The following questions / concerns were discussed:
- Are we making ecologically relevant assessments? Are regulatory protection goals explicit and clear? Are they set in relation to environmental quality? How do prospective and retrospective approaches differ?
- Are all species of equal importance, or are there keystone species that are more important than others?
- Is a generic PNEC derived from an SSD overly simplistic in terms of ecological representativeness or should we develop representative assemblages/communities (archetypes) to represent different typologies? Should protection goals account for local community composition?
- How does aquatic community sensitivity vary with species composition?
- How can knowledge of chemical MoA help construct SSDs for HC5 estimation?
- What are the research needs?
Due to emerging views, the discussion followed the questions only loosely, instead taking a ‘high level’ view of the issues being raised.
A theme of the group’s discussions was that there was no such thing as ‘the’ ecosystem and consequently no single benchmark that would apply in all circumstances, in the sense of predicting accurately between ‘safe’ and ‘impacted’ conditions. It is reasonable to expect that assemblages that may be exposed to stressors like chemicals will vary in composition from place to place and over time. Examples of the more extreme ecosystem types might be communities associated with polar regions (characterised by low diversity but large numbers of organisms) or sub-tropical seas (where biodiversity is typically very high). It is also reasonable to suppose that these different ecosystem types will show a range of sensitivities to stressors, and while some may exhibit high sensitivity to a particular set of stressors (e.g. eutrophication), others may be more sensitive to the presence of other stressors (e.g. toxic chemicals).
Often, there is a requirement to set a generic criterion - one that is intended to apply across a large area (perhaps even a continent) - and is independent of environmental variables. Ideally, the most sensitive ecosystem is protected by the generic criterion. Logically, when this protects the most sensitive of a suite of systems (ecotypes) then other systems which are less sensitive would be protected too. That would imply a margin of safety when such a criterion is applied to less sensitive systems. In order to identify risk management measures, we need to acknowledge the variability between ecosystems, and try to understand the normal operating range of traits/species abundance in these different ecosystem types. This is very important to ensure that risk managers do not jump to the wrong conclusion and take action to address a less important pressure (false positive), or fail to fix a problem that really does need attention (false negative). For making more ecologically relevant assessments, we could envisage a distribution of ecosystem sensitivities. The group speculated whether it might be possible to place such ecosystem types on a scale of sensitivity to a particular stressor, and to extrapolate from ones where we have a good understanding of chemical sensitivity to ones where we do not.
Taking stock of current approaches, the group felt that most HC5s seem to be accurate (based on limited corroboration from field and mesocosm studies) in the sense that HC5 values often correspond to the absence of visible or measurable biodiversity changes in field or mesocosm studies. But is that merely a coincidence? We can see no underlying reason why ecosystems should ‘tolerate’ risks (NOEC-exceedance) to a maximum of 5% of species without measurable adverse changes to structure or function. As presented by Posthuma (section 3.1), there is no ecology (yet) in SSDs, nor in the definition of the ‘safe’ level. On the other hand, requirements of SSDs to capture a diverse spectrum taxonomic representation does suggest that practitioners broadly view ecological and taxonomic considerations as relevant and important.
The group also agreed that functional aspects of ecosystems are more resilient than structural aspects (e.g. primary production could go unaffected even if several species of algae were impacted by a stressor). This implies that a focus on protection against impacts on ecosystem structure would generally protect major ecosystem functional aspects too. However, it must be recognised that some species have specific value, because they provide important ecological services, they are charismatic, or are rare. However, we have little appreciation of a relationship between scarcity and sensitivity to stressors (are they rare because they are sensitive?). We seek to protect structure, and thereby function, in a generic (non-ecological) way and, based on the corroboration from field evidence, generally seem to succeed. However, we should be careful that the proportion of species at risk does not lead to functional ecosystem change e.g. if pollinators were in the affected fraction of species. Scenarios that look at specific receptors need be systematically considered but this is much more challenging when the aim is to develop a generic criterion, as opposed to a well-defined ecosystem or habitat .
In the current regulatory paradigm, the (largely unknown) variability between different ecotypes is dealt with by focussing on structural protection in our thresholds. We have been seeking to protect all but the most sensitive 5% of species by setting an HC5-NOEC based threshold criterion, based on clean water laboratory studies and most sensitive endpoints, in the expectation that such a threshold will provide adequate protection to a wide range of communities. With hindsight, where SSDs are used to estimate a threshold for a single stressor, the approach appears to be protective, but the approach may be too simplistic: we usually rely on information about the sensitivity of a sample of closely related individuals of a species to the stressor over a fixed period of exposure. We are not making effective use of the wider insights that are now available about (1) ecological, (2) chemical, (3) exposure, or (4) toxicological influences on risk. The following examples illustrate this over-simplicity.
- Ecological aspects: this includes an understanding of interdependencies between species through food webs (system level), or the relationship between sensitivity and traits such as reproductive strategy and feeding behaviour (species level). Instead of describing communities of organisms in terms of the species they contain, they could be described in terms of the traits they exhibit, or their dependencies on each other. Some aspects of food web architecture may be common to many ecosystem types, so this is where our effort should focus initially. Past studies for example have already looked into QSSRs (Quantitative Species Sensitivity Relationships), where not only chemical sensitivity but also relationships with body size and so on were studied. Trait-related studies are on-going in this respect.
It might be expected that the responses of organisms in extreme environments (e.g. highly saline) or where there is adaptation to stressors e.g. in metal mine tailings) are linked to the physiological adaptations needed for those environments.
- Chemical behaviour: There is now a much better understanding of the importance of water chemistry on bioavailability of metals and hence their toxicity to aquatic organisms. Bioavailability is now explicitly incorporated into SSDs for some metals. The influence of pH on weak acids could readily be accommodated in a similar way, giving rise to more environmentally relevant estimates of risk. The effect should be to lead to more accurate assessments of risk so that any remediation is directed to where it is really needed.
- Chemical exposure: The exposure profiles of many substances are known to affect the response of organisms, and these can be characterised. For example, many household chemicals typically give rise to low-level chronic exposure from point sources, whilst some insecticides applied to arable and tree crops are more likely to give rise to short episodes of exposure such as following accidental overspray or run-off shortly after application. Sessile organisms in estuaries are likely to experience diurnal variations in exposure to chemicals with tidal ebb and flow.
- Toxic mode of action: Information about a chemical’s mechanism of action is important in helping to identify taxa (or perhaps traits) that are likely to be particularly sensitive. Existing guidance in the EU Common Implementation Strategy ‘CIS’ Technical Guidance on EQS Derivation (EC, 2011) acknowledges this and suggests that information about mode of action may be used to adjust assessment factors. Alternatively, an understanding of mode of action can help focus attention on critical data gaps that may be filled by testing particular species, or some of the non-testing alternatives suggested below.
How might such higher level thinking be incorporated into our hazard and risk assessments? One approach may be to use existing and emerging tools for generating information about sensitivity of different species to chemicals (e.g. QSARs, ‘read across’ tools, Web-ICE, dynamic energy budget tools [DEBTOX], to simulate different ecosystem types, and the consequences for thresholds if we were to simulate exposure of a community dominated by certain taxa or trophic level e.g. primary producers. That is, there is a need to distinguish between the protective success of our conventional generic methods (deriving one criterion to cover all eventualities), and the need to tailor the generic approach to specific circumstances when this is required (e.g. site-specific thresholds).
In a tiered system, the generic criterion serves as a starting point, which protects all systems. This is intended to be protective, but provides different margins of safety to different ecosystems. At a higher tier of assessment, SSDs for specific scenario (a specific area, species composition, or community exposed to with other stresses, etc.) may utilise information drawn from the 4 fields mentioned above. The aim would be to refine the SSD output to the system of interest. As a strategy, the group felt that there is ample opportunity to create ‘what if’ scenarios with SSDs, so the assessor can decide whether the criterion is, or is not, sensitive to adding scientific insights and data. By incorporating some of the principles mentioned in (1) – (4), it may be possible to simulate the effects of, for example, changing the set of tested species data in the SSDs to mirror the ecosystem under study. Low impacts of such simulations would suggest that the criterion is robust, whilst high impacts suggest specific attention for the factors causing the change, which may feed into risk management activities, or help guide the generation of new data.
The more intensive approach to hazard and risk assessment we are advocating could add a lot more work. How do we know when enough is enough? How can we decide when a next tier is necessary, and when we should stop? The group felt that the alternative scenario outputs, and the confidence estimates around an HCx provide a useful prompt for more (or less) scrutiny. A large uncertainty should be a major trigger for the sorts of systems thinking advocated above. Field data may also have a stronger role to play as a line of evidence in defining and using thresholds. The EU Technical Guidance on EQS derivation (EC, 2011) already refers to the use of field data in informing the size of assessment factors to be used.
In summary, the group felt the time was right to move away from using SSDs as a purely statistical construct applied to poorly understood species sensitivity data to one in which SSDs provide the framework for a more process-based approach. We envisage a statistics-related approach will remain at the core of our assessments but it can be enriched with fundamental insights of the kinds stated in (1) – (4). This would be greatly assisted by some sort of over-arching guidance (not a ‘rule book’) that would prompt the assessor to think about some of the factors that might be important (1-4 above) and the options for pursuing them further. The guidance might usefully adopt a hierarchical structure (as suggested above) that takes the assessor through a series of ‘things to think about’ in some logical sequence. The aim should be to stimulate a broader approach to the assessment and not to ‘fossilise’ the science by being over-prescriptive or wedded to particular tools. Any over-arching guidance should, however, be clear about what tools are available, along with their strengths and limitations.