The SOLUTIONS project has developed and tested the scientific basis for these recommendations, and provides tools and services to utilize them [33].
Collation and curation of ecotoxicity data to apply CBM
The application of CBMs requires predicted or measured concentrations of chemicals and ecotoxicity data. Exposure data can be obtained from monitoring (e.g. according to WFD-prescribed approaches) or from modelling (e.g. [17]). We produced a curated set of ecotoxicity data (ecotoxicity test data and read-across data) to enable application of CBMs for a wide array of chemicals [15]. The database contains more than 250,000 raw data records—covering a suite of tested compounds and tested species—which can be used for the mixture assessment purposes described below. In daily practice, water quality assessors commonly use ‘digested’ data, derived from such raw data records. At present, it is not feasible to publish this database, due to the fact that it contains a subset of REACH study results that are in part proprietary (see https://iuclid6.echa.europa.eu/reach-study-results, accessed August 13, 2019). The combined data could, however, be used for research when, e.g. median effect data are used. Such uses are described below. Note that the European Chemicals Agency and data owners continue to improve accessibility of the REACH study results, which would change the availability of the raw data set.
Utilizing the data for mixture assessments
The curated data set [15] can be used to derive per-chemical risk quotients (RQ), and thereupon to derive indications regarding the WFD objective of protection against chemical pollution effects.Footnote 1 As discussed above, RQ results that are simply based on the ratio of the concentration and the EQS may have no meaningful ecological interpretation towards the type and magnitude of risk of the exposure if ΣRQ > 1. To address the complexities of interpreting RQ and ΣRQ to evaluate the WFD goals of protection and ecological impact magnitudes, we developed and applied innovative methods, by stepwise removal of causes of interpretation bias [34]. According to this tiered system, the assessor starts with available exposure and effect threshold data (either EQSs, or NORMAN-based PNECs), to evaluate whether ΣRQ < 1. If so, the assessment can stop, because the mixture risk for the measured compounds implies sufficient protection. If the lowest-tier results in ΣRQ > 1, the assessor obtains improved mixture risk information by (stepwise) removing unjustified assumptions. Details are explained in [34]. Applied to a series of sites, the approach allows for ranking the expected magnitude of impacts of the mixtures at the sites, so as to help prioritizing measures. Various case studies (see below) were executed with these improved CBM approaches. Note that a Europe-wide study on chemical pollution was made by Malaj et al. [4], whereby these authors derived the exposure-to-effect quotients for ambient concentrations in European waters to the effect end points of three selected species (LC50 or EC50s for an algal, an invertebrate and a fish species). The results of this assessment showed that ambient (measured) concentrations exceeded the impact end points of those species to different degrees. This provides evidence for the conclusion that organic chemicals likely affect those species if they would be exposed to those water bodies, for individual chemicals. In comparison to an EQS-based assessment in which the RQ is directly derived from the exposure/EQS ratio, this interpretation is straightforward, and not potentially biased by the interpretation problems of the EQS-based mixture assessment methods [34].
Assessment of toxic pressures of chemicals and their mixtures for species assemblages
To predict the fraction of species affected by mixtures, SOLUTIONS made expansions and improvements regarding the use of species sensitivity distributions (SSDs) in impact assessment, closely aligned with the WFD-Annex II obligation to assess “the likelihood of impacts”. The collated ecotoxicity database (see above) allowed for deriving SSDs for more than 12,000 compounds. The use of SSDs as the CBM method results in the derivation of toxic pressures (per chemical) or mixture toxic pressures (for mixtures), expressed as (multi-substance) potentially affected fraction of species [29]. The research team utilized an expert user modelling pipeline to apply the SSD-based CBM, as described in a project deliverable [35]. An associated (Dutch) project constructed a software tool for Dutch water boards (accessible via https://www.stowa.nl/publicaties/ecologische-sleutelfactor-toxiciteit-hoofdrapport-deelrapporten-en-rekentools, “Tool Chemiespoor”; in Dutch). This CBM approach was used in case studies, for example to derive insights into the spatial variation of the (multi-substance) potentially affected fraction of species (msPAF) resulting from modelled mixture exposure concentrations across European surface waters [15] and from measured concentration in Dutch surface waters [36]. In the European case study, the model was used to characterize whether mixture exposures are likely to cause insufficient protection, which is based on re-use of the so-called ‘95%-protection criterion’ (defined as PAF-NOEC < 0.05) for mixtures (as msPAF-NOEC < 0.05). The model was also used to provide a quantitative metric that is empirically associated with species loss (msPAF-EC50). The derivation of the toxic pressure of chemical pollution utilizes the model used for deriving EQSs in its inverse form [9, 11], implying conceptual consistency between deriving EQSs and toxic pressures. The mixture toxic pressure metric PAF-NOEC relates to the WFD environmental objective of protection, whilst the msPAF-EC50 metric empirically relates to impacts on the ecological status [37]. Mixtures matter for ecological status. According to these findings, assessors can use (measured or predicted) concentrations of chemicals in a mixture in combination with the pertinent SSDs and mixture models [15] to derive mixture toxic pressures. Applied to a series of sites allows for ranking the expected magnitude of impacts of the mixtures at the sites, so as to help prioritizing measures.
Case studies: prioritization of mixture-impacted sites and of chemicals in mixtures
The case study results provide evidence for the applicability of the improved CBMs and the utility of their outcomes for prevention, ranking of mixture impacts across sites and identification of drivers of mixture risks (including currently not considered chemicals) and management.
European and national scale
Applied to predicted environmental concentrations for more than 22,000 water bodies situated across Europe, these studies suggested that a large fraction of European surface waters are insufficiently protected against adverse effects of chemical emissions, and that the expected impact magnitude of contemporary pollution (expressed as msPAF-NOEC and msPAF-EC50) varies widely across water bodies [35, 38]. These across-site risk ranking results are in line with the aforementioned assessments of Malaj et al. [4] and results of Kortenkamp et al. [14]. These CBM-based results show that chemical pollution is a stress factor that threatens water quality across Europe, with different expected impact magnitudes across water bodies, and suggesting an important role of mixtures of components that are currently not considered. Moreover, the results presented not only a clear ranking of sites regarding mixture risks, but also the relative dominance of some chemicals in causing that (see also the subsequent example). The derivation of mixture toxic pressures (and the ranking of sites and compounds) is a straightforward assessment which is geared towards large-scale data analyses for water system level analyses. It has therefore not only been applied to predicted exposures, but also to (Dutch) national monitoring data. This yielded national water quality assessment outcomes for mixtures (site and compound ranking), despite differences in sets of monitored chemicals between different water boards [36].
Basin and water body scale
Various studies considered mixture risks for water bodies and basins based on measured concentrations. Munz et al. [39] identified CBM-based mixture toxicity differences between sites up- and downstream of wastewater treatment plants, and were able to identify drivers of mixture toxicity. Gustavsson et al. [40, 41] also showed a relative dominance, now for pesticides in Swedish streams and of monitored substances in coastal waters. These authors communicated those results via so-called ‘waterfall graphs’, to communicate that some chemical are ‘drivers of impacts’ (Fig. 1).
Massei et al. [42] identified mixture risks and drivers for mixtures of pesticides and biocides measured in surface waters of seven large European river mouths. Lindim et al. [43] studied pharmaceutical mixtures in Swedish freshwaters, and also identified key drivers of mixture toxicity. Finally, based on reviews of typically emitted compounds from different land uses, Posthuma et al. [44] simulated the mixture risks of those, providing evidence for different land uses being drivers of mixture ‘signatures’, again with some compounds dominating mixture risks. That is, different land uses cause vastly different packages of emitted chemicals, and vastly different temporal emission and exposure patterns.
Case study implications
All these case studies show that the systematic application of CBM approaches vastly improves the current practice of evaluating chemical pollution in the context of the WFD, in which a limited number of pre-defined priority compounds are assessed one by one. In fact, all the SOLUTIONS case studies flagged chemicals that are not on the WFD list of priority substances or on the corresponding lists of river basin-specific pollutants as mixture risk drivers in various European aquatic ecosystems. Extension of the consideration of a wider array of chemicals is warranted, as all chemicals may threaten the ecological status because all have the potential to cause that (given the observations collated in the ecotoxicity database).
It was further shown that mixture risks were often driven by only a few compounds, with the dominant compounds showing strong spatiotemporal variations. Although this, at first sight, could mean that water quality management could focus on a new fixed list of prioritized compounds—those identified as dominant via the CBM analyses—this is not the logical conclusion to be drawn. Every assessment scale (a defined area, with its emissions and hydrological characteristics) will result in its own rank order of sites and chemicals. We are already used to the fact that different scales result in different priority lists, when going from the European scale (the current 45 priority substances) to the river basin scale (currently approximately 300 river basin-specific pollutants, summed over the EU basins). A further step in downscaling would similarly result in different lists of dominant chemicals for different areas. This process can be followed down till the local water body scale. There only one specific chemical might dominate (e.g. one pesticide in a field ditch), whilst it may be far from dominant for the larger surrounding area (if the pesticide is not used there). Hence, there is always dominance of some chemicals in ambient mixtures, but the dominating chemicals vary among water bodies and over time. The latter follows from dominance changes due to, e.g. pesticide use. The WFD environmental goal of good ecological status may not be reached due to any chemical. Therefore, the WFD text defines pollution as the chemicals (no restriction) that pose a risk to maintaining or reaching the good status (Article 4, and the associated WFD-Common Implementation Strategy (CIS) Document #3, [45]). It appears that the consideration of potentially all chemicals has been lost in practice since the CIS document. Assessors should consider all chemicals and their mixtures, and can apply the improved CBMs to do so. Scale-dependent identification of dominant chemicals provides the chance to identify effective management steps per certain scale of activities.
Anticipating the effects for future emission scenarios and mitigation measures
CBMs can be used to explore foreseeable water quality changes based on future emission scenarios and to predict or retrospectively evaluate abatement success. The former was shown by Van Gils et al. [38]. Exploratory modelling of alternative chemical management scenarios showed a surprising effectivity of a focus on the most hazardous compounds, as identified in chemical safety assessment policies. The latter was also shown by Gustavsson et al. [40]. CBMs can be utilized, therefore, in the context of the solution-focused risk assessment paradigm, which asks for evaluating alternative management or chemical substitution scenarios. CBMs also fit well into the WFD assessment and management cycle [46], as temporal trends in pollution levels can be evaluated. The application of the approach also demonstrated that the risks and relative importance of various compound groups in relation to land use and waste water treatment plants varied [39]. Application to ‘think tank’ scenarios on future pollution, and evaluation of alternative abatement scenarios, was productive in that it showed which chemical groups and which focus in selecting abatement strategies would reduce predicted impact magnitudes most [30]. These examples also underline how monitoring data (WFD-Annex V) analysed with the CBMs can help to evaluate water quality status and trends. The solution-focused risk assessment approach implies that assessors explore the ‘solution space’ to define optional risk reduction scenarios [31]. Assessors depend on using the CBMs to evaluate mixture risks under the selected management options (as effect-based methods cannot be applied to expected concentrations), provided that there is a method to predict future concentrations. At present, such a method is available for the European scale [17], and work is in progress to develop a similar model for the Netherlands. For local cases, assessors may use available hydrological information to predict expected concentrations of alternative solution scenarios.
Summarizing and communicating results on complex mixtures
SOLUTIONS developed methods to summarize and communicate complex results. For sites, the relative importance of chemicals was suggested to be communicated as ‘waterfall graphs’, Fig. 1 [40, 41]. For the water system level analyses of chemical pollution, SOLUTIONS developed chemical footprints [16]. Aligned with the SOLUTIONS integrated Model Train, the footprinting allows summarizing local mixture toxic pressure, its origins (whether or not sources upstream contribute to local mixture stress) and its downstream impacts (evaluating effects elsewhere, caused by water flows). Regarding abatement, such summaries are key to assess whether abatement should focus on upstream sources of pollution, on local chemical emissions or on effects of downstream (sensitive) protection end points, or on combinations of these approaches. Currently available results have so far been used to illustrate how this approach operates and what type of results can be obtained [17]. The available EU-wide model can be used to derive these footprint results for selected areas and water bodies.
Lessons for improved chemical assessments
The use of CBMs in the case studies clearly emphasized the need for sufficiently sensitive chemical analytical procedures. Ideally, the level of quantification (LoQ) should be around 1/100th of the EQS, or, more realistically, the LoQ should at least approximate the single-substance EQS. SOLUTIONS developed and tested the Kaplan–Meier estimation method to handle compounds with insufficiently high LoQs [47]. Also, the expansion beyond the approximate 300 priority substances and river basin-specific pollutants requires additional hazard data. Repositories on hazard data (such as those of REACH, NORMAN, or the SOLUTIONS curated database of effect data) can be used as a source of such data for the CBM applications, provided that various key aspects are considered. Those are—at minimum—that ecotoxicity data used for a CBM could represent outdoor exposure conditions, and that data used have a transparent and reproducible origin [15, 47]. The consequences of neglecting proper management and choice of (eco)toxicity data are large, as presented in the report of Arle et al. [7]. These authors reported an array of EQS values for RBSP across European basins, whereby the minimum and maximum EQS values for one-third of the listed substances differed up to 10-fold from each other across countries, and more than half (53%) of all the substances differ by more than 10-fold and up to 105-fold from each other. This relates in part to the use of different assessment factors for deriving EQSs.
In general, the practical experiences from the case studies clearly emphasize that the ecotoxicity data repositories that form the basis for all CBM-based methods require substantial improvements in transparency, traceability, consistency and, last but not least, data quality.
The need for the use of improved CBMs
The current use of CBMs has two impacts on water quality assessment practices that negatively affect the likelihood of reaching the WFD environmental goals. This is caused by the fact that the indicator system sensitively reacts to extra chemicals becoming monitored and is at the same time highly insensitive to water quality improvements that occur upon abatement investments. These act as ‘hidden triggers’ that counteract reaching the WFD environmental objectives, as the first makes the assessor reluctant to add compounds to a monitoring plan and the second makes the assessor reluctant to invest in abatement as improvements remain hidden. The use of only two classes for chemicals (an exposure concentration is classified as either lower or higher than the EQS) is the root cause of this practical problem. The proposed improved CBM methods [14, 15, 34] provide refined insights into chemical pollution, required to inform managers on the needs to take protective or restorative management action. The quantitative insights provided by the improved CBMs deliver key insights for management prioritization and planning. The SOLUTIONS case studies showed this and how the use of the improved CBMs substantially—and the resulting ranking of mixture risks among sites and compounds—refines the information for water management prioritization and planning. Examples of the improved efficacy of refined CBM approaches outside SOLUTIONS have started with a landscape-level ‘one pesticide’ assessment for water bodies across the USA in 1996 [48]. Today, such assessments have expanded to mixtures and they are currently in the stage of gaining global appreciation (examples listed in [15]).