Volume 6, Issue 2 p. 225-239
Original Research
Free Access

Efficiency of sediment quality guidelines for predicting toxicity: The case of the St. Lawrence river

Mélanie Desrosiers

Corresponding Author

Mélanie Desrosiers

Cemagref UR MALY, 3bis quai Chauveau CP 220, 69336 Lyon Cedex 9, France

Centre d'expertise en analyse environnementale du Québec, Ministère du Développement durable de l'Environnement et des Parcs du Québec, 2700 Einstein Street, Quebec City, Quebec G1P 3W8, Canada

Cemagref UR MALY, 3bis quai Chauveau CP 220, 69336 Lyon Cedex 9, France.Search for more papers by this author
Marc P Babut

Marc P Babut

Cemagref UR MALY, 3bis quai Chauveau CP 220, 69336 Lyon Cedex 9, France

Search for more papers by this author
Magella Pelletier

Magella Pelletier

Environment Canada, Science and Technology Branch, 105 McGill Street, Montreal, Quebec H2Y 2E7, Canada

Search for more papers by this author
Caroll Bélanger

Caroll Bélanger

Environment Canada, Environmental Protection Operations Division, 105 McGill Street, Montreal, Quebec H2Y 2E7, Canada

Caroll Bélanger has retired from Environment Canada.

Search for more papers by this author
Suzie Thibodeau

Suzie Thibodeau

Environment Canada, Environmental Protection Operations Division, 105 McGill Street, Montreal, Quebec H2Y 2E7, Canada

Search for more papers by this author
Louis Martel

Louis Martel

Centre d'expertise en analyse environnementale du Québec, Ministère du Développement durable de l'Environnement et des Parcs du Québec, 2700 Einstein Street, Quebec City, Quebec G1P 3W8, Canada

Search for more papers by this author
First published: 05 February 2010
Citations: 12

Abstract

Multitiered frameworks that are designed for risk assessment of contaminated sediment rely on sediment quality guidelines (SQGs) at the first tier or screening level. In the case of contamination by multiple pollutants, results can be aggregated under indices such as the mean quotient. A decision is then reached (e.g., to dispose of dredged materials in open water) without further investigation, provided that the SQGs or the specific values of indices or quotients derived from the SQGs are not exceeded. In this way, SQGs and quotients play a critical role in environmental protection. As part of the development of a tiered framework to assess the environmental risk of materials dredged from the St. Lawrence River, we evaluated various quotients based on SQGs available for this river with a data set that matches chemistry and toxicity test endpoints. The overall efficiency of all tested quotients was rather low, and we then examined factors such as sediment grain size, nutrients, metal-binding phases (e.g., Al, Fe), and dissolved organic carbon to explain misclassified samples. This examination led to the design of a modified tier 1 framework in which SQGs are used in combination with decision rules based on certain explanatory factors. Integr Environ Assess Manag 2010;6:225–239. © 2009 SETAC

INTRODUCTION

The St. Lawrence River, one of the world's largest waterways, is an essential transport route for northeastern North America. Its contribution to the economy of the region is expected to increase as seaway-based transport is deemed more sustainable than many other types of transportation. Periodic dredging is required at various sites along the waterway, mainly for the maintenance of the St. Lawrence Seaway system and harbor installations. The assessment and management of risks to the environment posed by such dredging projects are required as part of a sustainable navigation strategy (D'Arcy and Bibeault 2004).

Most industrialized countries regulate and manage contaminated sediments and dredged materials in their waterways. Consequently, various ecological risk assessment (ERA) approaches have been developed, and sediment quality guidelines (SQGs) were included in many cases (Babut et al. 2005, 2006; Alvarez-Guerra et al. 2007; Apitz et al. 2007). In Canada, Environment Canada regulates ocean and estuarine dredging operations and open-water disposal through Environment Canada's Disposal at Sea Program (DAS) under the Canadian Environmental Protection Act, 1999 (CEPA 1999). The DAS Program administers a permit system for sediment management at sea based on a multitier process revised in 1999 (CEPA 1999; Agius and Porebski 2008). Environment Canada and the Ontario Ministry of the Environment recently developed a risk-assessment framework to evaluate contaminated sites (Environment Canada and Ontario Ministry of the Environment 2007). This new framework was developed primarily for the Laurentian Great Lakes and is intended mainly to provide guidance in making remediation decisions (Chapman and Anderson 2005; Environment Canada and Ontario Ministry of the Environment 2007).

In parallel, a process aimed at developing an ecological risk assessment guideline for the management of contaminated sediments was launched in 2004 in the province of Quebec by Quebec's Ministère du Développement durable, de l'Environnement et des Parcs (MDDEP), and Environment Canada as part of the sustainable navigation strategy requirements of the St. Lawrence Plan. This guideline is currently being developed specifically to evaluate dredging operations and open-water disposal in the freshwater section of the St. Lawrence River. Most disposal sites in this portion of the river are considered to be moderately dispersive; thus, dredged material released at a dumping site generates a deposit on the bottom. Subsequently, the deposit may be subject to erosion during higher flow-velocity periods or may be covered by the deposition of upstream materials during periods of lower flow. Eroded particles are deposited downstream, particularly in shallow sections. Dredged materials may contain a range of contaminants that could potentially affect aquatic organisms in the water column and in bottom sediment at deposit sites as well as at spawning areas downstream. Therefore, benthic invertebrates and pelagic organisms were deemed relevant biological targets for disposal sites, and fishes and benthic organisms for fish spawning areas.

Sediment quality guidelines have been determined for the majority of contaminants present in St. Lawrence River sediments based on the approach of the Canadian Council of Ministers of the Environment, using a percentile distribution of concentrations where toxicological effects were observed on aquatic organisms (CCME 1995). Five threshold values were derived to protect sediment-associated life: the rare effect level (REL), the threshold effect level (TEL), the occasional effect level (OEL), the probable effect level (PEL), and the frequent effect level (FEL). The PEL and TEL values were obtained by CCME guidelines (CCME 1995), whereas the remaining effect levels were determined specifically for sediment risk assessment in Quebec (Environment Canada and MDDEP 2007). These SQGs define all of the intervention levels needed for sediment management in Quebec under a diversity of contexts: prevention of sediment contamination, management of dredged sediments, and remediation of contaminated aquatic sites (Environment Canada and MDDEP 2007).

Two of the SQG effect levels are used to manage dredged sediments in the St. Lawrence River: the occasional effect level (OEL) and the frequent effect level (FEL; Table 1). Together, these effect levels categorize sediment into 3 classes of contamination that require different management decisions. Class 1 sediments, with contaminant concentrations below the OEL, can be disposed of in open water or be used for other purposes, because the probability of detecting adverse biological effects is deemed low. Class 2 sediments, with contaminant concentrations between the OEL and the FEL, have a higher probability of adverse biological effects, so open-water disposal can be considered a valid option only if proper toxicity tests demonstrate that the sediments will not adversely affect the receiving environment. Finally, class 3 sediments, with concentrations above the FEL, are prohibited for open-water disposal, because the probability of detecting adverse biological effects is very high. Class 3 sediments must be treated or safely contained (Environment Canada and MDDEP 2007). Sediment may be categorized according to the highest observed classification among all contaminants covered by the SQGs or by means quotients to characterize the risk of multiple contaminants at the screening level (Long and MacDonald 1998; Ingersoll et al. 2000; MacDonald et al. 2000; Fairey et al. 2001; Marvin et al. 2004).

Table 1. Ranges of metal and organic chemical concentrations in sediment from sampling areas on the St. Lawrence River, particularly in its 3 fluvial lakes (Saint-François, Saint-Louis, Saint-Pierre) and in the Montréal Harbor area
SQGs Lake Saint-François Lake Saint-Louis Lake Saint-Pierre Montréal Harbor
OEL FEL
As 7.6 23 2.1–8.3 1.9–11 1.2–3.9 0.57–40
Cd 1.7 12 <DL–1.7 <DL–2 <DL–1 <DL–3
Cr 57 120 10–41 12–92 16–63 15–380
Cu 63 700 13–57 10–58 16–62 15–3600
Hg 0.25 0.87 0.05–0.6 0.09–9.9 <DL–0.29 0.04–0.9
Ni 47 12–41 11–75 15–51 10.8–310
Pb 52 150 10–38 6–38 6–50 7.1–190
Zn 170 770 49–330 31–312 62–210 61–550
PCBtot 0.079 0.78 <DL–2.248 <DL–0.0772 <DL–0.392 0.0024–0.527
PAH high
 Benzo[a]anthracene 0.12 0.76 0.1–0.27 <DL–4.3 <DL–0.5 <DL–2.8
 Benzo[a]pyrene 0.15 3.2 0.09–0.31 <DL–5.1 <DL–0.7 <DL–2.9
 Chrysene 0.24 1.6 0.01–0.25 <DL–16 <DL–0.75 0.06–4.6
 Fluoranthene 0.45 4.9 0.17–0.44 <DL–9 <DL–0.67 0.1–5.5
 Pyrene 0.23 1.5 0.13–0.35 <DL–6.7 <DL–0.64 0.1–5.7
PAH low
 Acenaphtene 0.021 0.94 <DL <DL–0.58 <DL–0.04 <DL–1
 Acenaphtylene 0.030 0.34 <DL <DL–0.02 <DL–0.04 <DL–0.35
 Anthracene 0.11 1.1 <DL–0.04 <DL–1.2 <DL–0.13 <DL–5.9
 Fluorene 0.061 1.2 <DL <DL–0.59 <DL–0.04 <DL–2.6
 Naphthalene 0.12 1.2 <DL <DL–0.44 <DL–0.08 <DL–0.35
 Phenanthrene 0.13 1.1 0.04–0.1 <DL–4.2 <DL–0.3 0.07–9.7
  • Sediment quality guideline thresholds (occasional effect level [OEL] and frequent effect level [FEL]), which were developed for sediment dredging management in the province of Quebec, Canada, are also provided (Environment Canada and MDDEP 2007). Concentrations are presented in mg/kg. <DL indicates that minimum concentration was below the detection limit.

This paper evolved from the development of screening-level ERA guidelines for open-water disposal of dredged sediment in the freshwater section of the St. Lawrence River. The ERA guidelines are still in development and will be used if a promoter decides, in the context of a dredging project (port creation, maintenance of a navigation channel, etc.), to dispose of the dredged sediment in open water. In that case, the risk assessment process aims to answer the following question: “Will dredged sediment disposed in open waters have adverse effects on the receiving aquatic environment?”

The specific objectives of this study were to assess the ability of Quebec SQGs to predict sediment toxicity, to examine relationships between contaminants present in sediments and toxicity tests results, and to use these findings to improve the proposed assessment guideline for open-water disposal of dredged sediment. To achieve these objectives, the following 3 questions were addressed:
  • How do different aggregation methods compare?

  • What factors can explain discrepancies in sediment classification based on SQGs and toxicity tests?

  • What practical approaches can be adopted to deal with so-called confounding factors (nutrients, organic matter, etc.)?

MATERIALS AND METHODS

Study area and sediment sampling

This study focused on the fluvial section of the St. Lawrence River, Canada, a waterway that flows east from the Laurentian Great Lakes for over 240 km before reaching the Lake Saint-Pierre outlet (Figure 1). Sampling areas were located mainly in sedimentation zones (fluvial lakes, harbor zones, and river plumes) as identified by fine-particle deposition, potential dredged areas, and past or present point sources of anthropogenic contamination. During the fall of 2004 and 2005, sediments were sampled at 59 stations in 3 fluvial lakes (Lake Saint-François, Lake Saint-Louis, and Lake Saint-Pierre) and in the Montréal Harbor area (Figure 1). Ten stations were located at Lake Saint-François, 21 stations at Lake Saint-Louis, and 15 at Lake Saint-Pierre. Finally, 11 stations were sampled in the Montréal Harbor zone and two others downstream from the Island of Montréal (Figure 1). Among and within sampling areas, a wide variation in sediment grain size, organic matter content, and concentrations of nutrients, organic chemicals, metals, and metalloids was observed, with higher chemical concentrations observed in the Montréal Harbor area (Table 1; Desrosiers et al. 2008).

Details are in the caption following the image

Study area of the 59 stations in the St. Lawrence River (Canada) located in the 3 fluvial lakes (Saint-François, Saint-Louis and Saint-Pierre) and Montréal Harbor area.

Surface sediments were taken with a Shipek grab sampler (400 cm2). For each station, 20 to 25 L of sediments were collected and placed in clear polyethylene bags. Samples were placed in a bucket with ice for 24 to 30 h until their arrival at the laboratory, where they were stored in a cold chamber (4 °C). The sediments were then sieved through 2-mm mesh, manually homogenized, and subsampled for individual analyses within 24 to 48 h after sampling. Sediment porewater was extracted by 2 centrifugations, the first with whole sediment (3000 g, 20 min) and the second with retrieved porewater (10 000 g, 30 min) to remove suspended particles. Subsamples of interstitial water were kept for toxicity tests and measurements of dissolved organic carbon (DOC).

Chemical and biological data set

In this study, we used a data set for the St. Lawrence River, comprising 59 sediment samples generated specifically for this project and measurements of inorganic and organic contaminant concentrations, biological toxicity on benthic and pelagic organisms, and ancillary environmental characteristics (i.e., sediment grain size, nutrient concentrations, etc.). All of the chemical and biological methods used here were based on Quality Assurance and Quality Control (QA/QC) standardized protocols. We briefly summarize standard chemical methods (Tables 2 and 3) and provide a short description of toxicity tests, data treatment, and statistical methodologies.

Table 2. Summary of analytical methods for inorganic chemicals, nutrients, organic matter, and sediment grain size
Variables Matrix Methods Device Detection limit Reference
Al Sediment Argon plasma emission spectrometer after total recoverable extraction (HCl 2.4 N/HNO3 8 N; 3/1) Optima 3000DV; Perkin Elmer 12.0 mg/kg CEAEQ 2003a
As 0.27 mg/kg
Ca 17.0 mg/kg
Cd 0.22 mg/kg
Cr 3.0 mg/kg
Cu 2.1 mg/kg
Fe 18.0 mg/kg
Mn 1.1 mg/kg
Ni 0.6 mg/kg
Pb 1.2 mg/kg
Zn 2.5 mg/kg
Hg Sediment Thermal decomposition with atomic absorption DMA-80; Milestone 0.035 mg/kg

CEAEQ 2007

Total sulfur Sediment Infrared detection LECO SC-444 50 mg/kg

CEAEQ 2006a

Total organic carbon (TOC) Sediment Titration 0.05%

CEAEQ 2006a

Total Kjeldahl nitrogen (TKN) Sediment Colorimetric method Technicon model II 100 mg/kg

CEAEQ 2006b

Total phosphorus (TP) Sediment Colorimetric method Technicon model II 200 mg/kg

CEAEQ 2006b

Sediment grain size Sediment Hydrometric sedimentation Hydrometer: type 152H 0.1%

Pelletier et al. 2008

Dissolved organic carbon (DOC) Porewater Infrared detection Shimadzu model TOC-5000A 0.20 mg/L

CEAEQ 2003b

Table 3. Summary of analytical methods for organic chemicals
Variables Matrix Methods Device Detection limit Reference
PCBs Sediment Congener method performed by gas chromatography/mass spectrometry extracted with acetone/hexane (60:40) and dichloromethane GC/MS; Agilent, GC 6890N, MS 5973N 2–6 µg/kg

CEAEQ 2003c

Purification with silica and freshly activated copper
PAHs Sediment Performed by gas chromatography/mass spectrometry extracted with acetone/hexane (60:40) and dichloromethane GC/MS; Agilent, GC 6890N, MS 5973N 0.02–0.10 mg/kg

CEAEQ 2003d

Purification on silica
Pesticides organochlorine Sediment Performed by gas chromatography/mass spectrometry extraction with acetone/hexane GC/MS; Thermo Quest, GC trace GC et MS trace MS 1–18 µg/kg

CEAEQ 2003a

Purification on Florisil
Pesticides organophosphate Sediment Performed by gas chromatography/mass spectrometry extraction with ethyl acetate GC/MS; Agilent, GC 6890N, MS 5973N 5–260 µg/kg

CEAEQ 2003b

Pesticides aryloxyacid Sediment Performed by gas chromatography/mass spectrometry GC/MS; Agilent, GC 6890N, MS 5973N 1–7 µg/kg

CEAEQ 2006b

extraction with NaHCO3 and on C-18 column
Purification on silica gel
Pesticides toxaphene Sediment Performed by gas chromatography/electron capture detector (GC/DCE) GC/ECD; Hewlett Packard, GC 5890 série II, ECD 3.5 mg/kg

CEAEQ 2003f

Petroleum hydrocarbons (C10–C50) Sediment Gas chromatography flame ionization detector (GC-FID) GC/FID; Hewlett Packard, GC 5890 series II, FID 30 mg/kg

CEAEQ 2002

Extraction with hexane

Toxicity tests

Toxicity tests were performed on whole sediments with two benthic organisms (a chironomid, Chironomus riparius, and an amphipod, Hyallela azteca) and on porewater with two pelagic organisms (a rotifer, Brachionus calyciflorus, and the alga, Pseudokirchneriella subcapitata). H. azteca and C. riparius survival and B. calyciflorus reproduction were considered the most sensitive endpoints in this study for assessing the risks posed by contaminated sediments in the St. Lawrence River, and the growth rates of C. riparius, H. azteca, and P. subcapitata were inefficient for toxicity assessment (Desrosiers et al., in preparation; data not shown).

H. azteca and C. riparius were cultured in the CEAEQ laboratory according to standard methods (Environment Canada 1997; AFNOR 2003, 2004). Cultures were kept under static conditions with soft aeration, and 50% (v/v) of the water was changed weekly. Both organisms were fed with a suspension of TetraMin fish food (Tetrawerke) 3 times per week. All test organisms were acclimated to test conditions prior to testing.

Whole-sediment toxicity tests were conducted according to standard procedures (Environment Canada 1997; AFNOR 2003, 2004). From 3 to 14 d following the sediment sampling, toxicity tests were conducted at 23°C ± 1°C, with continuous gentle aeration and a 16-h light:8-h dark photoperiod (1,000 lux). Five replicate test chambers (600-mL beakers containing 100 mL sediment, 400 mL water, and 10 test organisms) were used for each sample (field sediments and control). For the C. riparius toxicity test, control sediment consisted of siliceous sand (106–250 µm diameter; Multi-Sable Ltée) previously conditioned in order to initiate microorganism colonization, one of the sources of food for invertebrates (Verrhiest et al. 2002). For the H. azteca toxicity test, control sediment consisted of 5% siliceous sand (500–1,000 µm diameter; Givesco), 20% Loire sand (250–500 µm diameter; Amilab), 63.25% thin siliceous sand (106–250 µm diameter; Multi-Sable Ltée), 10.0% kaolin clay (Silumine Art Techni-Céram), 0.1% calcium carbonate (Anachemia), 1.5% α-cellulose (Sigma-Aldrich), and 0.15% TetraMin fish food (Tetrawerke). Water used in the test was dechlorinated tap water (pH 8.0, conductivity 239 µs/cm, hardness 72 mg CaCO3/L). Chironomids were fed with 1.5 mL of a suspension of TetraMin (4 g/L) daily (AFNOR 2004). H. azteca were fed with 0.75 mL YCT (yeast, cereal leaves, trout chow; 2.2 g/L) daily (Environment Canada 1997). At the end of the 7-d exposure for chironomids and 14-day exposure for amphipods, the surviving organisms were retrieved from the sediment.

B. calyciflorus test organisms were 2-h-old females hatched from cysts obtained from Microbiotests, according to AFNOR procedures (AFNOR 2000). Toxicity tests were conducted according to AFNOR procedures and Snell and Moffat (1992). Note that, to minimize carryover of nutrients and EDTA, the P. subcapitata algal cultures used to feed the rotifers were centrifuged twice (2000 g, 15 min; IEC 3000) and resuspended in NaHCO3 (15 g/L). After 48 h, the total number of surviving female B. calyciflorus per well was counted under a microscope (×9–12; Leica Wild M8, Leica Microsystems). To simplify the discussion and allow for a better comparison with the whole-sediment toxicity test (undiluted), we prefer to present the results obtained with essentially undiluted porewater. Consequently, only the results of organisms exposed to a higher concentration (90%) are presented in this paper.

Sediments classification and chemical aggregation methods

Our main objective was to evaluate the predictive ability of SQGs for use in the management of dredged sediments, particularly when contamination by multiple pollutants was observed. In the context of developing an ERA framework for sediment management, it was essential to cross-validate sediment chemical screening and toxicity test results, to determine the percentage of misclassified samples, and to examine environmental characteristics that may explain observed toxicity without chemical evidence.

Sediment from each station was categorized as class 1 (contaminant concentration < OEL), class 2 (OEL < concentration < FEL), or class 3 (concentration > FEL) according to the highest observed classification among all contaminants covered by the SQGs (Table 1; Environment Canada and MDDEP 2007). Aggregated quotients were also considered as an alternative to characterize the risk of multiple contaminants at the screening level, similar to approaches presented in earlier studies (Long and MacDonald 1998; Ingersoll et al. 2000; MacDonald et al. 2000; Fairey et al. 2001; Marvin et al. 2004). We were interested in the application of an aggregation method that compares contaminant concentrations to their SQGs and is useful as a management tool for toxicity prediction. Consequently, we evaluated different kinds of quotients, 5 of which are presented here. Specifically, we used 8 metals (As, Cd, Cr, Cu, Hg, Ni, Pb, Zn), total polychlorinated biphenyls (PCBs), and 11 polycyclic aromatic hydrcarbons (PAHs) for quotient calculations (Table 1). SQGs exist for 13 PAHs (Table 1), but dibenzo[a,h]anthracene and 2-methylnaphtane were not detected in our sediments. When chemical concentrations were under the detection limit (DL), half of the detection limit was used. However, pesticides were not included in this paper, because they were always under the detection limits (Table 3). The predictive ability of this classification was assessed assuming that class 2 sediment represents a threshold where biological toxicity may be observed and more studies are needed for decision making. Consequently, the OEL threshold was used for quotient calculations.

We first used a mean quotient (Qmean1) including all of the contaminants covered by the SQGs:
equation image(1)
where Ci corresponds to the concentrations measured for each contaminant (As, Cd, Cr, Cu, Hg, Ni, Pb, Zn, total PCBs, PAH1, PAH2 . . . PAH11) and where SQGsi corresponds to the OEL sediment quality guideline determined for each contaminant.
The second mean quotient (Qmean2) was similar, but in this case the mean of the 11 PAHs ratios was used.
equation image(2)
The third quotient (Qmean3) was based on a previous investigation of the St. Lawrence River in which certain contaminants were found to have a greater influence on biological toxicity or were bioaccumulated (Desrosiers et al. 2008). In that study, we observed that H. azteca mortality was correlated mainly with Cd, Pb, and Zn, although we observed few effects on the growth. Additional unpublished evidence indicated that inhibition of B. calyciflorus reproduction was correlated with Cd, Pb, Zn, total PCBs, and 11 PAH congeners (Desrosiers et al., in preparation; data not shown). In this context, Cd, Cu, Pb, and Zn remain important chemicals to consider, and each of them represented a weight of 1 in the equation. Conversely, As, Cr, Hg, and Ni, which were related less strongly or not at all to toxicity in our study area, were grouped as 1 factor. Finally, only the 6 low-molecular-weight PAHs (recognized as more toxic than high-molecular-weight PAHs) were considered and grouped as 1 factor.
equation image(3)
The fourth quotient (Qadd) was based on concentration additivity. In this quotient, all concentrations measured for each contaminant (As, Cd, Cr, Cu, Hg, Ni, Pb, Zn, total PCBs, PAH1, PAH2 . . . PAH11) were divided by their own SQG. The summation of each ratio was done as follows.
equation image(4)

The last aggregation method tested in this paper, the SQI, was developed for the Laurentian Great Lakes (Grapentine et al. 2002; Marvin et al. 2004), and calculations were performed in SQI 0.1: Sediment Quality Index Calculator (an MS EXCEL workbook that contains macros; Canadian Council of Ministers of the Environment 2006). The SQI can be considered a Qmean that includes 3 elements: scope (i.e., the percentage of a variable over the guidelines), amplitude (i.e., the magnitude by which failed variables exceeded SQGs), and frequency area (i.e., the proportion of individual measurements over the guidelines within a group of sites; Grapentine et al. 2002; Marvin et al. 2004). We considered all stations independently; as a consequence, only the scope and the magnitude were used. The maximum value obtained with SQI was 100, and the sediment could be considered nontoxic with values over 80 (Marvin et al. 2004).

Predictive efficiencies and statistical analyses

The predictive efficiencies of the different quotient methods were evaluated with the approach already developed by Shine and collaborators (Shine et al. 2003; Vidal and Bay 2005), where Sensitivity = B/(A + B), Specificity = C/(C + D), Toxicity predictive value = B/(B + D), No toxicity predictive values = C/(A + C), Overall efficiency = (B + C)/(A + B + C + D), type I error (false positives) = (D/[D + B]) × 100, and type II error (false negatives) = (A/[A + C]) × 100.

In these equations, A represents the number of samples with significant biological toxicity and a Qmean <1 (or SQI between 80 and 100), whereas B represents the number of significantly toxic samples with a Qmean >1 (Figure 2; or SQI <80). C and D represent samples with no significant toxicity and below and over the threshold, respectively (Qmean = 1, SQI = 80; Figure 2). For a toxicity test result to be considered significantly toxic, the mortality of H. azteca and C. riparius had to be ≥20% above a mortality level that was considered natural variability in the controls. Toxicity tests with B. calyciflorus had a threshold of ≥40% inhibition of reproduction. These toxicity test thresholds were significantly different from controls (t test; p < 0.05). Type I error or false positives refer to sediment classified as toxic using SQGs when in fact they are not. On the other hand, type II error or false negatives refer to sediment classified as nontoxic when in fact they are toxic (Adams et al. 2005; Wenning et al. 2005a).

Details are in the caption following the image

Schematic of the toxicity threshold and Qmean categories used to evaluate predictive efficiencies of 4 types of aggregation methods (adapted from Shine et al. 2003; Vidal and Bay 2005).

Univariate regression tree or partition analysis was performed in the JMP IN 5.1 statistical package (SAS Institute). We used this method to determine the hierarchy of factors that could explain the toxicity observed in class 1 samples. A great advantage of this method, which produces partition trees, is that it gives significant variables with their hierarchical importance and includes the relevant thresholds. It is useful to develop management tools, and it can categorize explanatory factors as continuous or categorical.

RESULTS AND DISCUSSION

Toxicity prediction based on chemicals measurements

Sediment samples were assigned to a quality class (1, 2, or 3) according to the contaminant with the highest concentrations relative to OEL and FEL thresholds. This classification was performed using SQGs developed for Quebec sediments and the St. Lawrence River (Environment Canada and MDDEP 2007). Using this method, 10 stations were assigned to class 1; of these 10 stations, 4 were located in Lake Saint-Pierre, 4 in Lake Saint-Louis, 1 downstream from the Island of Montréal, and 1 in Lake Saint-François (Table 4). A further 29 stations were assigned to class 2; 11 were located in Lake Saint-Pierre, 8 in Lake Saint-Louis, 6 in Lake Saint-François, and 3 in the Montréal Harbor area (Table 4). Finally, 20 stations were assigned to class 3; 9 of them were located in the Montréal Harbor area, 8 in Lake Saint-Louis, and 3 in Lake Saint-François (Table 4).

Table 4. Separation of sampling stations into 3 sediment classes according to sediment quality guidelines developed for the St. Lawrence River
Class Stations Area H. azteca C. riparius B. calyciflorus
1 51 LSF × ×
16 LSL × ×
17 LSL × ×
22 LSL × ×
23 LSL × ×
9 PM ×
13 LSP ×
15 LSP ×
33 LSP
34 LSP ×
2 52 LSF × ×
53 LSF ×
56 LSF × ×
57 LSF × ×
58 LSF ×
59 LSF ×
18 LSL × ×
19 LSL ×
20 LSL ×
21 LSL ×
24 LSL
26 LSL ×
41 LSL
43 LSL
46 LSL
2 PM × ×
5 PM × ×
8 PM
11 LSP × ×
12 LSP × ×
14 LSP ×
30 LSP × ×
31 LSP
32 LSP
35 LSP ×
36 LSP
37 LSP
38 LSP ×
39 LSP ×
3 50 LSF ×
54 LSF × ×
55 LSF × ×
25 LSL
40 LSL ×
42 LSL × ×
44 LSL ×
45 LSL
47 LSL
48 LSL
49 LSL ×
1 PM ×
3 PM × ×
4 PM × ×
6 PM × ×
7 PM × ×
10 PM
27 PM × ×
28 PM × × ×
29 PM × × ×
  • Stations are identified that had significant biological toxicity results measured by tests with whole sediment (H. azteca and C. riparius mortality) and with sediment porewater (B. calyciflorus reproduction inhibition). Dashes identify stations where a toxicity test was not conducted. LSF = Lake Saint-François, LSL = Lake Saint-Louis, PM = Montréal Harbor area, LSP = Lake Saint-Pierre

For sediments in class 1, the probability of detecting adverse biological effects should be low, insofar as the sediments are considered safe for disposal in open water (Environment Canada and MDDEP 2007). Nevertheless, we observed significant sediment toxicity at 9 of the 10 stations, depending on toxicity test endpoints. H. azteca and C. riparius mortality was observed at 8 stations, and either both species were affected (4 stations) or only 1 of them (4 stations). B. calyciflorus reproduction was significantly inhibited at 2 stations, and 1 of those was also significantly toxic for H. azteca mortality (Table 4).

In class 2, the probability of detecting adverse biological effects is considered to be higher and is supposed to increase with chemical concentrations in sediment (Environment Canada and MDDEP 2007). In this case, open-water disposal can be considered a valid option only if toxicity tests do not demonstrate adverse effects. In our data set, significant toxicity was observed at 20 of the 29 stations assigned to class 2. Both H. azteca and C. riparius mortality was significant at 7 stations, H. azteca-only mortality occurred at 7 stations, and C. riparius-only mortality occurred at 3 stations. In total, H. azteca and C. riparius mortality represented 48% and 34% of toxicity incidence, respectively. The reproduction of B. calyciflorus was affected at 28% of the stations, and 2 of those were also toxic for H. azteca (Table 4).

Finally, in class 3, the probability of detecting adverse biological effects is expected to be high, and open-water disposal is prohibited (Environment Canada and MDDEP 2007). Toxicity was observed at 15 of the 20 stations assigned to this class. Sediments were demonstrated to be significantly toxic at 9 stations for both mortality tests (H. azteca and C. riparius) and at 5 stations for only 1 of the 2 toxicity tests. The inhibition of B. calyciflorus reproduction was significant at 4 stations; 3 of those stations were also toxic to H. azteca, 2 were also toxic to C. riparius, and only 1 was toxic to B. calyciflorus alone (Table 4).

SQGs, quotients, and indices

Quotients and indices provide a potential alternative for classifying sediments when multiple contaminants and the interactions among them are considered. Ingersoll and collaborators (2005) observed an increase in the incidence of toxicity to H. azteca and C. riparius as a positive function of mean quotient (Qmean). However, the ability of SQGs to predict the occurrence of toxicity had a wide range (Vidal and Bay 2005).

In our study, the 3 Qmeans that were assessed had similar efficiencies for predicting toxicity (Table 5). The first mean quotient (Qmean1) had type II (false negative) errors of 54%, 48%, and 26% for H. azteca mortality, C. riparius mortality, and B. calyciflorus reproductive inhibition, respectively (Table 5). These proportions of type II errors suggest a better performance of B. calyciflorus. In fact, B. calyciflorus's ability to predict toxicity was much lower (0.3; Table 5) than that of C. riparius and H. azteca, with a mortality endpoint of 0.6 (Table 5). This difference is due to the higher proportion of type I (false positive) errors at 71% for B. calyciflorus compared with 40% for mortality tests. The high proportion of type I errors with B. calyciflorus might also be explained by its higher toxicity threshold (40%) compared with mortality tests, which would have reduced the sensitivity of B. calyciflorus to toxicity.

Table 5. Efficiency of 4 types of aggregation methods (based on the OEL threshold) used to predict toxicity of multiple contaminants in St. Lawrence River sediments
A B C D n Sensitivity Specificity Toxicity prediction No-toxicity prediction Overall efficiency Type II error Type I error
Qmean1
H. azteca mortality 19 13 16 19 157 0.41 0.64 0.59 0.46 0.51 54.3 40.9
C. riparius mortality 15 12 16 8 51 0.44 0.67 0.60 0.52 0.55 48.4 40.0
B. calyciflorus reproduction 5 4 14 10 34 0.44 0.58 0.29 0.74 0.55 26.3 71.4
Qmean2
H. azteca mortality 19 13 18 7 57 0.41 0.72 0.65 0.49 0.54 51.4 35.0
C. riparius mortality 16 11 17 7 51 0.41 0.71 0.61 0.52 0.55 48.5 38.9
B. calyciflorus reproduction 5 5 16 8 34 0.50 0.67 0.38 0.76 0.62 23.8 61.5
Qmean3
H. azteca mortality 18 14 18 7 57 0.44 0.72 0.67 0.50 0.56 50.0 33.3
C. riparius mortality 15 12 18 6 51 0.44 0.75 0.67 0.55 0.59 45.5 33.3
B. calyciflorus reproduction 4 6 17 7 34 0.60 0.71 0.46 0.81 0.68 19.0 53.8
SQI
H. azteca mortality 16 16 13 12 57 0.50 0.52 0.57 0.45 0.51 55.2 42.9
C. riparius mortality 13 14 15 9 51 0.52 0.63 0.61 0.54 0.57 46.4 39.1
B. calyciflorus reproduction 4 6 11 13 34 0.60 0.46 0.32 0.73 0.50 26.7 68.4
  • Definitions of variables used to evaluate the predictive abilities of the methods are presented in Materials and Methods.

The Qmean2 and Qmean3 provided results similar to those for Qmean1 (Table 5). Nevertheless, we observed for these mean quotients a slight decrease in type II errors (0–7%) and an increase in the overall efficiency (0–13%; Table 5). With Qmean3, the proportion of misclassified stations was lower, but the proportion of type II errors remained about 50% and 45% for H. azteca and C. riparius mortality, respectively (Table 5). Toxicity test was predicted at a frequency equal to or higher than 60% with these 3 quotient methods (Table 5).

The results obtained with the Qadd (Figure 3) and SQI methods were similar to mean quotient results, even though SQI accounts for scope and magnitude (Table 5). Furthermore, as with the Qmeans, the lack of a relationship between Qadd or SQI and toxicity incidence (Figure 4) prevented the determination of quality classes specific to our study area, as recommended for SQI (Grapentine, Painter, et al. 2002; Marvin et al. 2004).

Details are in the caption following the image

Relationships between the percentage of H. azteca (A) and C. riparius (B) mortality and the inhibition of B. calyciflorus reproduction (C) with the additive quotient (Qadd). Circles, triangles, squares, and lozenges correspond to the Montréal Harbor, Lake Saint-Pierre, Lake Saint-Louis, and Lake Saint-François, respectively.

Details are in the caption following the image

Relationships between the percentage of H. azteca (A) and C. riparius (B) mortality and the inhibition of B. calyciflorus reproduction (C) with the sediment quality index (SQI; 0 = poor quality class, 100 = excellent; Marvin et al. 2004). Circles, triangles, squares, and lozenges correspond to the Montréal Harbor, Lake Saint-Pierre, Lake Saint-Louis, and Lake Saint-François, respectively.

When we compared a classification of sediment based on individual chemicals or on mean quotients, we observed very similar results, and the same samples were misclassified. All stations classified as class 1 had a Qmean1 lower than 0.5. Those of class 2 had a Qmean1 between 0.5 and 1.0 except for 1 stations, 1 of which displayed only one chemical in class 2, the other of which had concentrations very close to class 3 for several chemicals. Quotients for stations in class 3 were above 1.5. Many of the samples assigned to class 1 (no impact) were toxic, whereas some nontoxic samples were found in class 2 and class 3. During this study, we focused our attention on type II errors (false negatives). The reduction of this type of error is obviously important in terms of environmental protection, insofar as hazardous materials would be erroneously disposed of in open water. Conversely, type I errors (false positives) do not entail any environmental risk: class 3 sediments, which display high concentrations of contaminants, would require an abandonment of the open-water disposal option.

In summary, 2 main observations were made in the comparison between screening of sediment toxicity by SQGs and organism toxicity tests. First, FEL thresholds predicted well the potential toxicity of highly contaminated sediments (class 3). The type I error observed for class 3 sediment was explained mainly by high concentrations of a single contaminant, Hg, a metal that presents an important risk of bioaccumulation in aquatic organisms rather than direct toxicity to invertebrates. Although some sediment samples assigned to class 3 were not toxic to invertebrates, the related management decision remains correct. Second, we observed that a great proportion of class 2 and 3 sediments were toxic to aquatic invertebrates. The type II errors in class 1 represent a real management problem, and the causes of these misclassifications must be identified. Two main hypotheses were proposed: 1) unmeasured contaminants were present that contributed to sediment toxicity, and 2) environmental factors (e.g., sediment grain size, nutrients, metal-binding phases) had a direct influence on toxicity test organisms by increasing their sensitivity or playing a role in contaminant bioavailability. These 2 hypotheses are evaluated in the next sections.

Unmeasured contaminants

A limitation of toxicity screening with SQGs is that a chemical is accounted for only if an SQG is available (Ingersoll et al. 2005; Long et al. 2006). The predictive efficiency of this approach is therefore dependent on the appropriateness of the list of substances monitored. Note that the method we used to determine efficiency was somewhat different from that used in previous studies (Field et al. 1999) in which predictive efficiency of SQGs and quotients was assessed after screening the data sets to eliminate samples presenting unexplained toxicity. Because our purpose was to assess the guidelines' suitability, such a screening approach seemed inappropriate. Accordingly, the first hypothesis to explain type II errors in this study was the presence of nonmeasured contaminants. By definition, eliminating this source of errors would entail complementary sampling and analyses, which were not possible in the context of the current study. Therefore, in this section we review possible explanations for misclassified samples observed in class 1.

Agricultural activity and pesticides can exert a significant pressure on ecosystems (Hela et al. 2005), and recently a risk-based approach was developed in Canada by Environment Canada for ranking pesticides and their potential risk to aquatic life (Whiteside et al. 2008). Pesticides may cause toxicity to invertebrates (Anderson et al. 2005) and modify community structure or species traits (Anderson et al. 2005; Liess and Von Der Ohe 2005; Liess et al. 2008). In this study, we analyzed 92 pesticides (organochlorine, organophosphate, aryloxyacid, toxaphene; Table 3) in sediment at stations 1 to 26. These substances were below quantification limits in 99% of the measurements. Consequently, these analyses were not performed on samples 27 to 59. Most of the class 1 false-negative samples belonged to the stations of the first cluster (i.e., stations 1–26), so pesticides are not the most probable cause of this unexplained toxicity. However, pesticides were not measured in sediment from a misclassified station (51) in Lake St. François (Table 4), which is located near the mouths of the Salmon and St. Regis Rivers (Figure 1). Near this station, some pesticides have been identified in the water column and sediments in 1989 (Fortin et al. 1994) and in 1992–1993 (Rondeau 1996). Three other class 1 stations with significant toxicity were located in Lake Saint-Pierre (13, 15, and 34). These stations were affected by the Maskinongé River and the Bayonne River, which drain watersheds affected by municipalities and extensive agriculture or intensive farming in their downstream sections (Robitaille 1997, 2005; Giroux 2007; Pelletier et al. 2008). No pesticides were identified in sediments from sites 13 and 15. However, agricultural activities, such as livestock farming, can release into the aquatic environment nutrients, veterinary drugs (including biocides), and other substances that were not analyzed. Furthermore, we observed a significant increase in B. calyciflorus reproduction (110%) at station 34. Different hypotheses could explain this result, such as the presence of pharmaceutical compounds (e.g., antibiotics or antihypertensive drugs) that can affect B. calyciflorus reproduction (Ferrari et al. 2004) or the presence of human or livestock hormonal substances. Alternatively, the stimulatory effect could be a laboratory artefact, because rotifer reproduction can be enhanced by environmental factors such as food availability during the toxicity test (Snell and Boyer 1988; Pavon-Meza et al. 2005).

Municipal wastewater represents another source of contaminants in the St. Lawrence River. Sewage could be a source of pharmaceuticals (Kolpin et al. 2004; Gros et al. 2007), some of them being rather persistent (Bendz et al. 2005). Their toxicity is currently not very well known in most cases (Fent et al. 2006; Gros et al. 2007). However, some data on toxicity effects of some pharmaceuticals and personal care products on C. tentans and H. azteca are available, and mean lethal concentration varied between 0.4 and 47.3 mg/L (Dussault et al. 2008). Concentrations of similar products are lower (<0.5 mg/L) in the surface water of the St. Lawrence River and in Montréal wastewater (Garcia-Ac et al. 2009). Their accumulation in sediment is still unknown for this ecosystem.

Another misclassified station (9) was located downstream of the Montréal area (Figure 1, Table 4). This station is influenced by highly turbid waters originating from the Des Mille Îles River, the Des Prairies River, and the Assomption River. The Des Mille Îles and Des Prairies Rivers collect urban discharges, and the Assomption River watershed is dominated by intensive agricultural activities. Therefore, we can hypothesize that an array of contaminants from these activities was present in sediments but that these were not measured (Rondeau 1996).

Four misclassified stations in class 1 were in Lake Saint-Louis (Table 4), three of which were located along the north shore (16, 17 and 22) and the last (23) located on the south shore (Figure 1). These stations are in a mixing zone for waters from the Ottawa and the St. Lawrence Rivers, which are generally considered to be only weakly contaminated. However, in this area, high concentrations of organotin contaminants, such as tributyltin (TBT), were observed by Pelletier (2008). Because organotins have been historically considered as a problem in marine environments (Cardwell et al. 1999), their impacts on freshwater ecosystems should not be overlooked. Organotins are persistent in sediments, and benthic invertebrates may be directly exposed to them as they ingest and burrow into sediments (Bartlett et al. 2004). These authors found the TBT chronic toxicity test on H. azteca survival seemed to be the most sensitive endpoint.

Station 22 may be affected by groundwater originating from the City of Montréal area, which is potentially contaminated from industrial activities or by surface water from small contaminated creeks on the north shore of Montréal Island (Deschamps et al. 2005). On the south shore, station 23 may be influenced by the Saint-Louis River plume; however, the chemical concentrations in sediment were lower than those observed at stations (45–49) located directly in the plume, and the level of toxicity at station 23 was rather weak, with 20% and 23% mortality for C. riparius and H. azteca, respectively. The difference from a “nontoxic” sample (response under 20%) was thus small.

In summary, reducing the uncertainty associated with nonmeasured contaminants would entail additional sampling and analyses, which were not possible in the context of the current study. Pesticides or their metabolites cannot be ruled out, although they are not the most probable factor. Industrial chemicals might explain type II errors at a few stations. Most concerns are related to contaminants carried by municipal wastewater or agricultural activities: pharmaceuticals, personal care products, or biocides. Organotins could also explain type II errors at some stations.

Importance of confounding factors and approaches for dealing with them

Another possible origin of type II errors may be related to physical and chemical characteristics of sediments (Word, Albrecht, et al. 2005; Word, Gardiner, et al. 2005), which may introduce uncertainty into the application of SQGs (Chapman et al. 2005). The responses of organisms used in laboratory toxicity tests may be influenced by characteristics such as grain size, nutrients, organic matter, or other contaminant-binding phases (Word, Albrecht, et al. 2005). Compared with class 2 and class 3 stations, misclassified class 1 stations were associated with lower concentrations of Al, Fe, Mn, sulfur, total organic carbon (TOC), and clay, all of which bind contaminants. The reduced concentrations of these phases could contribute to increased contaminant bioavailability. Regression tree analyses were performed with all possible explanatory variables in the data set. This multiple regression method identified which variables best explained toxicity. The main advantage of the regression tree method is that it classifies predictive variables according to their weight in the statistical model and establishes effect thresholds. We focused the analyses on type II errors in class 1. The numbers of B. calyciflorus reproduction test results for this class were too low (Table 4), so regression trees were performed only for H. azteca and C. riparius data.

Hyallela azteca mortality data were significantly partitioned only by total sulfur (TS) concentration. When TS concentrations were above 1400 mg/kg (r2 = 0.64; Figure 5A), all samples induced toxicity to H. azteca, whereas only 1 was toxic below this threshold. Sulphates and sulphides strongly bind metals and can reduce metal availability (Word, Gardiner, et al. 2005). However, sulphides are also recognized as highly toxic to invertebrates (Knezovich et al. 1996; Wang and Chapman 1999).

Details are in the caption following the image

Regression partition model explaining toxicity test responses of H. azteca (A) and C. riparius (B) for sediment assigned to class 1 only. The numbers of samples with toxicity test/total number of samples are presented in parentheses. Numbers indicate the sampling stations.

For C. riparius mortality, the regression partition model offered a weak explanation of the overall variance (r2 = 0.41; Figure 6B). Nonetheless, a greater proportion of mortality (4/5) occurred when the porewater dissolved organic carbon (DOC) concentration was lower than 6.5 mg/L. Organic matter is known to influence the bioavailability of chemicals (Bervoets et al. 1997; Watzin et al. 1997; Di Toro et al. 2005).

Details are in the caption following the image

Tier 1 decision tree for ERA of dredging sediments.

It is also possible to transfer the model developed for H. azteca to C. riparius with sulfur as the first partitioning factor. Four of 5 samples were toxic to C. riparius if TS was >1400 mg/kg, and only 2 of 5 were toxic when TS was <1400 mg/kg. Without further verification, the COD-based model developed for C. riparius is difficult to transpose to H. azteca: the same numbers of samples (3/5) are toxic on both sides of the COD threshold (6.5 mg/L).

The assignment of stations to class 1, 2, or 3 sediment was not significantly modified when a TS threshold of 1400 mg/kg was added to any of the quotient approaches. In class 1 sediment, quotients remained low (e.g., Qmean1<0.5). Likewise, when the class assignment was based on single chemicals and TS, the class being determined by the most penalizing one or TS, false negatives remained low at 10% and 20% for H. azteca and C. riparius, respectively. Five stations (16, 17, 22, 23, and 51) were reclassified into class 2, where, according to the proposed framework, a toxicity assessment would be performed. Three other stations (9, 13, and 15) were still erroneously assigned to class 1, but, as explained above, this seems likely to be due to unmeasured contaminants.

CONCLUSIONS: IMPLICATIONS FOR ERA GUIDELINE DEVELOPMENT FOR DREDGED SEDIMENT IN THE ST. LAWRENCE RIVER

Whatever the approach used, whether a classification was based on the chemical with the highest concentration relative to its effect levels or on mean quotients, we observed very similar results in terms of predictive ability, with a high proportion of type II errors in class 1 sediments or in sediments with quotients <1. From the perspective of sustainable management of dredged materials in the St. Lawrence River, the class 1 sediments that were toxic are the greatest concern. Inhibitory effects associated with compounds usually not addressed in priority substances lists, such as sulfur, were shown to correlate with a significant proportion of toxicity.

The apparently poor efficiency of St. Lawrence SQGs for predicting toxicity in class 1 materials could be an argument for modifying the assessment process design at tier 1 and for introducing toxicity tests at the first screening stage. We have demonstrated, based on regression tree analysis, that high sulfur concentrations were significantly associated with unexpected toxicity in this class; a total sulfur threshold allows the sorting of probably hazardous samples from samples that are uncontaminated in terms of priority compounds. Consequently, we argue that SQGs should be kept as the principal element for the first tier in the risk assessment process, but the initial screening should include a total sulfur measurement. If sulfur exceeds 1400 mg/kg, the sample would probably be toxic, and then the assessment would continue to tier 2, in which toxicity tests would be performed (Figure 6). If sediments were still assigned to class 1 after sulfur verification, they could be disposed of in open water or used for other purposes (Environment Canada and MDDEP 2007).

For class 2 sediments, with contaminant concentrations between OEL and FEL thresholds, a simple classification based on the most penalizing chemical concentration efficiently predicted toxicity in 69% of the stations (Table 4). Consequently, the probability of detecting adverse biological effects is relatively high in this sediment class. In this case, open-water disposal can be considered a valid option only if toxicity tests demonstrate that sediments will not adversely affect the receiving environment (Environment Canada and MDDEP 2007). Class 2 sediments pass to tier 2, in which a battery of toxicity tests is applied, as is recommended in the SQGs management framework (Figure 6).

Our study shows that class 3 sediments, those with contaminant concentrations exceeding FEL thresholds, are either toxic or contain a high concentration of Hg, a chemical that presents an important risk of bioaccumulation in aquatic organisms but is not readily toxic to benthic invertebrates. These observations are consistent with the very high probability of observing adverse biological effects expected for class 3 sediment, and, consequently, within the management framework proposed for the St. Lawrence River (Environment Canada and MDDEP 2007), open-water disposal of class 3 dredged material is prohibited (Figure 6). In this case, sediments must be treated or safely contained. The risk of either of those two options will be evaluated based on a specific ERA guideline for contaminated soil and management under terrestrial conditions (CEAEQ 1998; Beaulieu 1999).

Screening using SQGs and statistical analyses may not be accurate indicators of which chemicals in the sediments are the cause of toxicity. Additional analyses, such as toxicity identification evaluations and laboratory toxicity tests of clean sediments spiked with known chemicals, may be needed to determine causality accurately (Wenning et al. 2005b; Long et al. 2006). Furthermore, the effect of mixture is not completely considered in the quotient approach, a situation that is exacerbated by the lack of knowledge about interactions between chemicals or influences of geochemical properties (Batley et al. 2005).

Future steps involve the development of several tools to validate the conclusions of the present study. First, a comparison of the toxicity test results with the benthic communities living in these sediments will be performed. Second, it is necessary to characterize the presence of emergent contaminants in sediment, to determine their toxicity, and to evaluate the need to establish new SQGs for these substances. Finally, the main conclusions of this study should be validated by a case study assessment of future dredging activities in the St. Lawrence River.

Acknowledgements

This study is a part of a larger collaborative program funded by the third and fourth phases of the St. Lawrence Plan for Sustainable Development with the active participation of Environment Canada (Environmental Protection Operations Division and Science and Technology Branch), the Ministère du Développement durable, de l'Environnement et des Parcs du Québec (Centre d'expertise en analyse environnementale du Québec; Direction des évaluations environnementales; Direction du suivi de l'état de l'environnement), and Cemagref from Lyon (France). The project was also associated with the Sustainable Navigation Strategy for the St. Lawrence River, which includes aspects such as sustainable dredging management, contaminated sites restoration, and sediment quality guidelines revision for contaminated sediment. We acknowledge the Commission Permanente de Cooperation Franco-Québecoise for travel funding during this collaborative project. We address special thanks to the project steering committee members: C. Gagnon from Environment Canada; L. Boudreau, P. Michon, I. Guay, and G. Triffaut-Bouchet from the Ministère du Développement durable, de l'Environnement et des Parcs du Québec; and S. Masson from Parc Aquarium du Québec. We also acknowledge all fieldwork participants, particularly M. Arseneault, P. Turcotte, A. Lajeunesse, and G. Brault, who helped over the 2 sampling years. We also thank Heather Ferguson and John Chételat for English editing.