Additional supporting data are found in an appendix available in the online edition of IEAM Volume (2), Number (4). DOI: 10.1897/2005–024.1.
The development of relevant frameworks for assessing ecological risks posed by dredged material management does not only involve an appropriate selection of assessment and measurement endpoints but also requires a sound approach to both risk characterization and the associated uncertainty. A formal methodology addressing both aspects has been developed in France for freshwater sediment deposits in water. Both exposure and effects measurements are 1 st transformed into scores or classes. As far as possible, class boundaries are based on existing knowledge or expertise. Benthic organism exposure is based on a ratio of the deposit area to the burrow pit area, whereas pelagic species exposure is based on the ratio of porewater volume to water column volume. The combination of exposure and effect scores yields risk scores, or classes, which are linked to management decisions. Uncertainty is assessed with respect to a set of 4 predetermined criteria for exposure (the strength of association with the assessment endpoint, spatial and temporal representativeness, and the use of standard methods) and 4 criteria for effects (strength of association, the distinction between effect and no effect, sensitivity, and the use of standard methods). This approach was applied to 8 sediments from French canals contaminated to varying degrees.
Additional supporting data are found in an appendix available in the online edition of IEAM Volume (2), Number (4). DOI: 10.1897/2005–024.1.
The French Ministry of Transport and the French Ministry of Ecology currently support the development of general and specific guidance measures for managing the risks posed by dredging operations in continental waterways. The development of relevant frameworks for assessing the ecological risk of contaminated sediments or dredged materials not only requires an appropriate selection of assessment and measurement endpoints, but also a sound approach to risk characterization, highly desirable because it will both promote clear-cut decisions and facilitate communication to the public. Although an array of methods are available for risk characterization in aquatic media (USEPA 1998; Aldenberg and Jaworska 2000; ECB 2001), few are directly applicable for bed sediments or dredged materials. Consequently, assessors often refer to weight-of-evidence (WOE) evaluation procedures, which combine the results of multiple measurements or lines of evidence and make it possible to consider their strengths and weaknesses (Menzie et al. 1996; Burton et al. 2002).
The TRIAD approaches, as set out for instance in the dredged material assessment procedure in The Netherlands (Den Besten et al. 2003), could be viewed as a specific case of WOE assessment, combining 3 or 4 different lines of evidence: Chemistry, bioassays, field surveys, and sometimes bioaccumulation tests (Burton et al. 2002). A similar approach combining sediment chemistry, bioassays, and bioaccumulation tests was applied in Spain for coastal sediments (Casado-Martinez et al. 2005). A common feature of these frameworks and others is that they are tiered (DelValls et al. 2004). However, in most cases, there is no formal evaluation of uncertainty, as should be the case in any ecological risk assessment (USEPA 1998).
The Massachusetts Weight-of-Evidence Workgroup (Menzie et al. 1996) provided a framework for integrating the results of several measurements related to a single assessment endpoint. This procedure allows a rigorous consideration of the strengths and weaknesses and of the nature of uncertainty associated with each measurement. Although their conceptual framework was not developed specifically for sediments or dredged materials, it can be applied to this type of material, as Johnston et al. (2002) have done for an estuarine site downstream of a former shipyard. This site-specific study also provided a number of conceptual developments to the WOE approach initially proposed by Menzie et al. (1996) and established a basis for designing the risk characterization step of ecological risk assessment frameworks of bed sediments or dredged materials.
In application of the general assessment and management scheme recommended for freshwater sediments (Imbert et al. 1998), a multitiered ecological risk assessment framework was developed for dredged materials in freshwater systems (Babut and Perrodin 2001; Babut et al. 2004). This frame-work for dredged materials is based on chemical analyses at the 1st tier and bioassays at the 2nd. For the 3rd tier, 2 management options are currently being considered, deposits in water and deposits on soil. At this stage, the procedure is still standardized: Standard biotests are used as well as leaching assays for which fixed protocols are available. The site characteristics are used for scaling exposure. A 4th tier of sophisticated tests or site-specific approaches can be launched if deemed necessary. Taking 1 scenario as an example, the deposition of dredged material underwater in an open gravel quarry, Figure 1 shows the whole logical process for this management option.
The framework itself is presented in detail elsewhere (Babut et al. 2002; Den Besten et al. 2003). Briefly, at the 1st tier, metals and PAHs at least are analyzed. A mean quotient is calculated on the basis of probable effect concentrations (MacDonald et al. 2000). For quotient values lower than 0.1, the dredged materials present are considered a negligible risk, so the assessment process may end there. Conversely, for quotient values above 0.5, tier 3 would start directly. For quotient values between 0.1 and 0.5, sediment toxicity biotests, namely Chironomus riparius and Hyalella azteca should be realized. Depending on the results, the assessment process could either stop or be pursued in tier 3. This latter tier combines toxicity biotests with a leaching test allowing contaminant transfer to surrounding waters to be assessed; exposure is related to the deposit site dimensions.
This paper focuses on the risk characterization step at the 3rd tier, which is the process integrating the different lines of evidence into a single and intelligible figure. Because the available measurements are expressed in various units, and sediment tests do not rely on a dose-response relationship, a common and logical scale must be defined. The final aim is to provide appropriate guidance to local environmental managers on how to assess the risks involved with dredged materials, including an understandable, standard, and applicable approach to risk characterization. Moreover, because addressing uncertainty is an intrinsic part of the decision process (USEPA 1998), the risk characterization procedure and uncertainty assessment have to be addressed jointly to determine the level of confidence associated with the risk estimate.
This risk characterization approach was applied to 2 sets of data: A study of 3 sites in a French canal in the northeastern region, referred to as the CEBS study, and 5 sites from canals in the northern region, referred to as the NPDC study. Because no actual dredging operation was planned for these sites at the time of study, a virtual quarry was used for exposure calculations.
Although important in practice, the risk to groundwater quality will not be addressed in this paper, which is focusing on risk characterization for ecological targets.
The quarry is likened to a cross section of the alluvial groundwater. Therefore, the water will flow through the dredged material deposit, and contaminants in the deposit, if any, could be leached out over time. Aquatic organisms might be exposed to contaminants after release of the porewater into the water column. Sediment species might be affected in various ways, in particular when they recolonize the deposit. The following assessment endpoints have thus been proposed: 1) The deposit should not disturb the structure and abundance of benthic invertebrate populations, 2) the releases from the deposit should not have chronic effects on aquatic organisms, and 3) the releases should not degrade groundwater quality (i.e., that used for drinking water).
A summary of the conceptual model is given Figure 2. The 1st assessment endpoint involves biotests applied to bulk sediment with an insect larva (C. riparius) and an epibenthic crustacean (H. aztecd). For the 2nd assessment endpoint related to the water column, measurements include an alga (Pseudokirchneriella subcapitata), a pelagic crustacean (Ceriodaphnia dubia), or a rotifer (Brachionus calyciflorus); these tests are done with porewater. A leaching test examines the assessment endpoint concerning groundwater quality. The measurements related to the groundwater quality assessment endpoint and the subsequent risk characterization are not considered in this paper.
In technical terms, risk results from a combination of hazard—the intrinsic effects of a stressor—and exposure (USEPA 1998). However, this step of the overall risk assessment process is complicated in most cases by the cooccurrence of several measurement endpoints (e.g., toxicity tests) for a given assessment endpoint (e.g., benthic population maintenance) and even by the co-occurrence of several assessment endpoints.
The risk characterization approach proposed to the French authorities stems from the above-mentioned papers by the Massachusetts WOE Workgroup (MWWG; Menzie, 1996) and Johnston et al. (2002). Both effects and exposure data are converted and combined into classes of ecological risk.
The uncertainty analysis describes the degree of confidence in the assessment and can help the risk manager focus research on those areas that will lead to the greatest reductions in uncertainty (USEPA 1998). Because the qualitative aspects of this uncertainty are as valuable to risk managers as statistical measures of confidence (NRC 1983) cited by Goldstein (2003), the selected approach evaluates uncertainty against a set of predetermined attributes and provides an overall score combining individual scores for these attributes.
Effects assessment—As in a shipyard case study (Johnston et al. 2002), 3 classes of effects were determined; that is, 2 thresholds for effects are considered, whose definitions differ slightly from those used in the shipyard case study. In class 0, the effects are said to be similar to reference or control conditions, whereas in class 2 they should be substantially different. For sediment toxicity biotests, the thresholds were based on former studies in The Netherlands (Den Besten et al. 1995), whereas for aquatic toxicity biotests, they were taken from the existing regulations on hazardous wastes (MATE 1997, 2002). This latter option was selected because in Europe, dredged materials are considered waste when extracted from the riverbed. All thresholds for both assessment endpoints are given in Table 1.
Exposure assessment—Again, 3 classes of exposure were determined on a worst-case basis. Although in most cases the actual stressor would be made up of mixtures of toxic substances, these classes refer to exposure pathways and the corresponding media to which sediment or aquatic species are exposed. Similar to effluent effects assessments, which usually consider the proportion of wastewater in the receiving medium (Isnard 1998), aquatic species will be exposed to porewater fluxes mixed into the water column. For the benthos assessment endpoint, the classes of exposure refer to the ratio (DS/TS) of the surface covered by the deposit (DS) to the total surface of the quarry (TS). The thresholds for the DS/TS ratio were set at 25% and 50% (Table 1). For the water column assessment endpoint, these classes refer to the ratio (PWV/WCV) of the porewater volume (PWV) to the total volume of water in the quarry (WCV). Both volumes were estimated, the former according to the percentage of water in the bulk sediment, the latter from the dimensions of the quarry. The exposure threshold was set at 1% to get an exposure-to-effect ratio (i.e., a hazard quotient) of 1 or more when C. dubia or B. calyciflorus tests had EC20s below this value. In that case, the French waste classification system would consider the material hazardous (MATE 2002), equivalent to a class 2 in the proposed approach. Because a risk was deemed to occur when exposure or effects or both were in class 2 (Table 2), the exposure limit for class 2 could therefore be back-calculated from a hazard quotient of 1 and the class 2 threshold for effects. The set of thresholds for both assessment endpoints is given in Table 1.
Risk characterization—The evidence of risk was drawn from the combination of the outcomes of effects and exposure assessments; the definitions of risk are summarized in Table 2. These definitions reflect that risk classes, which can be expressed as numbers (0, 1, 2) as well as qualitative adjectives (negligible, intermediary, high), were considered equivalent to scores, which allowed the combination of individual risks to benthic organisms into an overall risk to benthos. Furthermore, no priority was given to exposure or effects; in other words, the risk resulting from the combination of effects in class 1 and exposure in class 2 was considered equivalent to the risk resulting from effects in class 2 and exposure in class 1.
These definitions apply to the 2 assessment endpoints related to benthic and aquatic populations; for each of them, the final risk score corresponds to the mean of the individual risk scores determined for the corresponding measurements. The risk characterization process for the assessment endpoint related to benthos is shown on Figure 3.
|Class of effects or exposure|
|Assessment endpoint||Measurement endpoint||0||1||2|
|Benthic invertebrate population disturbance (structure/abundance)||Effect|
|Chironomus riparius (survival and growth)||Mortality or growth decrease ≤10%||10% < decrease ≤50%||Mortality or growth decrease >50%|
|Hyallela azteca (survival and growth)||Mortality or growth decrease ≤10%||10%< decrease ≤50%||Mortality or growth decrease >50%|
|DS/TS, ratio between the deposit surface and the total surface||≤25%||25%< ratio ≤50%||>50%|
|Chronic toxicity to aquatic organisms||Effect|
|Algae Pseudokirchneriella subcapitata or Chlorella vulgaris||Cell number × 16 (pH increase <1.5)||IC50 < 1 0%|
|Ceriodaphnia dubia (survival and reproduction)||Adult female survival ≥80%, >3 broods for at least 60% females, 15 young per brood on average||EC20 < 1%|
|Brachionus calyciflorus (survival and reproduction)||Mortality < 10%; reproduction rate > 4||EC20<1%|
|PWV/WCV, ratio between pore water and water column volumes||≤0.1%||0.1%< ratio ≤1%||>1%|
Uncertainty evaluation—In a risk management perspective, it seems preferable to discriminate between actual uncertainty, which is related to knowledge as well as to measurement errors, and variability, which is inherent to population characteristics (Kelly and Campbell 2000). Among the known sources of uncertainty in ecological risk assessment (USEPA 1998), those related to the functioning of ecosystems or to the design of the conceptual model will have a nearly constant effect on the overall uncertainty if the conceptual model and the measurement endpoints are fixed. Accordingly, if the proposed framework is deemed applicable by the authorities, they would principally need to know about site-specific (or project) uncertainty. Provided the projects apply the same framework, use of the same set of tests and the number of uncertainty or weight attributes described primarily in the MWWG approach (Menzie et al. 1996) and in the shipyard case study (Johnston et al. 2002) can thus be reduced. Four criteria were selected for exposure assessment: 1) The strength of association between the measurement endpoint and the assessment endpoint, 2) spatial representativeness, 3) temporal representativeness, and 4) the use of a standard method. For the effects assessment, the 4 proposed criteria are 1) the strength of association between the measurement endpoint and assessment endpoint, 2) the distinction between effect and no effect, 3) the sensitivity of the measurement, and 4) the use of a standard method.
Although not mentioned in this list, data quality is considered a crucial issue; it is assessed separately, before the uncertainty assessment itself. If the quality of a given measurement was deemed acceptable according to current criteria (e.g., validity requirements of the standards), then the uncertainty would be evaluated as described below. Conversely, if the quality was poor, the measurements concerned should be taken again. Alternatively, in particular when the quality estimation is close to the requirements, a measurement could be accepted and a penalty added to the uncertainty final score.
The uncertainty for each measurement endpoint was then compiled following a score ordination approach (Jouany et al. 1982; Vaillant et al. 1995). In such an approach, each criterion—in our case the 4 criteria selected for exposure and the 4 for effects—can take 3 values: High (H), Medium (M), or Low (L), which represents the confidence in using the measurement to infer a risk. It is acknowledged 1st that the different criteria do not have the same influence on the final ranking. Rather than attributing an arbitrary weight to the criteria, they are simply ranked according to their expected influence on the parameter in question. The expectations here stem from assessors' experience; their acceptability for risk managers would be enhanced by obtaining a consensus among experts. A penalty table is built starting from the right (i.e., the criterion having the less weight) and filling the rows to the left, as shown in Table 3 which gives 1 of the application's insights into these principles for 2 criteria. Rows are monotonically ordered in the penalty table because, in that case, there is no interaction between criteria; for instance, the fact that a toxicity test is sensitive (criterion 3) would not interact with its relevance as a measurement related to an assessment endpoint (criterion 1). With 4 criteria and 3 values per criterion, the maximum number of ranks is 81 (i.e., 34). Because the current number of measurement endpoints for exposure or for effects is much smaller, the ranks are scaled from 1 to 10. The complete penalty table for 4 criteria with the corresponding ranks can be found in the Appendix Table.
For a given measurement, the uncertainty evaluation yields an ordered series of 4 letters; the corresponding score, or rank, is found in the penalty table.
In a 1st step, this scoring approach is applied to each measurement. Then, the arithmetic mean of these scores (or ranks) is calculated for each assessment endpoint. The uncertainty evaluation process for 1 assessment endpoint, namely benthos populations maintenance, is summarized in Figure 3.
Chemical analyses were processed according to existing standards: NF T 90–112 for metals, except mercury, which was analyzed according to NF EN 1483; and NF EN ISO 11885 for arsenic. Sediment samples were 1st mineralized according to NFISO 11466. PAHs and PCBs were extracted by a mixture of dichloromethane and acetone at 100 °C under 100 bars (accelerated solvent extraction procedure) and analyzed by GC-MS, GC-ECD or HPLC fluorimetry; various standards were used (NF T90 115, NF EN ISO 6468, XP X33 012, EPA 525, EPA 610), depending on the substance and the matrix.
Bioassays were carried out according to existing standards; C. riparius following the French Standards AFNOR XP T 90 339–1, H. azteca AFNOR XP T 90 338–1, B. calyciflorus AFNOR T 90–377 (microplate), and C. dubia AFNOR T90 376. Algal tests were done according to AFNOR NF T 90–375 with P. subcapitata in microplates (72-h exposure), but Chlorella vulgaris instead of P. subcapitata was used in the CEBS study, with a protocol derived from the former standard AFNOR NF T 90–304.
|Criterion 1||Criterion 2||Rank|
|H||H-H = 1|
|H||M||H-M = 2|
|L||H-L = 3|
|H||M-H = 4|
|M||M||M-M = 5|
|L||M-L = 6|
|H||L-H = 7|
|L||M||L-M = 8|
|L||L-L = 9|
EC20s and associated confidence intervals (B. calyciflorus, C. dubia) were determined by logistic regression (Isnard et al. 2001).
The surface of the quarry was set equal to 2,500 m2 (50 × 50 m) and the depth to 10 m. Assuming a sediment deposit 2 m thick on average, a 500-m3 deposit would cover 10% and a 3,000-m3 deposit 60% of the total quarry surface. The sediment moistures ranged from 28% (NPDC study site 13990) to 70% (NPDC study site 17000; Table 4). Thus, for a 2,000-m3 deposit, the porewater volume to water column volume ratio (PWV/WCV) would vary between 0.024 at site 13990 and 0.061 at site 17000; according to the proportion of fine particles (i.e., D50% [particle diameter corresponding to 50% of the particle size distribution]), all the sediments but 1 were predominantly made of silt, clay, and organic matter (Table 4).
Various metals and arsenic were analyzed in bulk sediment and porewater along with PAHs (16 compounds) and PCBs (congeners 28, 44, 52, 101, 105, 118, 138, 153, 170, 180, 194, and 209). According to mean quotient values, calculated as described in (MacDonald et al. 2000), the most contaminated location was the 17000 site (NPDC study), with a mean quotient value of 2.28. This value was explained by rather high concentrations of cadmium, chromium, lead, and zinc. Sum-PAHs concentrations in sediments ranged from 5863 μg/kg (dry weight) to 37,640 μg/kg (dry weight); only 2 samples displayed concentrations above the probable effects concentration of 22,800 μg/kg (dry weight; MacDonald et al. 2000). PCBs were low or not measurable, except at site 17000, where the sum-PCB concentration exceeded the probable effects concentration.
Porewater concentrations exceeded the chronic toxicity water quality criterion available in France (Oudin and Maupas 2003) for cadmium, chromium, copper, nickel, and zinc and the PAHs in the NPDC study (almost all samples) and exceeded the water quality criterion for arsenic and copper at B22 in the CEBS study. Ammonium was also measured in porewaters; the concentrations ranged from less than the quantification limit for CEBS sites to 254 mg/L in site 17000 porewater. Porewater at other NPDC sites was either of poor (site 13990) or bad quality (12570, 12730, and 12800) according to French water quality criterion (Oudin and Maupas 2003).
Among the CEBS sediments, B13 was the most hazardous. C. riparius survival was 88.6% (26.1), not significantly different from the control, but the mean growth in test replicates was only 38% of the control. H. azteca survival for this sediment was 41.4% (24.8), but the control test did not match the validity criteria of the standard. According to Table 1, this sediment was then in class 2 for effects on C. riparius and in class 1 for those on H. azteca. Site B22 sediment induced a moderate C. riparius growth inhibition at 69% of the control and did not affect its survival or either endpoint for H. azteca. Therefore, this sediment was in class 1 for C. riparius and in class 0 for H. azteca. Site B2 sediment was not toxic to any of the tested species. All the NPDC sediments but 1 (site 17000) moderately affected C. riparius survival; the survival rates were, however, above 80%, so sediments 13990, 12570, 12730, and 12800 were all in class 1 for that endpoint. Conversely, no effect could be observed on growth, except for site 17000, where the observed growth was about 70% of the control. According to Table 1, this sample was therefore categorized as class 1. These sediments also highly affected H. azteca survival, except for site 13990; survival rates ranged from 0.0% at 17000 to 12.9% (11.1%) at 12730; they thus all belong to class 2 for this endpoint. Again, no effect was found on H. azteca growth.
Porewater test results are presented in Table 5. Brachionus calyciflorus tests were successfully carried out on all porewater samples; in the NPDC study, reproduction EC20s ranged between 5.9% (4.0–7.6%) at 17000 and 90.1% (53.1–133.5%) at 13990. In the CEBS study, site B2 and B22 porewaters were slightly more toxic to B. calyciflorus than B13. Because of practical problems (i.e., porewater volumes that were too small), C. dubia results were realized on waters collected after decanting the sediment samples. In the NPDC study, the C. dubia reproduction EC20s ranged between 4.7% (1.5–10.0%) and32.9% (26.4–38.6%) for sites 12800, 12730, and 17000, whereas for the other sites, responses were close to the control. In the CEBS study, the water samples collected after decantation strongly affected the survival of C. dubia; moreover, the observed effects seem inconsistent because they varied independently of the test concentrations. Neither survival nor reproduction EC20s could thus be determined. For the same reasons as for C. dubia, algal tests were applied to waters collected after decantation. In the NPDC study, all samples induced rapid growth, up to 200% compared with the control for sample 17000. Because this situation was not anticipated in the problem formulation phase, no score could be assigned for algal tests applied to NPDC samples. In the CEBS study, algal growth for sample B2 was not distinct from the control; however, splashes formed in the wells, suggesting that an immeasurable effect could have occurred. Seventy-two-hour EC50s for the B13 and B22 sites were 5.5% (0.5 %) and 7.2% (0.9%), respectively. Accordingly, both results were assigned a score of 2.
The classification of effects for the assessment endpoint related to benthos is reported in Table 6. For each site, the class was 1st determined for each measurement (e.g., C. riparius survival); the score for the measurement endpoint (e.g., C. riparius) was set equal to the highest score among the existing measurements for that endpoint. Assuming a 2,000-m3 deposit, that is, a surface ratio (DS/TS) of 40%, the risk for benthic organisms would range from negligible at site B2 to high at sites 17000, 12570, 12730, 12800, and B13 (Table 7). Figure 4 shows the influence of the deposit size on the risk to both sediment species, and for the assessment endpoint as a whole.
|Site||Moisture (%)||Grain size (D50, μm)||Porewater volume (m3)||PWV/WCV|
For the same deposit volume, the PWV/WCV ratio, which represents the exposure of aquatic organisms, would never exceed 0.061; therefore, the class would be 0 for all sites. Thus, the maximum risk scores for pelagic species would be 1 (negligible) at sites 17000, 12730, and 12800. Even an increase in the deposit volume up to 3,000 m3 would not yield PWV/WCV ratios above 0.1 (Figure 5).
The uncertainty scores for effects measurements at each site are reported in Table 7, along with the uncertainty scores for both assessment endpoints. No evaluation is provided for exposure because the deposit sites are virtual. In the uncertainty evaluation process, some elements are not site or study specific: this is the case for the criteria Strength of Association between the Measurement Endpoint and Assessment Endpoint and Use of a Standard Method. For these criteria, the value will essentially be the same at each site. Conversely, for the 2 other criteria, the score can vary as a function of the actual data.
Strength of association between measurements and assessment endpoints—The C. riparius test was rated H for this criterion, whereas H. azteca was rated M because the former species is endobenthic and feeds on sediment exclusively whereas the latter is epibenthic and feeds either on sediment or from the water column (ASTM 2000; Bonnet 2000). Ceriodaphnia dubia and B. calyciflorus tests were both rated H for this criterion because both organisms are invertebrates commonly found in surface waters and because both tests use chronic endpoints. The growth of unicellular algae such as P. subcapitata or C. vulgaris also seems appropriate for describing chronic toxicity in the water column; these species are often used to assess the toxicity of chemicals (Lewis 1995). However, the tests were done on supernatant water in both studies and not on porewater. Therefore, this criterion was rated M for all the sites.
Distinction effect/no effect—The values for these criteria should be adjusted according to the biotest results. No particular difficulty or bias was noted during any of the C. riparius tests for all sites. Therefore, the value H was attributed to this test in all cases. H. azteca data were more dissimilar; in the CEBS study, the control test did not match the validity criteria of the standard, yielding a value of L for the 3 samples. Conversely, a value of H was assigned to the NPDC sediments because the control was valid, and the survival rates were clearly either statistically different from the control or not different. In the CEBS study, C. dubia did not yield consistent dose-response curves, so this attribute was rated L for sites B2, B13, and B22. Although the curves appeared more consistent in the NPDC study, they remained incomplete. Therefore a value of M was attributed for these sites. Conversely, EC20s with their confidence intervals could be calculated for B. calyciflorus tests in both studies. The value H was thus attributed. For 6 of the 8 sites tested, it was impossible to show any toxic effects to algae because of a confusing growth effect (NPDC study) or the formation of splashes (B2); consequently, the score assigned to these sites was L for this criterion. For the other 2 sites (B13 and B22), the scores were M because the EC50s could be determined but the confidence intervals were only estimated.
|Brachionus calyciflorus||Ceriodaphnia dubia||Algae|
|Site||EC20 (%)||Cl||EC20 (%)||Cl||EC50 (%)||Cl|
|Overall risk for benthic organisms||1.50||2.50||2.50||2.50||2.50||1.00||2.50||1.50|
|Overall uncertainty (benthos)||2.04||2.04||2.04||2.04||2.04||3.15||3.15||3.15|
|Overall risk for pelagic organisms||0||1||0||1||1||0||1||1|
Sensitivity—The C. riparius is generally considered an organism with rather poor sensitivity, whereas H. azteca is usually considered a rather sensitive organism (Ankley et al. 1991; West et al. 1993; Kemble et al. 1994). These statements could be refined according to the contaminants actually found in the samples because the sensitivity of each organism depends on the nature of the contaminant (e.g., metals, PAHs; Phipps et al. 1995). Nevertheless, both CEBS and NPDC sites displayed a range of contaminants, including metals, PAHs, and sometimes PCBs; thus, introducing differences between sites for this criterion would be inaccurate. Consequently, C. riparius was given a value of M at all sites, whereas H. azteca was rated H. Ceriodaphnia dubia and B. calyciflorus are also rather sensitive species (Nebeker et al. 1984; Giesy et al. 1990; Snell and Moffat 1992; Snell and Carmona 1995) and were thus both given the value H for this criterion. Both algal species (P. subcapitata and C. vulgaris) are usually considered sensitive test organisms (Cullimore 1975; Blanck et al. 1984), so the value H was assigned to all measurements.
Use of a standard method—French (AFNOR) standards are available for all the tests used in these studies, and these standards were applied at all sites, so this attribute was rated H in all cases, except for H. azteca tests, and for the algal tests in the CEBS study. For H. azteca tests, this criterion was rated M because the standard is still being developed. The same score M was also given to algal tests in the CEBS study because the tested species differed from that requested by the standard.
The sequence for C. riparius tests was thus H-H-M-H at all sites; according to the penalty table in the appendix, this sequence ranks 4th out of 81 combinations (4/81), yielding a score of 0.49. For H. azteca, the resulting sequence was M-H-H-M for NPDC sediments, ranking 29/81, with a score of 3.58. Moreover, the final sequence for CEBS sediments was M-L-H-M (47/81), yielding a score of 5.80. Note that the sequences are identical for all sites of a study only because all the tests were done in parallel. The final sequence for C. dubia tests was H-L-H-H in the CEBS study (19/81, 2.35) and H-M-H-H (10/81, 1.23) in the NPDC study. For B. calyciflorus, the sequence was H-H-H-H (1/81, 0.12) at all sites. The final sequence for algal tests varied from M-L-M-H at B2 (CEBS study; 49/81, 6.05) to M-M-M-H at B13 and B22 (40/81, 4.94). Algal results in the NPDC study were M-L-H-M (47/81, 5.80).
Risk scores and related uncertainties are summarized in Table 7; in the tested exposure conditions, namely a 2,000-m3 deposit, the risk to benthos would be negligible to high, depending on the origin of the dredged materials, with an uncertainty ranging from about 20% to slightly more than 30% of the total span. For pelagic species, the risk would be negligible in all cases, with an associated uncertainty slightly below 25% of the span.
This joint approach to risk characterization and uncertainty assessment is made up of several parts: Hazard assessment, exposure assessment, risk characterization, and uncertainty evaluation. It characterizes a risk to selected components of the ecosystem and estimates the associated uncertainty. Nevertheless, the overall relevance of this approach basically depends on the relevance of each component.
In particular, it is assumed that the exposure of benthic organisms is best represented by the ratio of the surface covered by the dredged material deposit. This proposal is quite different from that of Johnston et al. (2002), who based their exposure assessment on classes of chemical quality using toxicological benchmarks. This approach was considered inappropriate because it is in fact a rough risk characterization rather than an exposure assessment. It thus became somewhat redundant with the effect assessment on the basis of a set of bioassays. For the benthic organisms living in the quarry, hazard is essentially attributed to known or unknown chemicals carried by particles of contaminated dredged materials; the exposure is basically linked to the quantity of dredged materials brought into the quarry. In that case, what could be the best descriptor for exposure within this assessment endpoint? Assuming that benthic organisms live mostly in the top few centimeters of sediments (Rasmussen 1984; Ward 1992; Tachet et al. 2000), the recovery/recolonization process taking place after dredged material deposition will start from deposit-free areas; the higher the proportion of deposit-free areas, the more efficient the recovery. Therefore, the critical factor is the ratio of the deposit surface to the quarry's total surface area, rather than the deposit volume. In addition, the exposure assessment for the water compartment is also based on a ratio of the porewater volume to the water column volume. This concept is similar to the dilution factor commonly used for liquid stressors such as effluents (e.g., Costan et al. 1995; Vindimian et al. 1999). The use of ratios for both compartments accordingly provides the advantage of coherence.
In this study, the benthos exposure ratio calculation is somewhat rough because it is based on an average deposit thickness. This was taken because no deposit area was available. In the real world, the deposit would be thicker in its central part and thinner on the edges. Nevertheless, an average thickness could be estimated, so the use of an average by itself is not an issue. What has to be addressed is the prediction of the deposit area and the related uncertainty.
This approach to risk characterization was designed to complete a risk assessment framework on the basis mainly of standard methods. In this context, standardization is expected to help achieve sound risk assessment studies at a reasonable cost. The risk characterization and uncertainty assessment approach itself should also be standardized to a certain extent to support the data processing step, ensure consistency, and limit variability of the assessors' conclusions.
The application of classification (ranking) approaches is customary in environmental quality studies (e.g., Wildhaber and Schmitt 1996; McDonald et al. 2002) and in hazard assessment approaches (see, e.g., Mitchell et al. 2002; Boxall et al. 2003; Gupta and Edwards 2003). Some risk assessment applications exist, particularly those that use discrete data or whose influence on risk is not accurately known, and a discrete but robust output is sufficient, as in some instances of pesticide management (Spugnoli and Vieri 1998; Sanchez-Bayo et al. 2002; Padovani et al. 2004). Because at least some of the data used for characterizing ecological risks of sediments are discrete, it seems logical to apply a ranking approach to the characterization of the ecological risk of sediments, as MWWG (Menzie et al. 1996) and later Johnston et al. (2002) did. One of the issues, however, is the number of classes for effects, exposure, and finally risk. For instance, Johnston et al. (2002) used 5 classes for exposure, whereas this approach relies on 3 classes only, although this might be less discriminating. They finally retained 4 classes for risk classification. The capacity for assessing the magnitude of risk, or for discriminating among different classes of risk, is nevertheless not independent of the options available to risk managers. Three conclusions are currently available to them; the considered option could either be accepted, if the risk was negligible, or rejected, if the risk was too high, or the conclusion could be uncertain. Sediment quality assessment schemes governed by empirical guidelines, such as effect range low or medium (ERL/ERM; Long et al. 1995) or threshold effect level and probable effect level (TEL/PEL; MacDonald et al. 1996; Smith et al. 1996), also frequently rely on 3 classes, which in turn means the influence of either type I or type II errors can be minimized on the classification result. Assuming that the optimal number of classes for risk is 3 in general, 3 classes for effects and for exposure seem sufficient in most cases.
The evidence of risk as summarized in Table 2 reflects the sum of the respective scores for effects and exposure. This rather simple scheme leads to ranking risk on a constant basis—the same class will be assigned to a given risk score, whatever the combination of exposure and effects. High risk is obtained either by the combination of moderate effects and high exposure or moderate exposure and strong effects. Moreover, exposure and effects assessments are given the same importance, which was not the case in the scheme adopted by Johnston et al. (2002): in their scheme, only high or elevated exposures (i.e., classes 3 and 4, respectively) can yield a high risk. However, 2 of the combinations displayed in Table 2 could be problematic: effects in class 0 and exposure in class 2 or effects in class 2 and exposure in class 0. For effects on pelagic species, a score of 0 means that the test would yield results that are not statistically distinct from the control, whereas for chironomids, it corresponds to conditions close to or stricter than control conditions in the Organization for Economic Cooperation and Development guidelines (Den Besten et al. 1995) or in the current French standard (Table 1). Thus, classifying these samples at intermediate risk would be a cautionary decision, which can be justified because exposure classes were defined arbitrarily and because of uncertainty associated with toxicity testing. In vitro toxicity testing is based on the assumption that the toxicological sensitivity in the laboratory is predictive of, or comparable to, the sensitivity of the test organisms in the field (Underwood 1995 cited in DeWitt et al. 1999). Such a validation is seldom practiced or even possible (Chapman 1995), particularly when nonindigenous species are used. This is the case for H. azteca, which is not present in European freshwaters. Although it can be agreed that it is highly desirable to use indigenous instead of nonindigenous species (Cairns 1993 cited in Chapman 1995), it was deemed preferable to have 2 test species in the procedure to account for several routes of exposure. This could also be an issue, however, specifically for H. azteca; according to the review by Wang et al. (2004), test conditions increase H. azteca exposure to contaminated sediments compared with the field, although DeWitt et al. (1999) were less conclusive. Reducing this uncertainty associated with toxicity testing within this framework would take validation studies on actual deposits. This has not been possible until now but should be envisaged in the future to improve the framework.
Furthermore, the same cautionary approach should prevail when effects are in class 2 and exposure in class 0 because the exposure class limits were set arbitrarily. Therefore, these 2 combinations should not be considered problematic per se; the main difficulty here is rather to establish relevant exposure class limits, which now depends on field experiments attempting to validate this approach.
Setting class boundaries is another critical issue. These boundaries should ideally be related to ecosystem functioning. Single-species bioassays are rather poor surrogates from this perspective because they are done on few individuals and their results are seldom extrapolated to the population level (Congdon et al. 2001) and because they cannot account for interactions among species (Cairns 1983; Kimball and Levin 1985). Therefore, assigning hazard thresholds to bioassay results with respect to ecosystem protection is somewhat tricky, if not completely arbitrary. This general consideration led regulators to adopt arbitrary limits. For instance, the hazard threshold for toxic waste leachates is set at 1 % dilution in France (MATE 2002). This regulation, however, cannot be used for setting effect thresholds adapted to benthic species.
Although they use 3 classes of increasing hazard, Johnston et al. (2002) defined them essentially in terms of probability of effect: When the measured biological variables are similar to the reference or control condition, the sample belongs to the No Effect class, whereas it is assigned to the Probable Effects class when results are statistically different from the reference or control condition. Otherwise, the sample would be classified as Potential Effects. In this scheme, however, it seems difficult to clearly fix the boundary between No Effects and Potential Effects. Similarly, in many sediment quality assessment studies (e.g., USEPA 1999; Ingersoll et al. 2002), with toxicity tests on benthic species, the responses are usually qualitative (i.e., the sediments are classified as either toxic or nontoxic). This in effect makes it possible to determine only 2 classes of hazard and does not help to evaluate the magnitude of the hazard. Although it is not possible to determine dose-response curves with such benthic toxicity tests, the fact that more or fewer individuals display the effect sought (e.g., mortality or growth inhibition) is an indication of the magnitude of effect, exactly as in toxicity tests involving pelagic species. Recent attempts to infer effects on chironomid populations from conventional C. riparius test results support this statement (Péry et al. 2002, 2003). Rather than trying to insert a hazard class on a qualitative and faulty statistical basis, it was deemed preferable to determine 2 classes of increasing hazard within the category of sediment actually classified as toxic, as did Den Besten et al. (1995). For this reason, in this scheme, the class boundary between the class 0 (no distinguishable toxicity) and class 1 (moderate or intermediate toxicity) corresponds to the validity conditions in control tests. The boundary between class 1 and class 2 derived from the above-mentioned Dutch studies (Den Besten et al. 1995) because the model developed by Péry et al. (2003) is not available yet for organisms other than chironomids.
Again, these sediment hazard thresholds stem from expert judgments; they cannot be based on indisputable observations or facts, nor can they really be discussed. The proposed thresholds should therefore be considered as interim values and reassessed later, when field data stemming from pilot studies or application of the framework would become available.
In the proposed framework, the initial ecological status of the burrow pit is ignored; it is thus assumed to be good and potentially disturbed by the deposit. However, monitoring this ecological status both before the deposit and during recovery would help to check the relevance of the assessment endpoints and to calibrate the effects and exposure of class boundaries.
Class boundaries for exposure of pelagic organisms derive directly from the hazard thresholds to avoid obtaining risk quotients of more than 1 in any class of exposure. Conversely, class boundaries for exposure of benthic species cannot be based on a similar rationale: benthic organisms are supposed to colonize the deposit from the surrounding areas. Thus the magnitude of exposure is related to the proportion of deposit-free areas, because the scenario considered occurs in a closed space. The time necessary for recolonization is not explicitly considered in this approach.
Compared with the 11 attributes listed by MWWG (Menzie et al. 1996), only 4 criteria for assessing uncertainty related to hazard and 4 for exposure were selected. Because this group did not clearly distinguish between effects and exposure assessments and preferred to consider an array of lines of evidence, there are in fact few differences between the 2 approaches. For hazards, 3 attributes are more or less redundant: the strength of association between the measurement endpoint and assessment endpoint, the stressor specificity, and the correlation of the stressor to the response. Johnston and coworkers (2002) pooled these attributes into a single one, which should make it possible to improve the applicability and the transparency of this procedure. Another difference stems from the framework design; it is assumed here that there is no need to assess the attribute said to be quantitative by MWWG (Menzie et al. 1996) because the set of measurement endpoints for each assessment endpoint is already fixed in the current scenario. Thus, uncertainty related to this attribute would be the same among all the studies on the basis of this framework and would not influence the results. Moreover, for the same reason, most attributes related to study design in the Johnston et al. (2002) study are no longer useful for the hazard component of uncertainty assessment. Conversely, some of these attributes are still essential for the exposure component (i.e., spatial representativeness and temporal representativeness). The MWWG left open the possibility of assessing data quality either as 1 attribute among others or as a preliminary attribute. This latter option was preferred because the measurements can be taken if necessary (Menzie et al. 1996).
Score ordination provides several advantages compared with other scoring methods. First, attributes are assumed to have various influences on overall uncertainty, which would not be the case if the final score was set equal to the mean value of attribute scores. Moreover, a specific weight for each attribute, which could be controversial and lack transparency, does not need to be determined. It is quite easy to handle, and fits well with most of the requirements listed by Burton et al. (2002) for WOE: Robustness, methodology, sensitivity, appropriateness, and transparency. Although proposed for WOE, these criteria can indeed have a broader application. The robustness of score ordination could be debatable because this item refers to the consistency in interpretation irrespective of when and where conducted; in the proposed context however, inconsistencies would probably come rather from the attribute scoring step than from the score ordination approach itself. Methodology refers to the ease of use, which seems obvious in this case. With 81 possible combinations for 4 attributes, this approach is not only sensitive but is also appropriate according to the criteria outlined by Burton et al. (2002); that is, it is applicable in a wide range of conditions. It also provides some flexibility in the framework design; some measurement endpoints could be added without changing the uncertainty assessment methodology.
All algal tests in the NPDC study showed growth significantly above that of control tests; this unexpected growth might have masked the porewater toxicity to algae and thus lowered the overall risk for pelagic species. It might also indicate that eutrophication could be an issue for the burrow pit. This issue was not identified in the problem formulation phase and should be addressed in future versions of the conceptual model. For instance, an assessment endpoint accounting for a eutrophication risk could be that the releases from the deposit should not stimulate phytoplankton growth. With this perspective, the choice of appropriate tests either for the measurement of toxicity to algae or for assessing the growth potential has to be reviewed. However, the standard used for algal toxicity tests (AFNOR NF T 90–375) is characterized by low amounts of nutrients in the culture medium. Thus, even limited N or P in porewaters would yield increased growth compared with the control, suggesting that it would be rather difficult to calibrate this type of effect with the current standard. Although some attempts have been made to assess the growth potential of unicellular algae as a function of nutrient bioavailability (Nyholm and Lyngby 1988; Dorioz et al. 1998), their suitability to the issue at hand remains questionable. Dorioz et al. (1998) used 2 different experimental designs, 1 in which sediment particles and algae were mixed and another in which sediment particles were removed by filtration at 0.22 μm. Their algal potential growth test was not standardized. Nyholm and Lyngby (1988) discussed the use of an algal bottle test, based on a former USEPA standard, combined with mathematical modeling. In this case, the test design is nevertheless oriented to the identification of growth-limiting factors.
As mentioned in the Results section, many chemical analyses were performed on bulk sediments and porewaters. Although bulk sediment analytical results were used in tier 1 of the framework (not discussed in this paper), the porewater data were not considered in tier 3. Although several metals (Cd, Cr, Cu, Ni, and Zn) and PAHs were present at rather high concentrations in the porewaters of most samples, only slight toxic effects could be observed in some instances. The most toxic samples (B13 and B22) conversely displayed effects on algae, whereas the contaminants measured in porewater remained at presumably nontoxic levels, except for Cr or Cu (both sites) and As (B22 only). These analyses were performed in the context of a test within the risk assessment framework and were planned with the perspective of providing some insights into the origin of toxicity, rather than complementary lines of evidence. It appears that they are not really essential to the proposed risk characterization approach. Thus, they would probably not be retained in the framework of the routine application.
An approach to ecological risk characterization of freshwater dredged material deposits has been proposed and was tested on a limited number of sediments containing a range of contaminations. This approach includes an estimate of the associated uncertainty, which is assessed following a score ordination approach on the basis of 4 criteria each for exposure and effects. The risk characterization approach itself is based on scores, so that classes of effects, classes of exposure, and finally classes of risk can be determined. These classes of risk are linked to management decisions: Disposal Allowed; Disposal Prohibited; and Uncertain Response, Requires Further Investigation. Thus, an almost standard approach for assessing the risks of these dredged materials in the case of disposal in burrow pits could be developed on this basis. Because the exposure assessment accounts for the pit characteristics, the manager can adjust the deposited amounts to maintain the ecological risk at an acceptable level. This flexible approach fits the requirements of robustness, methodology, appropriateness, and transparency quite well (Burton et al. 2002). More case studies are needed to substantiate sensitivity.
In the scenario and conditions tested, and according to the exposure boundaries adopted, the risk to benthos appears more critical than the risk to aquatic (pelagic) species. Although the exposure hypotheses for the water column were rather conservative, strong effects on benthic species can occur before any risk to aquatic species are observed. Again, this conclusion was reached on the basis of few cases and should not be considered generally applicable until a wider range of situations are assessed.
The main difficulties encountered while designing this approach reside in exposure assessment. It is assumed that exposure for benthos is best represented by the surface of the deposit. Moreover, the definition of class boundaries is a key issue: Could microcosm or mesocosm studies help improve their definition or validate the proposed values? At least some expertise in ecology is needed here.
Both exposure and effects class boundaries would benefit from benthos monitoring before the deposit and during the recolonization phase.
This study was supported by grants from the Ministry of Equipment and Transport and Voies Navigables de France (VNF). The program was coordinated by Christophe Charrier, Centre d'Eudes Techniques Maritimes&Fluviales (CETMEF) and David Bécart (VNF). We are grateful for the contribution of the Technical Committee, which supervised this program: Geneviève Golaszewski (Ministry of Ecology and Sustainable Development), Claude Alzieu (IFREMER), Michel Albrecht (CETMEF), Bernard Solente (VNF), and Hubert Verhaeghe (Artois-Picardie Water Agency). The authors thank Linda Northrup for linguistic assistance and 4 anonymous reviewers who provided many valuable comments.
Disclaimer—The views presented in this paper are those of the authors and do not necessarily represent opinions or statements endorsed by the French Ministry of Ecology, the CETMEF (Centre d'Etudes Techniques Maritimes&Fluviales), or VNF (Voies Navigables de France).