level exposure 4x higher than most
  models

10
Dermal Contact (Generic Site)
BaP Relative Dose (normalised to Risc-Human)
11
Dermal Contact Models

• CLEA has smaller dose as contaminant is allowed
  to absorb plus volatilise from skin
• Vlier- Risc-Human limits exposure to 2 hrs/day,
  reflecting skin permeability
• Risc-Human is very low because its soil-on-skin
  adherence is fixed 10x lower than that in other
  models

12
Soil to Indoor Air
Benzene concentrations in mg/m3
46
0.07
Note UMS concentration is 650x higher than RBCA
13
Indoor Air Soil Algorithms

• Flow thru cracks
• Flow thru concrete matrix
• Concrete weathering
• Indoor air 1 of soil gas
• User input for soil gas intrusion

• RISC RBCA Toolkit
• Vlier- Risc-Humaan
• JAGG
• UMS
• SFT 9906

RBCA allows for source depletion and averages
concentration over exposure duration
14
Generic Site Conclusions

• Soil ingestion and groundwater migration models
  are all similar (one order magnitude)
• Vegetable ingestion model results surprisingly
  uniform (one order magnitude)
• Dermal contact models more variable (two orders
  magnitude)
• Indoor air models, particularly UMS code, have
  highest variability (3 orders magnitude)
• Differences attributed to fixed parameters
  (locked cells) or algorithms (indoor air)

15
Test Cases

• Lube plant with chlorinated plume
• Will show predicted vs. actual GW conc.
• Manufactured gas plant with PAHs in soil
• Will show soil ingestion results vs. generic site
• Fly ash landfill with heavy metals
• Chemical plant with chlorinateds pesticides in
  soil
• Petrol filling station with BTEX MTBE

16
Soil Ingestion Generic vs. Test Case
Relative Doses BaP Soil Ingestion
750
17
Overall Conclusions

• Consistent defensible results possible where fate
  transport / chemical / exposure parameters well
  understood
• Where model defaults are used, significant
  differences (3 orders magnitude) can occur, even
  for simple soil ingestion case
• Risk harmonisation in Europe will require
  research on algorithms (dermal contact and indoor
  air) better standardisation of exposure
  parameters