SHE-Net - Outcome of the third workshop

Risk assessment: from contaminated soils to humans via the foodchain

The third and final SHE-NET workshop was held at the University of Nottingham on 2nd and 3rd July, 2008. The objective of the workshop was to consider current approaches to risk assessment modelling for organic contaminants in soils, with an emphasis on human exposure.The focus was principally on exposure assessment modelling, organised into 3 themes:

  1. models and methods
  2. science underpinning the models and uncertainties
  3. model intercomparisons

There were twenty participants at the workshop. In addition to the organising committee, contributors were drawn from five UK universities, three contaminated land consultancies in Sweden and the UK, and from three national research organisations in the USA, Belgium and the UK.

Introductory Presentations

Four brief presentations were given by the members of the organising committee. Two of these (presented by Chris Collins and Lindsay Bramwell) summarised the major conclusions of the previous two workshops.

George Shaw then introduced the theme of the workshop, describing how the rigorous methodology of Quantitative Risk Assessment (QRA) can establish objective thresholds, such as soil guideline values (SGVs), for contaminants, but with the major caveat that uncertainties still remain resulting from imperfect understanding of relevant pathways, mechanisms and parameters, model representation of specific pathways (e.g. plant uptake), toxicology of many of the substances of concern and availability of data for model validation.

Raquel Duarte-Davidson then provided a perspective on the current status of risk assessment for soil contaminants from the point of view of the UK Health Protection Agency (Duarte-Davidson.ppt). This emphasised the use of a tiered risk assessment approach from simple risk screening, through generic risk assessment to detailed site- and case-specific QRA. The presentation considered what determines how sophisticated a risk assessment should be and the appropriate selection of risk assessment methodologies. It was concluded that, in identifying key research needs, both deterministic and probabilistic approaches should be encompassed to ensure that the method used is fit for purpose.

Following the introductory presentations, eight invited speakers provided talks on specific topics under the three workshop themes.

Theme 1: Models and Methods

Current challenges in risk assessment modelling for organic contaminants.

Paul Nathanail (Land Quality Management, UK)

This presentation described how QRA can possibly only provide a qualitative insight based on quantitative analysis, and emphasised that QRA modelling was always likely to be less desirable than making direct measurements ('Measure if you can, model if you must.'), an example being the recent trend in the USA to measure human exposure to volatile organic compounds rather than modelling (e.g. using the Johnson-Ettinger model). In support of risk assessment, it was stated that while QRA is an expensive process it is much less expensive than risk management or mitigation, which can be avoided if appropriate QRA is used. There is a trend in the UK towards the use of deterministic exposure models, but this can lead to 'creeping conservatism' since explicit representation of parameter uncertainty is replaced by selection of possibly over-pessimistic values (e.g. of soil ingestion). The forthcoming Soil Framework Directive will provide an opportunity for harmonisation of QRA across Europe.

Modelling to derive guideline concentrations for organic contaminants in soils in Sweden.

Mark Elert (Kemakta Konsult, Sweden)

The development of SGVs for organic contaminants in Swedish soils was described.These values represent 'no risk' thresholds with respect to human health and environmental protection and are designed to be protective of most sites in Sweden.They are derived using generic methods and models largely based on European (Dutch) and US EPA models.The newest version of the Soil Criteria Model will be Excel based and openly available for public use, which raises the question of what level of expertise is needed to draw reliable conclusions from risk modelling.A recent change in emphasis is towards less conservative (more realistic) estimates of exposure but more stringent limits for health effects and release.

Exposure assessment using the CLEA model.

Ian Martin (Environment Agency, UK)

This presentation described the tiered risk-based approach used by the CLEA model to predict likely exposure of humans to organic contaminants in soils. In particular it focussed on the question of what type of produce is eaten in the UK (based on the Expenditure and Food Survey) and what proportion of this is homegrown compared with other countries. Also covered were the questions of how much land area is required to provide homegrown foods, plus the uncertainties of soil adhesion to vegetables and the effects of food preparation on contaminant concentrations. Along with the unknown impact of climate change and population migration on food production and consumption in the UK, these topics were identified as gaps in our current understanding and, therefore, targets for future research.

Risk assessment, multi-criteria decision analysis, and adaptive management techniques for addressing model uncertainty and reliability.

Patrick Deliman (US Army Engineer Research and Development Center, USA)

This talk focussed on the use of adaptive risk-based planning to guide projects such as the restoration of contaminated sites (e.g. post Hurricane Katrina damage to coastal Louisiana). Risks and benefits associated with alternative management strategies are difficult to quantify and it is important to consider uncertainties and variability of models, parameters and scenarios associated with management options (as well as stakeholder value judgments). Several examples were used to describe how risk assessment and planning for situations with a limited knowledge base and high uncertainty/variability require coupling of traditional risk assessment and planning with multi-criteria decision analysis (MCDA) to support regulatory decision making.

Theme 2: Science Underpinning the Models and Uncertainties

Mechanistic models in exposure assessment modelling.

Piet Seuntjens (VITO, Belgium)

This presentation addressed the questions of whether we need mechanistic models in exposure assessments, how mechanistic models can be incorporated in exposure assessments and why shouldn't we measure exposure instead of modelling? It was concluded that mechanistic models are more generic than empirical models and allow extrapolations from one situation to another. They also help to underpin or derive simpler empirical models. Mechanistic models in tiered risk assessments allow extended impact analysis with deeper insights into causes as well as effects, though there is still a role for models of reduced complexity (see Neil Crout's presentation). Finally, model development and exposure measurement should both be part of an iterative process, building up confidence in risk assessments for specific cases.

Systematic model simplification – benefits and approaches.

Neil Crout (University of Nottingham, UK)

Following on from the previous presentation, this talk addressed the issue of model complexity and how this could lead to over-parameterisation (especially when suitable data for model verification are often difficult to find). Even in the case of apparently mechanistic models, greater complexity may not result in greater realism. It was emphasised that we need to know whether a model's formulation really can be supported by observations. Reduction in model complexity, using formalised methods, can reduce uncertainty in models, provides a way to test model formulation and tests the understanding built into our models.

Conceptual site model development.

Judith Nathanail (Land Quality Management, UK)

One of the legal cornerstones of risk assessments of contaminated land in the UK is the question of whether 'pollutant linkages' exist at specific sites. Rigorous and formalised site descriptions – Conceptual Site Models (CSMs) – need to be constructed to establish the existence of pollutant linkages with confidence. A new software-based approach to construct CSMs was described, which can be used to construct complex matrices of pollutant linkages (together with associated uncertainties) to be evaluated further using exposure and QRA models.

Theme 3: Model Intercomparison

European risk model comparison study.

Lawrence Houlden (Archon Environmental, UK)

Many exposure and QRA models have been written and are currently in use in different countries around the world. One of the questions arising from the first SHE-NET workshop was whether it is possible to develop one core model? One way to answer this is to compare the performance of the many models available using a standard intercomparison scenario. This presentation described the (to date) most complex and extensive model intercomparison of human health risk models, undertaken by the NICOLE group. Models (or in the absence of models, 'methods') from 17 countries were compared, leading to the conclusions that results were comparable for cases where fate and transport of a specific substance are well understood, but that 'generic' parameters can lead to large disagreements between model predictions (ie. uncertainties) and that test sites and cases suitable for model testing are limited.

The final conclusion was that risk managers need critically to assess model assumptions and how software is applied, which echoes the question raised earlier in the workshop by Mark Elert as to the level of expertise needed to draw reliable conclusions from risk modelling.

Open Forum Discussion

The workshop concluded with a 90 minute open forum discussion based on a series of topics and questions raised during the meeting.General questions raised by individual contributors were as follows.

  • How reliable are models in the hands of 'non-expert' users?
  • Is there any evidence of harm to humans in non-occupational exposure (e.g. background dioxin exposure in Sweden > TDI)
  • What is 'background' – new versus old (legacy) contamination?
  • Habit / diet data – do we know enough, especially given changing climatic and population trends?
  • Balance of benefits – nutrition vs. exposure risk in home-grown vegetables.
  • Mechanistic vs. empirical or 'reduced' models – what are the benefits and what is the problem being solved?
  • Model complexity vs. uncertainty – an inverse relationship?
  • How to use models in exposure assessments?
  • Making complex decisions – does common sense always make sense? Are there any surprises, ie. 'unknown unknowns'?
  • Are the modelling tools available 'fit for purpose'? How do we judge from intercomparison studies?
  • Conservatism in models – is this an 'insurance policy' or should ever improving accuracy be sought? How near to the cliff edge should we aim to stand?
  • All models are wrong - models can only be validated in experiments in which exposure is lethal, which are impossible to carry out.
  • Auditing in CSM development and exposure assessment.

In addition, several wider and more general questions were raised, as follows.

  • 'Measure if you can, model if you must' (attributed to Colin Ferguson, originator of the CLEA model).
  • OPINION OF THE 'REGULATORS' Need clear audit trail, defensible conclusions and means of making complex decisions.
  • OPINION OF THE 'PRACTITIONERS' Need to be satisfied they are 'getting it right', but 'how right'? Accreditation and duty of care in QRA.
  • OPINION OF THE 'RESEARCHERS' What needs to be done to improve exposure assessment? Better use of existing information in models and care needed when models 'borrow' the same algorithms from each other.
  • DO WE NEED BETTER MODELS, OR BETTER DATA? New chemicals, newly identified synergistic effects, new perceptions, new crops/foodstuffs.
  • CAN WE HARMONISE MODELS, especially with respect to key algorithms and physico-chemical parameters?

Workshop Conclusions

Modelling of human exposures to organic contaminants in soils is less desirable and defensible than making direct measurements, though measurement is not possible in many (perhaps most) cases. Furthermore, modelling allows risks to be evaluated before exposures have occurred (e.g. in 'what if' scenarios), whereas measurement may only indicate significant exposures after they have occurred. QRA modelling, therefore, remains essential.

A number of well known QRA models exist in several countries, each designed to evaluate exposures of human receptors to organic contaminants via environmental pathways, including ingestion of soil and plant material. These models achieve comparable results for cases in which fate and transport of a specific substance is well understood, but use of 'generic' parameters (e.g. for ingestion of foodstuffs) can lead to large disagreements between model predictions. Use of similar or identical algorithms in different models, e.g. for assessing plant uptake of organic contaminants or human exposure to vapour phase chemicals, suggests that development of a core modelling approach should be possible if these algorithms can be successfully validated.

The current tiered approach to risk assessment should be reflected in the development of QRA models, with models of different levels of complexity appropriate for generic risk assessment at one of the spectrum, to detailed site- and case-specific QRA at the other end. Scientific underpinning should be provided by 'mechanistic' models, but these may be legitimately simplified to avoid over parameterisation and over confidence in the realism of QRA models. 'Creeping conservatism' (ie. the use of over-pessimistic parameters) is to be avoided.

With changing climate there might be increased pressure to use marginal land for food production, including contaminated sites. Changing climate and population migration will also lead to changes in consumption of specific foodstuffs, which might not be represented in existing QRA models.

Test sites and cases suitable for model validation are limited. More multi-site, urban validation studies are required to provide improved understanding of human exposures to organic contaminants in soils, and improved confidence in QRA models.

SHE-NET Workshop 3. University of Nottingham, 2-3 July 2008

Organising Committee

Shaw, George

University of Nottingham, UK

Duarte-Davidson, Raquel

Health Protection Agency, UK

Collins, Chris

University of Reading, UK

Bramwell, Lindsay

University of Newcastle, UK

Dowding, Alan

Food Standards Agency, UK

Martin, Ian

Environment Agency, UK


Al Mahaini, Talal

University of Nottingham, UK

Bell, David

University of Nottingham, UK

Cave, Mark

British Geological Survey, UK

Crout, Neil

University of Nottingham, UK

Deliman, Patrick

US Army Corps of Engineers, USA

Harrad, Stuart

University of Birmingham, UK

Houlden, Lawrence

Archon Environmental, UK

Judith Nathanail

Land Quality Management, UK

Kelly, Shona

University of Nottingham, UK

Mark Elert

Kemakta Konsult, Sweden

Paul Nathanail

Land Quality Management, UK

Piet Seuntjens

VITO, Belgium

Sakrabani, Ruben

Cranfield University, UK

Wragg, Joanna

British Geological Survey, UK

Page navigation


Search Form

A-Z lists