The idea behind expressing ecological protection goals in terms of ecosystem services is to make a connection between ecosystems and what people get out of them in terms of marketed goods and nonmarketed welfare. The initial motivation was to draw attention to the importance of ecosystems in supporting the economy and hence the importance of what we get out of them. This has involved approaches drawing attention to the total value of ecosystem services on the planet, in particular nations and mapping across landscapes (Costanza et al. 1997; Balmford et al. 2002; Naldoo et al. 2008). From a public relations point of view these might be useful exercises (Gomez-Baggethun et al. 2010). However, it ought not to be surprising that global services add up to big numbers, because without them we would not survive. More importantly, though, it is often unclear in these kinds of studies whose values are being represented, and yet values are dependent on socioeconomic circumstances (Pearce 1998).
Framing environmental protection goals in terms of ecosystem services is gaining momentum (Fisher et al. 2009). Here our focus will be on how the ecosystem services framework is and can be applied in the ecological risk assessment (ERA) of chemicals. We provide 2 contrasting examples of how the ecosystem services framework is currently being applied in regulatory risk assessment, and we discuss the challenges and knowledge gaps that need to be addressed if such a framework is to substantially improve ERAs and their ability to inform management decisions. We make the point that formulating ecological protection goals in terms of ecosystem services only makes sense if they can be used in managing environmental impacts, and if they are useful in informing the risk assessments behind these. The presumption is that what is being managed matters to the extent that it is valued by those affected and that services express that value. The ecosystem services approach therefore has a part to play in ensuring that risk assessments are value relevant.
Several consequences arise from using an ecosystem services framework to its fullest potential in a risk assessment context. The first is that it makes explicit that there are tradeoffs as well as positive couplings among services and hence a need to value services appropriately. The second is that we need robust models that can link impacts on service-providing units with the delivery of services over appropriate spatial and temporal scales. The third is that we need predictive systems models that can confidently link the kinds of effects that are measured in risk assessments (that are typically on individual organisms) with impacts on service-providing units (that are often populations or groups of populations).
There are substantial challenges—and not just for the science. For example, an interesting caveat that arises from the ecosystem services philosophy is that if the services are connected directly to markets then they ought to be managed through market forces of supply and demand. As fish and lumber get less plentiful, the price should rise and the demand should fall. However, there are market failures. For example, subsidies can interfere, but also, and most importantly, there are services external to the market associated with such things as recreation and just knowing that nature is there. Cost-benefit analysis following the valuation of these services is used to internalize externalities, and when the benefits of interventions outweigh the costs, intervention is justified from a welfare economic point of view (Hanley and Barbier 2009). The benefit depends on the likelihood of the impact being avoided and the value that those affected put on it. To manage this kind of situation, we need the likelihood of impact from the risk assessment and the value of the service. This emphasizes that the impact should be expressed in service- and hence value-relevant terms. The ecosystem services philosophy focuses on making a connection between the structure and function of ecological systems and delivery of the valued services—so-called ecological production functions (Daily and Matson 2008).
CONCEPTUAL FRAMEWORK AND IDEALIZED IMPLEMENTATION OF THE ECOSYSTEM SERVICES APPROACH IN ERA
For assessing ecological risks of chemicals, there is a wide range of legislation across jurisdictions that broadly defines protection goals. These might be based on general protection goals explicit or implicit to the legislation (e.g., in pesticide and industrial chemicals legislation in the European Union [EU]) (Regulation (EC) Number 1107/2009; Regulation (EC) Number 1907/2006) and/or to particular situations where sites are likely to be exposed (e.g., Superfund sites in the United States) (Munns et al. 2009). Articulating environmental protection goals in terms of ecosystem services can be helpful for several reasons. First, the general protection goals formulated in most legislation are vaguely defined, from a scientific perspective, and hence not easily measurable. For example, the new Pesticide Regulation in the EU (EC 1107/2009) states that plant protection products, “shall have no unacceptable effects on the environment.” Formulating potential impacts on “the environment” in terms of the delivery of a range of ecosystem services assists in making the impacts measurable, allows for a systematic and transparent analysis of the types of impacts that need to be considered, and makes obvious the fact that delivery of all services cannot be maximized at the same place and time.
Another advantage of the ecosystem services framework is that the services can be subject to valuation using techniques developed by socioeconomists, and ideally these should be the values of those affected by the impact and any intervention designed to ameliorate the impact (Hanley and Barbier 2009). Defining who is adversely affected by the impact on ecosystem services and affected by the cost of the intervention will always be most difficult for broad-scale impacts; for example where there is regional or even global exposure, and here there may be differences between those who are affected by the exposure and those affected by the intervention. In all situations, whether considering global or local situations, it will always be easier to use existing values than to define the ones most appropriate to the circumstances because valuation is time consuming and expensive. However, such extrapolation, called value transfer, needs carrying out with care because values are very sensitive to socioeconomic circumstances and might therefore vary through both space and time (Pearce 1998). Finally, issues of distribution of impacts and the costs of intervention are matters that can ultimately only be resolved by governments.
An ideal implementation of the ecosystem services framework in ERA would start by asking the question, “What is it we want to protect and over what spatial and temporal scales?” The next step would be to systematically consider the ecosystem services that could be potentially at risk in the given situation. Then it would be necessary to determine which ecological entities are providing the services (i.e., service-providing units [SPUs]) and to derive quantitative relationships between the SPUs and service delivery (i.e., ecological production functions). Ideally the ecological production functions could be used to determine the type of data that would be needed to estimate risk to service delivery. This conceptual approach is illustrated in Figure 1.
Unfortunately, the idealized implementation of the ecosystem services framework is somewhat at odds with the actual practice of ERA. On the one hand, existing chemicals legislation requires protection of the environment in general terms. However, for historical reasons, the kind of data that are required to assess chemical risks are so far removed from the protection goals that much of the field of ecotoxicology has been preoccupied with trying to relate what is measured to what we want to protect (Forbes et al. 2008). Whereas an ecosystem services framework can be helpful in this regard, its implementation will be different than in the idealized approach indicated in Figure 1. In practice, we usually only have information on the effects of chemicals on individual organisms across a few species—and these are more often than not from acute exposures to unrealistically high concentrations with mortality as the test endpoint. Considering that generally SPUs do not consist of individual organisms but of populations or groups of populations, not to mention that sublethal responses to chronic exposures are not accurately or consistently predicted from acute mortality (Forbes and Calow 1999; Hanson and Stark 2012), there is a need to translate concentration-response relationships for individual organisms produced in risk assessment to concentration-response relationships for ecologically relevant responses occurring at higher levels of biological organization. This can be done through the application of predictive systems models (PSMs) (Forbes and Calow 2012). Outputs from the PSMs then need to be related mechanistically to service delivery through the derivation of ecological production functions. Changes in service delivery through impacts on the SPUs can then be valued using appropriate socioeconomic models. This is illustrated in Figure 2.
Naturally there are challenges here as well. One of the most important involves interspecies extrapolation because the species for which standard toxicity test data are produced as part of the risk assessment are not necessarily good representatives of the relevant SPUs. Thus decisions have to be made as to which species or groups of species to model as representative SPUs, whether they should be real species or generic species, and how the toxicity data generated for standard test species can best be extrapolated to the modeled species. Decisions on these important issues require further consideration and will likely involve extensive stakeholder dialogue.
In the next sections, we provide 2 contrasting examples to show how the ecosystem services concept is currently being applied in ERA and risk management of chemicals. We then go on to consider what challenges need to be met before the ecosystem services framework could be used to its full potential in support of ERAs.
APPLICATION OF ECOSYSTEM SERVICES BY THE EUROPEAN FOOD SAFETY AUTHORITY
The European Food Safety Authority (EFSA) is an independent agency that is responsible for risk assessment of all aspects of food and feed safety. The European Food Safety Authority has been charged with updating the Technical Guidance Documents (TGDs) for Pesticide Risk Assessment in light of the new European Regulation (EC 1107/2009). Ideally, the TGDs should be designed to help regulators decide what to protect, where to protect it, and over what temporal and spatial scale to protect it.
With the assistance of a working group of experts, the process that EFSA used to inform revision of the TGDs for Aquatic Ecotoxicology and Terrestrial Ecotoxicology was to start with the comprehensive list of ecosystem services outlined in the MEA (2005), and to consider which of these could be potentially impacted by pesticides. The next step was to identify key drivers (representative taxa or functional groups) involved in delivery of each service—in other words, SPUs. For each SPU, the level of biological organization needed to protect the service was identified (i.e., individual, [meta]population, functional group, ecosystem), as well as the relevant attribute or attributes of the SPU (e.g., biomass, process rate, etc.) needing protection. In addition, a transparent scheme was provided for risk managers to determine maximum tolerable impacts on SPU attributes in terms of magnitude of effect, temporal and spatial scales over which effects could be considered acceptable, and the degree of certainty that these maximum tolerable effects would not be exceeded (EFSA 2010; Nienstedt et al. 2012). It is important to note here that this exercise was intended as a one-time analysis designed to inform revision of the TGDs and will not be conducted every time a risk assessment is performed. However, important decisions still need to be made regarding the selection of vulnerable representatives of each SPU or key driver as well as how to extrapolate toxicity data from the standard test species to the representative SPUs. It is likely that PSMs will have an important role to play in the future for extrapolating standard toxicity test data to ecologically relevant responses in SPUs. However, the development of quantitative ecological production functions, that link impacts on SPU attributes to ES delivery, as well as valuation of the various ecosystem services potentially impacted by agriculture are beyond the scope of the EFSA exercise.
In many respects EFSA's application of the ecosystem services framework in ERA is an important step forward. Framing protection goals in terms of ecosystem services has proven to be a valuable communication tool; the revised guidance has the potential to improve the extrapolation from standard test endpoints to ecologically relevant responses (e.g., by allowing the use of appropriate mechanistic effect models); and the EFSA scheme for defining maximum tolerable impacts has the potential to increase the consistency and transparency of risk management decisions. However, it is not clear at this time what impact this exercise will have on the risk assessment of pesticides in Europe. For example, it is not likely to result in major differences in which species are tested. Because there has been no mention of including ecological production functions for making quantitative links between SPUs and ecosystem service delivery, it is unclear how impacts on SPUs will be evaluated. Finally, there are no indications that risk managers will use this opportunity to put values on ecosystem services or make tradeoffs among them explicit, and hence make their risk management decisions more transparent.
APPLICATION OF THE ECOSYSTEM SERVICES CONCEPT BY THE US ENVIRONMENTAL PROTECTION AGENCY
An important aspect of US Environmental Protection Agency (USEPA) interest in the ecosystem services framework has focused on the site-specific risk assessments associated with the Comprehensive Environmental Response, Compensation and Liability Act (CERCLA, more commonly known as Superfund). Here there is a pressing need to make more explicit links between ERA of hazardous substances at the contaminated sites and natural resource damage assessments (NRDAs). The ERAs have tended to emphasize ecological impacts, and the NRDAs require that impacts be valued so that appropriate action and compensation can be effected. Clearly, an ERA that is framed in terms of impacts on ecosystem services would facilitate the valuation process (Munns et al. 2009). The steps in this process would ideally be those defined in Figure 1, but even when this is focused on particular sites with particular ecological systems and hence services, there are acknowledged difficulties in carrying out such detailed analyses routinely. Therefore, there are calls for the translation between ERA and NRDA to be based on generic ecological services assessments that might be applicable in a variety of environmental management contexts (Munns et al. 2009). This would be based on the generic ecological assessment endpoints already defined by the USEPA (2003). However, these focus on structural and process characteristics such as the survival, reproduction, growth, and morphology of individuals, the abundance and production of populations, the biodiversity and processes of communities, and the extent and quality of habitats, and would have to be translated into valued services in much the same way as has been suggested for ecological benefits assessment by USEPA (2006). There will be at least 2 challenges. First, there will be challenges in making general connections between structure, process, and services, i.e., in appropriately extrapolating from one ecological situation to another. Second, care will be needed in extrapolating values from one place to another as described above.
WHAT NEEDS TO BE DONE AND CHALLENGES FOR GETTING THERE
We deal with these issues in turn for the ecological models, ecological production functions, and socioeconomic approaches.
Mechanistic effect models to link test endpoints to SPUs
Although implementation of an ecosystem services framework into ERA would ideally follow the logic outlined in Figure 1, in which the services to be protected defined the type and amount of data to be collected, the practical reality is that it is the type and amount of data that are routinely collected that, in the absence of a major paradigm shift, will continue to drive ERA. However, there are opportunities for greatly improving the extrapolation from what is typically measured (e.g., acute mortality of individual organisms) to value-relevant endpoints (e.g., the long-term persistence of populations in space and time). A range of mechanistic effect models exists to connect standard test endpoints to ecologically relevant effects on service-providing units. These models vary from very simple models that integrate chemical impacts on different life-history traits simultaneously (Forbes and Calow 1999; Hanson and Stark 2012), to models that incorporate density dependence (Palmqvist and Forbes 2008; Hayashi et al. 2009) and/or environmental stochasticity (Haridas et al. 2012), to those that can incorporate spatial and/or temporal variability in exposure (Wang and Grimm 2010), to extremely complex and realistic simulations of individual organisms interacting in virtual ecosystems (Topping et al. 2003).
Challenges for the ecological modeling include being able to deal with both too much and too little information. For most species there is a lack of basic ecological data (e.g., on life-history traits under different environmental conditions, behavior, mechanisms of density dependence, etc.) necessary to parameterize all but the simplest population models. Because it would be impractical to collect such detailed information for all species, one approach could be to identify a small subset of representative or focal species that can be studied in detail and modeled accurately. An alternative approach could be to create generic or virtual species that contain those characteristics thought to make them potentially most vulnerable to chemical impacts. Each approach has advantages and disadvantages, and decisions about which to use can probably best be made through extensive stakeholder dialogue and consensus.
In addition to the problem of having too few data, we are increasingly faced with the problem of having too many. Through advances in sensor and monitoring techniques, it is possible to measure environmental variables at a level of temporal detail and over a range of spatial scales that has not been possible before. Likewise, advances in various high-throughput techniques are facilitating the generation of huge amounts of information at molecular and metabolic levels, and there is the expectation that this information should be relevant for ERA (Van Straalen and Feder 2012). Although the identification of correlations between molecular responses and exposure to toxic chemicals may provide insights into which metabolic pathways are involved in the responses of organisms to chemicals, using such data to make predictions about ecologically relevant responses, so as to assess risks, will require the development of better computational approaches that are nonlinear, multidimensional, mechanistic, and quantitative (Forbes and Calow 2012).
An additional challenge for linking standard ERA test endpoints to impacts on SPUs is that there is currently a lack of widely accepted, user-friendly ecological models that have been sufficiently tested for this purpose. Regulators are understandably reluctant to use models that they do not trust, and given that most are not modelers themselves, they need to be provided with guidance on choosing appropriate models, evaluating their quality, and interpreting their outputs. In addition, there are concerns that decisions could become more complex and/or time consuming particularly if the outputs of ecologically complex models need to be evaluated. There have been a number of initiatives in recent years (USEPA 2009; Forbes et al. 2011; Thorbek et al. 2010) intended to bring together stakeholders, from academia, industry, and regulatory authorities to discuss the barriers that prevent mechanistic effect models from being used in ERA and to develop suggestions for overcoming such barriers. These efforts have resulted in an initiative to develop Good Modeling Practice for ERA (Grimm et al. 2009) as well as in various initiatives to train model users and to develop case studies.
Ecological production functions to link attributes of SPUs to ecosystem service delivery
Despite substantial progress in the development and application of mechanistic effect models that can link the kinds of data produced in standard ERAs to impacts on populations under ecologically realistic conditions, less attention has focused on developing models that can make quantitative and predictive links between populations or groups of populations and the delivery of ecosystem services. Development of such ecological production functions is essential for determining the magnitude (i.e., severity, spatial, and temporal scale) over which impacts can be considered acceptable and for identifying which attributes of the SPUs (e.g., growth rate, biomass, age/size distribution, etc.) are the most robust predictors of ecosystem service delivery. For most services, we do not know how ecosystem service delivery relates to the outputs of ecological models that are often framed in terms of population growth rates, population size, etc. There is a pressing need for research that tightly and robustly links easily measurable properties of populations or groups of populations to service delivery. For those services that traditionally have been traded in markets, such as production of food and fuel, it should be fairly straightforward to link service delivery to population attributes (e.g., fisheries production relies on population density and age/size distribution of fish). However, for other services, such as cultural, spiritual, and educational services, it will be more challenging to identify the most appropriate attributes of the SPUs that predict service delivery.
Socioeconomic approaches to put values on impacts to ecosystem services
Clearly, the degree to which ecosystem services thinking can improve ERA will rely on the use of appropriate economic valuation models that can put values on ecosystem services and quantify tradeoffs among them. Certainly there are challenges in putting values on ecosystem services especially for nonmarketed goods. More fundamentally, there are recurrent arguments about the extent to which nature can be valued. However, to the extent that we need to make decisions about providing more or less protection to ecosystems as compared with other aspects of what people need, then preferences are invariably expressed about more or less ecosystem services. Welfare economics makes the presumption that the best way to make decisions is by assessing the preferences of those affected in quantitative (monetary) terms and seeking to find policies that maximize the welfare (benefits over costs) of this group. One counter view is that some services have intrinsic value that is missed by the economic analysis. However, this often means that the things with intrinsic value should be given the value ascribed to them by those making this claim and/or that they be given very high (infinite) value (McCauley, 2006). In other words, this does not avoid making decisions on the basis of preferences—but the preferences might not reflect those of the public affected. As far as putting value on nonmarketed ecosystem services, welfare economists have made great strides in developing an array of sophisticated techniques that either reveal values or derive them from stated preferences to the extent that this is unlikely to hold ecosystem services analyses back (Hanley and Barbier 2009). However, these techniques are often very resource intensive and so value transfer is common, and here care needs to be exercised in ensuring that appropriate allowance is made for the variability in value that goes with socioeconomic circumstances (Pearce 1998).One important area of contention, though, is when there are different values across the group affected. The traditional way of dealing with this is to suppress the variability by using representative consumers and producers. More sophisticated models are now being developed that reflect the heterogeneity, and these will undoubtedly come to play a more important part in ecosystem services analyses (Beinhocker 2006). Finally, Norgaard (2010) has pointed out that the ecosystem services approach often takes a narrow view of the economy in partial equilibrium analyses—yet there can be broader knock-on effects into the general economy from both the impacts and interventions, and these need to be considered in more general equilibrium models. This is an inherent problem with simplistic cost–benefit analysis. Our view is still that risk assessments that are informed by the ecosystem services approach (i.e., that are value relevant) are likely to be more helpful in taking a view about the effects of likely impacts, and the interventions that are designed to address them, on the general economy than those based on the more ad hoc current approaches to risk assessment and management.
Except from a public relations point of view, mapping ecosystem services for the sake of it does not add much to environmental management and may sometimes mislead in terms of using inappropriate values. Ecosystem services can make a contribution to management by connecting ecosystem structure and process to what is valued, and analyzing risk in this context is a way of making risk assessment more policy- and value-relevant.
The EFSA approach, by using the ecosystem services framework to evaluate the general relevance of the tests that are carried out and the way they are used in risk assessment, is relatively straightforward. Given the difficulties in identifying the public affected, and hence which values to use, this approach may be the only one possible in the broad-scale risk assessments used for example for pesticides and industrial chemicals (e.g., as in REACH).
The USEPA approach of using the ecosystem services framework to ensure that impacts are appropriately assessed and valued at particular sites has potential but still requires considerable effort. The intent to shortcut by using generic ecological services assessments is understandable, but care will be needed in assigning appropriate values.
An important advantage of the ecosystem services approach is that once SPUs are expressed in the same (e.g., monetary) units it becomes much more straightforward to aggregate impacts across services and/or to consider possible tradeoffs between the services themselves in implementing different kinds of interventions. If SPUs are not measured in the same units, judgments have to be made about priorities, and hence preferences are expressed (often by the regulators) (Munns et al. 2009) to an extent that is not transparent and may not reflect those of the public affected. This is an argument for taking the analyses as far as possible toward valuation.
The biggest challenges in using an ecosystem services approach in ERA will be in making the ecological connections, and we see modeling as facilitating this. We are still a long way from the kind of ecological understanding that would facilitate routine use of ecosystem services in decision making. Particular challenges include the need to create a new paradigm for linking responses of biological systems at different levels of organization that is mechanistic, quantitative and predictive. It will be important to develop better understanding of how much (and which) complexity needs to be included in models in order to make robust management decisions. Finally there is a need for appropriate guidance on model development, documentation, and interpretation, as well as training for potential model users.
Aside from the challenges for ecology, the challenges in applying the economics are not unimportant. From an ecosystem services perspective, probably the most important of these is that though the analyses encourage a focused approach there will always be important connections between the costs and benefits associated with particular interventions and the economy at large. The importance of this may well have to be considered on a case by case basis.
In summary, using an ecosystem services framework to its fullest potential to support ERA will require the successful development of a suite of coupled Valuation Methods, Ecological Production Functions, and Mechanistic Effect Models as described above. This will require the establishment of strong multidisciplinary collaborations among ecologists, computer scientists, social scientists, and possibly others. In addition, buy-in from environmental decision makers and other stakeholders will be crucial. Some progress is being made in that EFSA, USEPA, and other agencies appear to be supportive of using ecosystem services to inform ERA, and that should facilitate more and better science being carried out in this area.
Finally, new legislation (e.g., REACH) is creating a need for value-relevant metrics that can be used in socioeconomic assessments. This should provide an incentive to develop risk assessment endpoints that are much more directly related to environmental protection goals than are the usual hazard/risk quotients.
This paper is one of 8 articles generated from the SETAC Special Symposium: Ecosystem Services, from Policy to Practice (15–16 February 2012, Brussels, Belgium). The symposium aimed to give a broad overview of the application of the ecosystem services concept in environmental assessment and management, against the background of the implementation of the European environmental policies such as the biodiversity agenda, agricultural policy, and the water framework directive.
This article is based on an invited talk presented at the 5th SETAC Europe Special Science Symposium on Ecosystem Services—From Policy to Practice held in Brussels, Belgium, 15–16 February, 2012. The ideas presented here have benefitted from interactions with many collaborators and students, most notably those involved in the EFSA Working Group on Ecotoxicological Effects and the European Union 7th Framework Programme Project CREAM (PITN-GA-2009-238148).