Journal of Soils and Sediments

, Volume 11, Issue 3, pp 504–517

Development of sediment quality guidelines for freshwater ecosystems

Authors

    • Institute of Environment & Sustainable DevelopmentUniversity of Antwerp
  • Ward De Cooman
    • Flemish Environment Agency
  • Vicky Leloup
    • Ecosystem Management Research GroupUniversity of Antwerp
  • Patrick Meire
    • Ecosystem Management Research GroupUniversity of Antwerp
  • Claudia Schmitt
    • Ecosystem Management Research GroupUniversity of Antwerp
  • Peter C. von der Ohe
    • UFZ, Department of Effect-Directed AnalysisHelmholtz Centre for Environmental Research, UFZ
SEDIMENTS, SEC 1 • SEDIMENT QUALITY AND IMPACT ASSESSMENT • RESEARCH ARTICLE

DOI: 10.1007/s11368-010-0328-x

Cite this article as:
de Deckere, E., De Cooman, W., Leloup, V. et al. J Soils Sediments (2011) 11: 504. doi:10.1007/s11368-010-0328-x
  • 448 Views

Abstract

Purpose

The development of Sediment Quality Guidelines (SQGs) is one of the remaining challenges for a better protection of aquatic biodiversity and in particular sediment dwelling organisms. So far, sediment quality assessment in Flanders was based on a comparison of chemical concentrations to the geometric mean of the concentrations at 12 reference sites. The study described in this paper addressed the need for more science-based guidelines. The developed guidelines are already incorporated into Flemish legislation.

Materials and methods

Based on a large sediment monitoring database, containing physico-chemical properties, concentrations of chemicals, macrobenthic community assemblages and ecotoxicological data, Sediment Effect Concentrations (SECs) were calculated as basis for the SQGs. The derived SECs were based on ecological effects, namely Lowest and Severe Effect Levels (LEL/SEL), as well as ecotoxicological endpoints, namely Threshold and Probable Effect Levels (TEL/PEL). The average values of the ecological and ecotoxicological SECs were used to distinguish five sediment quality classes.

Results and discussion

The ecological values were in general less stringent than the ecotoxicological values. However, the Lowest Effect Levels (95% of the benthic taxa can be present under this level) and Threshold Effect Levels (no toxic effect is expected under this level) did not differ significantly. Probable Effect Levels (concentrations above this level will certainly result in toxic effects) were generally lower than the Severe Effect Levels (above this level only 5% or less of the taxa are present). The SECs calculated in this study enabled us to correctly identify 87.9% of the sediments as toxic. The development of SQGs based on a combination of the LEL/SEL and TEL/PEL methods enabled us to underpin these SQGs based on field observations and will improve the assessment of sediment quality based on chemical parameters. Although sediments typically contain complex mixtures of contaminants, only a limited number of these contaminants will be measured. Additional application of bioassays for the overall sediment quality assessment is therefore recommended.

Conclusions

This study describes the development of SQGs in Flanders, which are based on ecological and ecotoxicological data derived from a TRIAD monitoring network. The combination of the LEL and PEL resulted in SQGs that were recently incorporated in Flemish legislation and for which the respective pore water concentrations were in the same order of magnitude as the Annual Average Environmental Quality Standards values for Water Framework Directive priority pollutants.

Keywords

EQSMacrobenthosPriority substancesSECSediment quality guidelinesSQGTRIAD

1 Introduction

In the past, sediment quality was often negatively influenced by contaminated surface waters. Nowadays, surface water quality has improved due to enhanced sewage treatment, source control and other measures and it is expected that sediments will now act as a source of contamination for macroinvertebrates and organisms of higher trophic levels (Salomons and Brils 2004; Teuchies et al. 2011). Consequently, achieving a “good ecological status” of surface waters, as required by the European Water Framework Directive (WFD, 2000/60/EC), may strongly depend on the management of sediments. An improved understanding of the relative impact of polluted sediment is necessary to identify and to manage areas where chemical pollution severely affects the ecological status (Fairey et al. 2001) and appropriate guidelines should be developed (den Besten et al. 2003; Chapman et al. 2005; Crane and Babut 2007).

The development of guidelines can be done based on Sediment Effect Concentrations (SECs). SECs are concentration levels which, using certain methods and assumptions, can be related to observed ecological and/or ecotoxicological effects (Shine et al. 2003). SECs based on ecotoxicological information are often based on single species toxicity tests measuring the effect of a specific toxicant by spiking artificial sediments not taking into account other natural stressors (e.g. low oxygen levels, competition). Unfortunately, standardised test organisms for toxicity tests are mainly organisms living in the water phase and the direct exposure route of contaminants from the sediment to the organisms is thereby neglected. In this case, equilibrium partitioning models are applied to derive sediment quality guidelines (SQGs) for sediments (Chapman et al. 1998; Crommentuijn et al. 2000; European Commission 2003). Besides this, the bioavailability of contaminants in spiked sediments can differ significantly from natural sediments, both due to the spiking procedure and ageing effects, as well as due to sediment characteristics. Moreover, toxicity tests are usually applied to test organisms that can be easily cultured and are not applied to rare or very sensitive species.

SECs can also be based on the co-occurrence of biota under certain conditions in the field. Field data give a more realistic view, but are much more difficult to interpret because these data—chemical, physical and biological—are a result of natural fluctuations and multiple stressors. Collecting field data is more time consuming and the data have a greater uncertainty (Connell et al. 1999). Ideally, the derivation of SECs should be based on a combination of both ecological (in situ communities) and ecotoxicological data (Swartz 1999; Environment Canada 2003; Ingersoll et al. 2000; MacDonald et al. 2000; Engler et al. 2005; Vidal and Bay 2005), but integrated monitoring datasets are scarce. Therefore, most studies concerning SQGs have used only either toxicity or field data.

Since the first SQGs were introduced in the 1970s (Engler et al. 2005), a number of tools have been developed for the assessment of sediment quality. Historically, sediment assessment was carried out by comparing the concentrations of contaminants with a reference value. However, chemical concentrations in sediments alone do not allow predicting the expected effects on the organisms, as the bioavailability of the contaminants could differ widely among different sediments. Derivation of effects-based SQGs was not possible until toxicity tests for sediments were developed and applied. In the late 1980s, the screening level-concentration approach was applied to derive values that were protective for 90% of benthic species. This approach is based on co-occurrence data of chemical concentrations and benthic communities at a certain spot. However these data were not obtained at exactly the same time and/or place. In 1985, the Sediment Quality TRIAD concept was developed (Long and Chapman 1985), being one of the first tools that used the same sediment sample for both chemical and biological analysis. This integrated assessment approach can be used to derive SECs as a basis for SQGs.

According to Article 16 of the WFD (2000/60/EC), environmental quality standards (EQS) should be worked out for certain groups or classes of pollutants found in water, sediment or biota. These standards should be implemented in the legislation of the Member States. The methodology to develop these guidelines is described in a Technical Guidance Document for Deriving Environmental Quality Standards under the Water Framework Directive. However, in Flanders, the northern part of Belgium, new SQGs have been already developed and were implemented into legislation in July 2010. These guidelines replaced the old guidelines which were based on the geometric mean of 12 reference sites (Babut et al. 2005). The guidelines are described in the Flemish regulations, considering quality standards for surface waters, groundwater and sediments, as target values that should be reached or maintained, but they are not considered as stringent remediation criteria.

The guidelines in Flanders are based on the evaluation of a huge monitoring dataset. The sediment monitoring programme is a fairly unique monitoring programme where chemical, biological and ecotoxicological data are generated from the same volume of sediment. More than one hundred chemical parameters are analysed, the benthic community is inventoried, and both the pore water and the sediment are tested with bioassays. The dataset is the result of a study of six years (1993–1999), during which a TRIAD assessment method (physico-chemical, ecotoxicological and ecological evaluation) was developed (de Deckere et al. 2000), and a regular monitoring programme, which started in 2000 applying the previously developed TRIAD method on a network of 600 sites. In this study, these data have been used to calculate different types of SECs. Lowest Effect Level (LEL) and Severe Effect Level (SEL) values were calculated as ecological SECs (Persaud et al. 1992). Threshold Effect Level (TEL) and Probable Effect Level (PEL) values were calculated as ecotoxicological SECs (MacDonald et al. 1992, 2003; Smith et al. 1996). These SECs were derived for single compounds for individual heavy metals, polycyclic aromatic hydrocarbons (PAHs) and polychlorinated biphenyls (PCBs). For PAHs and PCBs, the sum parameters were also derived, because these are more often used in other countries (Swartz 1999; Health Council of the Netherlands 2002). However, not all PAHs and PCBs have the same toxicity, and for sediment management it is interesting to have information on which specific pollutants play an important role in affecting the aquatic ecosystem. At the end, the proposed SQGs were compared to the annual average environmental quality standards (AA-EQS) for priority substances as worked out for the WFD, using an equilibrium partitioning model.

2 Material and methods

2.1 Sampling procedure and database

At all sites, 40 L of sediment was collected using a Van Veen grab sampler. With a few exceptions, all samples were taken in spring (March–June). Subsequently, the physico-chemical, biological and ecotoxicological characteristics of the sediments were determined. For the chemical analyses, the concentrations of 120 pollutants, including heavy metals, PCBs, organochloropesticides (OCPs), PAHs and mineral oil were measured. Associated with the chemical analyses, grain size and organic matter content were measured. The biological assessment is based on the composition of the macrozoobenthic community. The macroinvertebrates were determined to the lowest taxonomic level required for the calculation of the Biotic Sediment Index (De Pauw and Heylen 2001), which was mostly genus or family level. In addition to the biological and chemical evaluation, a sediment contact test (10 day mortality test with Hyalella azteca, ASTM 2007) was conducted. More details on sampling and analysing methods can be found in de Deckere et al. (2000).

The database used in this study mainly consisted of data on freshwater sites. A limited number of data were gathered at brackish sites. All sites belong to the river basins of the Scheldt, Meuse or Yser, and are situated in Flanders, the northern part of Belgium. The sampling sites were mainly located in low gradient streams and highly modified systems. The data were collected between 1995 and 2005 as part of the sediment monitoring network of the Flemish Environment Agency. For sites that were sampled more than once, only the most recent results were included in the analyses. This means that the data collected in 2000 and 2001 were not used for the derivation of the SECs as these were sampled again in 2004 and 2005. It resulted in 1027 unique sites of which, according to the TRIAD evaluation method of Flanders (de Deckere et al. 2000), 5% were rather unpolluted, 20% were slightly polluted, 35% were polluted and 40% heavily polluted.

2.2 Derivation of ecological and ecotoxicological SECs

The ecological SECs calculated in this study are the LEL and the SEL (Neff et al. 1986). The first step in the LEL/SEL calculation, see also Fig. 1, was to count the number of sampling sites where each taxon (genus or family see section 2.1) was found, but only taxa found at a minimum of five sampling sites were used in the further calculations. For those taxa, the 85th, 90th, and 95th percentile values of all concentrations measured for a certain pollutant at the sites where the taxa was present, were calculated. The values increased slightly with increasing percentile. However, the 90th percentile was proposed by Neff et al. (1986) and also used by MacDonald et al. (1996, 2003) to provide a more conservative estimate, supposing that extreme high concentrations in the sediment may be an aspect of specific sediment characteristics resulting in low biological availability. Therefore it was decided to use the LEL/SEL values based on the 90th percentile. The concentrations of some of the toxicants, especially the OCPs and PCBs, were often below the limit of detection (LoD). The LoD for PCBs and OCPs changed from 1 to 0.05 μg kg−1 DW in 2000 as a result of new analytical equipment. For the calculation of SECs it is necessary to estimate a value for measurements below the LoD. In cases where the percentage of LoD of a particular OCP or PCB was lower than 30%, half of the value of the LoD was used. When 30–80% of the data was below the LoD, values were obtained by using the Maximum Likelihood Estimation method (Helsel 2004). If more than 80% of a certain pollutant was reported below the LoD, no values could be estimated. The 90th percentile value was calculated based on each individual taxon for (1) each compound and (2) for the sum of ten PCBs (PCB101, PCB118, PCB138, PCB153, PCB170, PCB180, PCB28, PCB31, PCB49, PCB52) and the six PAHs of Borneff (fluoranthene, benzo(k)fluoranthene, benzo(a)pyrene, benzo(g,h,i)perylene, indeno(1,2,3-cd)pyrene, benzo(b)fluoranthene). Subsequently, the 90th percentile values of all taxa were ranked separately for each pollutant and the 5th percentile (LEL) and 95th percentile (SEL) values of this distribution were calculated, respectively.
https://static-content.springer.com/image/art%3A10.1007%2Fs11368-010-0328-x/MediaObjects/11368_2010_328_Fig1_HTML.gif
Fig. 1

Flow chart of the calculation steps for the lowest and severe effect levels

For the ecotoxicological SECs, the TEL and PEL were calculated (MacDonald et al. 1992, 1996, 2003; Smith et al. 1996), based on the results of the solid phase test with H. azteca (Vangheluwe et al. 2000). The first step of the TEL/PEL calculation was to divide the toxicity data of a certain compound into an “effect” and a “no effect” group, taking a mortality of 20% compared to the reference sediments as the limit between “effect” and “no effect”. Next, the average concentration of the compound was calculated for both the “effect” and “no effect” group. If the average value of the “no effect” group was higher than the average value of the “effect” group, TEL and PEL could not be calculated because there was no causal relation between the toxicant and the measured toxicity. If the average of the “effect” group was higher than the average of the “no effect” group, the data in the “effect” group with concentrations lower than the average value of the “no effect” group was not taken into account in the further calculations. The data of the “no effect” group with concentrations higher than the average of the “effect” group was also not taken into account. Secondly, the 15th and median percentile values of the concentrations of the compound in the “effect” group were calculated. For the “no effect” group, the median and 85th percentile were calculated. Finally, TEL was calculated as the geometric mean of the 15th percentile of the “effect” group and the median percentile of the “no effect” group. PEL was calculated as the geometric mean of the median percentile of the “effect” group and the 85th percentile of the “no effect” group.

2.3 Consensus-based SECs

The calculated SECs are used for underpinning a new classification of sediment quality in Flanders based on the so called “consensus values” and for the incorporation of SQGs into Flemish legislation. The consensus 1 values were calculated as the average of LEL and TEL and can be described as a long term objective or good ecological sediment status. The consensus 2 values were calculated as average of SEL and PEL and can be described as values above which toxic and in situ effects are most likely to be observed. If TEL and PEL could not be calculated, LEL and SEL were suggested as consensus values.

2.4 Predictability and sensitivity of the consensus values

The predictability of the consensus values was investigated using the monitoring data of 2000–2001 (299 sites), which were not included in the calculations of LEL, TEL, SEL, and PEL. The measured concentration of a compound in a sediment sample was compared to the consensus value 2. If a compound was present in a concentration higher than consensus value 2, a toxic effect is predicted. As second step, the prediction is compared to the H. azteca bioassay result of the same sample. If the mortality to H. azteca was >20% higher than the mortality in the reference sediment, the sample was considered toxic, and the prediction was considered as confirmed.

The predictive ability of the consensus values was finally calculated as the ratio of the number of samples that were correctly predicted to be toxic (ncpred) compared to the total number of samples that were predicted to be toxic (npred; MacDonald et al. 2000), according to Eq. 1:
$$ Predictability = {n_{{cpred}}}/{n_{{pred}}} $$
(1)
On the contrary, sensitivity is the ability to identify toxic samples based on the consensus values. It was calculated as the ratio of the number of samples that were correctly predicted to be toxic based on a specific compound or group of compounds (ncpred) compared to the total number of samples where the mortality of H. azteca was >20% (ntox = 282), according to Eq. 2:
$$ Sensitivity = {n_{{cpred}}}/{n_{{tox}}} $$
(2)

2.5 International guidelines

The consensus 1 values calculated in this study were compared to the Target Values of the Netherlands (Babut et al. 2003), the LEL and TEL values calculated for Florida inland waters (MacDonald et al. 2003), the limit class A values for the Italian Venice Lagoon (Apitz et al. 2007) and the Class II values for the German part of the Elbe (Heise et al. 2005). All these SQGs can be considered as “no effect” SQGs, based upon contaminant levels below which toxic effects are generally not observed. The consensus 2 values were compared to the SEL and PEL values for Florida (MacDonald et al. 2003), the limit class B values for the Venice Lagoon (Apitz et al. 2007) and the Quality Objective (QO) values for the Netherlands (Babut et al. 2003). These guidelines can be considered as “probable effect” SQGs, based upon contaminant levels above which toxic effects are generally observed.

Furthermore the consensus values 1, which were used to make up the final Flemish SQGs, have been transformed into concentrations for the water phase based on the partition coefficients equation (3) introduced by Di Toro et al. (1991):
$$ {K_{{oc}}} = {c_s}/{c_w} \times {f_{{oc}}} $$
(3)
where cw is the concentration in the water, cs is the concentration in the sediment, foc is the fraction of organic carbon in the sediment and Koc is the partition coefficient between water and sediment based on organic carbon. As consensus, the 5% value for organic carbon content was used in this study as this parameter was not analysed. The calculated values for the water phase have been compared with the legally binding AA-EQS for priority substances (EU Directive 2008/105/EC), as well as to the PNEC values that have been derived for the updated prioritisation process of priority substances (James et al. 2009).

2.6 Statistical analyses

All statistical analyses were done with the programme SAS 9.1. (SAS Institute Inc., Cary, NC, USA). The difference between the ecological and ecotoxicological values was statistically tested using Wilcoxon signed rank test. Only the individual compounds for which both the ecological and ecotoxicological values could be calculated were included in the statistical analysis and a difference was made between the main groups of pollutants (heavy metals, PAHs and PCBs). OCPs were not included because of the high number of data below the detection limit and the low number of OCP compounds included in this study.

3 Results

3.1 Description of the database

A summary of the value range of chemical concentrations and grain size characteristics at the sampling sites can be found in Table 1. The 10th percentile values of the measured concentrations for heavy metals varied between 0.03 mg kg−1 DW for mercury (Hg) and 47 mg kg−1 DW for zinc (Zn). The 50th and 90th percentile values were also highest for Zn (137 mg kg−1 DW and 609 mg kg−1 DW). Concerning PAHs, fluoranthene had the highest 50th percentile value (0.22 mg kg−1 DW) and pyrene had the highest 90th percentile value (1.2 mg kg−1 DW). For all PCBs, the 10th percentile values were the detection limit of 0.05 μg kg−1 DW. Highest PCB concentrations were measured for PCB 153 (50th and 90th percentile values 1.2 μg kg−1 DW and 13 μg kg−1 DW, respectively). At 98% of the sample sites, measured concentrations of PCB 169 were situated below the detection limit. Therefore, no SECs could be calculated for this compound.
Table 1

Physico-chemical properties of the sampling sites—10th percentile, 50th percentile and 90th percentile of the measured concentrations

Parameter

10th percentile

50th percentile

90th percentile

Fraction grain size x < 2 μm

2

7

21

Fraction grain size 2 μm < x < 63 μm

3

18

53

Fraction grain size 63 μm < x

28

75

94

Fraction total organic carbon

4

16

47

Asa

2.3

6.5

21

Cda

0.2

0.63

3.6

Cra

8

25

77

Cua

3.8

17

82

Hga

0.03

0.15

0.97

Nia

4.7

13

31

Pba

10

27

129

Sea

0.48

1.4

5.4

Sna

0.15

1.5

10

Zna

47

137

609

Acenaphthenea

0.0004

0.05

0.5

Acenapthylenea

0.0004

0.0004

0.5

Anthracenea

0.002

0.02

0.2

Benz(a)anthracenea

0.009

0.09

0.67

Benzo(a)pyrenea

0.01

0.1

0.64

Benzo(b)fluoranthenea

0.01

0.12

0.74

Benzo(e)pyrenea

0.01

0.14

1

Benzo(g,h,i)perylenea

0.01

0.09

0.52

Benzo(k)fluoranthenea

0.007

0.05

0.35

Chrysenea

0.01

0.12

0.78

Dibenz(a,h)anthracenea

0.002

0.02

0.1

Phenanthrenea

0.01

0.11

0.89

Fluoranthenea

0.02

0.22

1.5

Fluorenea

0.002

0.02

0.23

Indeno(1,2,3-cd)pyrenea

0.01

0.08

0.54

Naphthalenea

0.0007

0.05

0.64

Perylenea

0.007

0.04

0.2

Pyrenea

0.01

0.16

1.2

Extractable organohalogenesc

0.2

6

33

Non polar hydrocarbonsa

39

147

783

PCB 101b

0.05

0.67

8.9

PCB 118b

0.05

0.4

5.9

PCB 138b

0.05

0.9

10

PCB 153b

0.05

1.2

13

PCB 169b

0.05

0.05

0.05

PCB 170b

0.05

0.05

3.9

PCB 180b

0.05

0.7

8.9

PCB 28b

0.05

0.05

4

PCB 31b

0.05

0.05

3.4

PCB 49b

0.05

0.05

4.4

PCB 52b

0.05

0.05

6.7

4,4′-DDDb

0.05

0.05

3.5

4,4′-DDEb

0.05

0.35

6.4

Hexachlorobenzeneb

0.00005

0.05

0.05

aValues in mg kg−1 DW

bValues in μg kg−1 DW

cValues in mg Cl kg−1 DW

The benthic community found on the 1027 sites consisted of 127 different taxa. Twenty-one taxa were only found at one site. The most commonly present group was chironomidae non thummi plumosus, which was found at 820 sites. When excluding the taxa found at less than five sites, 51 of the 127 taxa were omitted. So, in total, 76 taxa were used for the calculations of LEL and SEL. On one hand, there is a relative high number of sensitive species according to the BBI index (De Pauw and Vanhooren 1983) and the SPEAR index (Liess and Von der Ohe 2005; Von der Ohe et al. 2007) among the excluded taxa (Fig. 2). On the other hand, the excluded taxa found at two to four sites were distributed over locations with one up to 28 contaminants above the calculated consensus 1 in a similar way as the taxa found at five sites or more. Furthermore, 32 out of the 51 species belong to taxa that are observed only in very rare cases in or on the sediments, such as Plecoptera, Ephemeroptera and Odonata (De Pauw and Heylen 2001).
https://static-content.springer.com/image/art%3A10.1007%2Fs11368-010-0328-x/MediaObjects/11368_2010_328_Fig2_HTML.gif
Fig. 2

Relative number of taxa that are found at one to four sites or at more than four sites. The taxa are divided into tolerance classes (left) based on the BBI (De Pauw and Vanhooren 1983) with 1 being the most sensitive class and 7 most tolerant class or on sensitivity (right) based on the SPEAR classification (von der Ohe et al. 2007), with 1 being sensitive species and 0 being non-sensitive species

3.2 Ecological and ecotoxicological SECs

Both the LEL and TEL values as well as the SEL and PEL values were always in the same order of magnitude (Table 2). For arsenic (As), selenium (Se), PCB118 and hexachlorobenzene, the TEL and PEL values could not be calculated because the average concentration of the “effect” group was lower than the average concentration of the “no effect” group. With regard to heavy metals, there are no clear patterns in the LEL and TEL values, sometimes LEL was higher and sometimes TEL was higher (see Table 2). There was no significant difference between LEL and TEL (p = 0.38). SEL was always higher than PEL and both differed significantly (p < 0.01). The values of Zn, especially for the SEL, were very high compared to the other heavy metals. When Zn was excluded, SEL and PEL of the heavy metals did not show significant differences (p = 0.16). Similar to the situation for the metals, LEL and TEL did not show significant differences for PAHs (p = 0.54) while SEL and PEL differed significantly (p < 0.001; see Table 2). Again, SEL was always higher than PEL. For PCBs and OCPs, with the exception of PCB 180 and PCB 101, TEL was always higher than LEL (see Table 2). For PCB 28, PCB 31, PCB 49, PCB 52 and 4,4′-DDD, the LEL values resulted in values below the detection limit. For PCBs and OCPs, LEL and TEL did not differ significantly (p = 1). SEL was again always higher than PEL, and both differed significantly (p = 0.01). For both PCBs and OCPs, the effect levels calculated based on the sum of the concentrations were lower than the summed values of the individual compounds included in the sum parameter.
Table 2

LEL, TEL, consensus 1, SEL, PEL and consensus 2 values calculated with the total dataset

Chemical

LEL

TEL

Consensus 1

SEL

PEL

Consensus 2

Asa

7.9

7.9

50

50

Cda

0.71

1.2

0.93

13

2.6

7.8

Cra

25

26

26

90

45

68

Cua

13

16

14

85

34

60

Hga

0.28

0.18

0.23

1.8

0.47

1.2

Nia

15

7.5

11

44

19

32

Pba

19

31

25

167

68

118

Sea

1.5

1.5

6.4

6.4

Sna

1,9

0.85

1.4

21

3.3

12

Zna

129

163

146

1300

305

800

Acenaphthenea

0.05

0.04

0.04

5

1.6

3.3

Acenapthylenea

0.01

0.04

0.03

8.8

1.6

5.2

Anthracenea

0.03

0.03

0.03

0.23

0.12

0.17

Benz(a)anthracenea

0.11

0.12

0.12

0.81

0.40

0.60

Benzo(a)pyrenea

0.16

0.12

0.14

0.81

0.40

0.60

Benzo(b)fluoranthenea

0.19

0.14

0.17

0.88

0.44

0.66

Benzo(e)pyrenea

0.25

0.17

0.21

1.4

0.48

0.93

Benzo(g,h,i)perylenea

0.12

0.1

0.11

0.6

0.3

0.45

Benzo(k)fluoranthenea

0.08

0.07

0.08

0.40

0.23

0.32

Chrysenea

0.14

0.16

0.15

1.2

0.48

0.83

Dibenz(a,h)anthracenea

0.02

0.02

0.02

0.16

0.07

0.12

Phenanthrenea

0.16

0.20

0.18

1.2

0.56

0.89

Fluoranthenea

0.21

0.30

0.25

1.6

0.88

1.2

Fluorenea

0.03

0.06

0.04

0.29

0.24

0.26

Indeno(1,2,3-cd)pyrenea

0.13

0.10

0.12

0.66

0.31

0.48

Naphthalenea

0.07

0.32

0.20

10

2.8

6.6

Perylenea

0.07

0.05

0.06

0.29

0.13

0.21

Pyrenea

0.25

0.23

0.24

1.2

0.69

0.94

6 PAHs Borneffa

0.95

0.86

0.91

5.14

2.58

3.86

PCB 101b

0.68

0.41

0.54

8.7

4.7

6.7

PCB 118b

0.43

0.43

6.9

6.9

PCB 138b

0.85

1.2

1

11

4.3

7.5

PCB 153b

1.2

1.8

1.5

13

6

9.7

PCB 170b

0.08

0.30

0.19

4.3

1.4

2.8

PCB 180b

0.81

0.07

0.44

9.4

1.6

5.5

PCB 28b

0.005

0.07

0.04

3.9

0.14

2.0

PCB 31b

0.005

0.06

0.03

3.4

0.25

1.9

PCB 49b

0.02

0.17

0.1

4.3

0.92

2.6

PCB 52b

0.01

0.18

0.1

7.0

2.2

4.6

Total PCB'sb

7.6

8.0

7.7

80

30

55

4,4′-DDDb

0.01

0.12

0.06

5.1

1.3

3.2

4,4′-DDEb

0.39

0.24

0.31

11

2.2

6.8

Hexachlorobenzeneb

0.0004

0.0004

0.72

0.72

Extractable organohalogenesc

7.9

0.002

5.0

50

4.8

27

Non polar hydrocarbonsa

147

0.161

154

865

392

628

aValues in mg kg−1 DW

bValues in μg kg−1 DW

cValues in mg Cl kg−1 DW

3.3 Predictability and sensitivity

Based on chemical concentrations of all measured compounds found in the sediments during 2000–2001, 144 out of 299 samples have one or more compounds of which the concentration exceeds the consensus value 2, meaning that toxic effects can be expected (Table 3). Based on the concentrations of the metals, PAHs or PCBs 85, 49 or 78 sites were predicted to be toxic, respectively. In all cases, around 97% of the samples that are expected to show toxic effect, had a mortality of >20% in the bioassay with H. azteca, indicating that they were toxic. However the sensitivity (see Table 3) clearly shows that correctly predicted toxicity based on metals, PAHs or PCBs can only explain a fraction of the toxicity observed over all the samples (29%, 17%, and 27%, respectively). However, if all measured contaminants are considered together, at least 49% of the samples that show toxic effects in the bioassay are correctly predicted to be toxic.
Table 3

The number of sites in the period 2000 and 2001 (n = 299) of which the chemical concentrations of certain groups of contaminants or of all measured contaminants is above the consensus 2 value (predicted toxic), the number of these locations that is showing toxic effects based on the bioassay Hyalella azteca (correctly predicted toxic) and predictability (correctly predicted toxic/predicted toxic) and sensitivity (correctly predicted toxic/total number showing toxic effects in the bioassay (n = 282))

Chemical

Predicted toxic

Correctly predicted toxic

Predictability

Sensitivity

Total metals

85

83

0.98

0.29

Total PAHs

49

48

0.98

0.17

Total PCBs

78

75

0.96

0.27

Total

144

139

0.97

0.49

3.4 International guidelines

The freely soluble water concentrations calculated based on the sediment consensus values 1 are for most compounds in the same order of magnitude as the AA-EQS or the PNEC values for surface waters (Table 4). Only the value for hexachlorobenzene is more than 7,000 times lower, so more stringent, if calculated based on the consensus values. For the other compounds, eight have consensus values that are more stringent than the AA-EQS (factors up to 11), while one is less stringent (factor 0.9). If the values are compared to the PNEC values, 15 values derived from the SQGs are more stringent (factors up to 63) and four are less stringent (factors down to 0.05).
Table 4

Consensus values 1 based on this study, Koc values for the organic compounds, based on the consensus values and the Koc values water concentrations, the AA-EQS for the European priority substances (Commission 2008), PNEC water (μg L−1) from the updated prioritisation exercise for priority substances (James et al. 2009), and in the last column are the final Flemish sediment quality guidelines as published on July 9th 2010

CAS

Chemical

Consensus 1 (μg kg−1 DW)

Koc

Water SQG (μg L−1)

AA-EQS (μg L−1)

PNECwater (μg L−1)

Final Flemish SQG (μg kg−1 DW)

7440-38-2

As

7,900

   

4.2

19,000

7440-43-9

Cd

930

  

0.08

0.08

1,000

7440-47-3

Cr

26,000

   

3.4

62,000

7440-50-8

Cu

14,000

   

1.4

20,000

7439-97-6

Hg

230

  

0.05

0.05

550

7440-02-0

Ni

11,000

  

20

3.8

16,000

7439-92-1

Pb

25,000

  

7.2

2.3

40,000

7782-49-2

Se

1,500

   

0.95

 

7440-31-5

Sn

1,400

   

1.5

 

7440-66-6

Zn

146,000

   

10.8

147,000

83-32-9

Acenaphthene

40

5,027

0.16

 

3.8

200

208-96-8

Acenapthylene

30

5,027

0.12

 

1.3

200

120-12-7

Anthracene

30

16,360

0.037

0.1

0.11

100

56-55-3

Benz(a)anthracene

120

176,900

0.014

 

0.012

150

50-32-8

Benzo(a)pyrene

140

587,400

0.005

0.05

0.022

150

205-99-2

Benzo(b)fluoranthene

170

599,400

0.006

0.03

0.027

200

192-97-2

Benzo(e)pyrene

210

599,400

0.007

   

191-24-2

Benzo(g,h,i)perylene

110

1,951,000

0.001

0.002

0.0082

130

207-08-9

Benzo(k)fluoranthene

80

587,400

0.003

0.03

0.017

200

218-01-9

Chrysene

150

180,500

0.017

 

0.07

210

53-70-3

Dibenz(a,h)anthracene

20

1,912,000

0.0002

 

0.0014

100

85-01-8

Phenanthrene

180

16,690

0.22

 

1.3

210

206-44-0

Fluoranthene

250

55,450

0.090

0.1

0.01

370

86-73-7

Fluorene

40

9160

0.087

 

2.5

100

193-39-5

Indeno(1,2,3-cd)Pyrene

120

1,951,000

0.0012

0.002

0.0027

140

91-20-3

Naphthalene

200

1,544

2.6

2.4

2.4

100

198-55-0

Perylene

60

599,400

0.002

   

129-00-0

Pyrene

240

54,340

0.088

 

0.0046

300

37680-73-2

PCB 101

0.54

127,900

0.00008

 

0.00003

0.4

31508-00-6

PCB 118

0.43

127,900

0.00007

 

0.00003

0.3

35065-28-2

PCB 138

1

213,600

0.00009

 

0.00002

0.7

35065-27-1

PCB 153

1.5

209,300

0.00014

 

0.00002

0.9

35065-30-6

PCB 170

0.19

356,800

0.00001

   

35065-29-3

PCB 180

0.44

349,700

0.00003

 

0.00002

0.6

7012-37-5

PCB 28

0.04

47,700

0.00002

 

0.00003

0.1

16606-02-3

PCB 31

0.03

47,700

0.00001

  

0.1

41464-40-8

PCB 49

0.1

78,100

0.00003

  

0.1

35693-99-3

PCB 52

0.1

78,100

0.00003

 

0.00003

0.1

72-54-8

4,4′-DDD

0.06

117,500

0.00001

 

0.00064

0.3

72-55-9

4,4′-DDE

0.31

117,500

0.00005

 

0.0006

0.5

118-74-1

Hexachlorobenzene

0.0004

6,195

0.000001

0.01

0.01

 
When comparing the different “no effect” guidelines for heavy metals, consensus values 1 are closest to the LEL and TEL values of Florida (Table 5). The class II of the Elbe, the limit class A for the Venice Lagoon and the Dutch Target Values, although generally higher, are the same order of magnitude. Regarding PAHs, only consensus value 1 and LEL and TEL for Florida have values which are quite alike for the individual compounds. The sum parameters for PAHs differed a lot between the different SQGs. The SQGs of the Netherlands and Venice Lagoon are a factor of 1,000 higher than the consensus value 1. Furthermore, the same compounds are not always used in the sum procedure. With consensus value 1, the sum parameter consists of the six PAHs of Borneff. With some of the other SQGs, the sum parameter consists of 10–11 PAHs. Also the values for the sum parameter of the PCBs differ between the SQGs of the different regions.
Table 5

International SQGs: LEL Florida and TEL Florida (MacDonald et al. 2003), Target value for The Netherlands (Babut et al. 2003), Limit Class A for Venice Lagoon (Apitz et al. 2007), Elbe Class II (Heise et al. 2005), consensus value 1 for heavy metals PAHs and PCBs and International SQGs: SEL Florida and PEL Florida (MacDonald et al. 2003), QO for the Netherlands (Babut et al. 2003) for PAHs, limit class B for Venice Lagoon (Apitz et al. 2007), consensus value 2 for heavy metals, PAHs and PCBs

Chemical

LEL (Florida)

TEL (Florida)

Target value (The Netherlands)

Limit class A (Venice Lagoon)

Class II (Elbe)

Consensus 1

SEL (Florida)

PEL (Florida)

QO (the Netherlands)

Limit class B (Venice Lagoon)

Consensus 2

No effect

Probable effect

As

6

5.9

29

15

20

7.9

33

17

55

25

50

Cd

0.6

0.6

0.8

1

0.6

0.93

10

3.5

2

5

7.8

Cr

26

37

100

20

160

26

110

90

380

100

68

Cu

16

36

35

40

40

14

110

197

35

50

60

Hg

0.2

0.17

0.3

0.5

0.4

0.23

2

0.49

0.5

2

1.2

Ni

16

18

35

45

60

11

75

36

35

50

32

Pb

31

35

85

45

50

25

250

91.3

530

50

118

Zn

120

123

140

200

200

146

820

315

480

400

800

Anthracene

0.22

0.03

3.7

0.17

Benz(a)anthracene

0.32

0.032

0.12

14.8

0.385

0.60

Benzo(a)pyrene

0.37

0.032

0.14

14.4

0.782

0.60

Chrysene

0.34

0.057

0.15

4.6

0.862

0.83

Dibenz(a,h)anthracene

0.06

0.02

     

Phenanthrene

0.56

0.042

0.18

9.5

0.515

0.89

Fluoranthene

0.75

0.111

0.25

10.2

2.355

1.2

Fluorene

0.19

0.04

1.6

0.26

Pyrene

0.49

0.053

0.24

8.5

0.875

0.94

Sum PAHs

4

1,000

1,000

0.91

100

1,000

10,000

3.86

Sum PCBs

0.07

0.0341

200

10

2

7.7

5.3

0.277

200

200

0.055

All values in mg kg−1 DW, only the sum of PCBs in μg kg−1 DW

Regarding the “probable effect” guidelines for heavy metals, there is no clear pattern (see Table 5). For some compounds, the LEL and PEL values are the highest, while for some it is the consensus value 2 or QO values. Again, the sum parameters for PAHs and PCBs of the Netherlands and Venice Lagoon are very high compared to the other SQGs.

4 Discussion

A whole variety of approaches can be used to develop SQGs, each having certain advantages and limitations (Contaminated Sediment Standing Team 2003; MacDonald et al. 2000). The method used for the derivation of SQGs is also strongly dependent on the future use of the guideline. Guidelines may be used in several ways: as indicators of the existing quality of a site, as guidance for determining whether a further site investigation or a remedial action is needed, as guidance for determining when risk assessment is necessary or to verify the success of remediation measures (Contaminated Sediment Standing Team 2003; Long and MacDonald 1998). One of the approaches to derive SQGs is the use of co-occurrence data, but the weak point is that the data are usually not collected at exactly the same location and time. In addition to using chemical, ecological and ecotoxicological data collected from the same volume of sediment, combining both ecological and ecotoxicological SECs is a strength of the SQGs derived in this study.

The ecological tools used in this study are LEL and SEL. The approach used to calculate LEL and SEL does not require a priori assumptions concerning the cause–effect relationships between environmental factors such as contaminants and the macroinvertebrate species (Persaud et al. 1992). It only considers the range of concentrations of a contaminant at sites where a species is present. This is in contrast to other methods such as multivariate analysis where the focus is to look for relations between environmental factors and the occurrence of species. LEL is the concentration of a chemical in the sediment below which 95% or more of the considered macroinvertebrate species should be able to survive. At concentrations above SEL, only 5% or less of the considered species will be able to survive. For the calculation of LEL and SEL, MacDonald et al. (1992) proposed that biological data of a minimum of 20 taxa should be available. According to Persaud et al. (1992), a minimum of ten taxa is sufficient, if the data cover the full tolerance range of each taxonomic group. Leung et al. (2005) also derived SQGs based on field data and proposed that only taxa found at a minimum of 30 sites should be included. The biological and chemical data mentioned by Leung et al. (2005) were not collected simultaneously with the biota. The composition of the macrobenthic community will, just as other indicators like abundance or richness, be influenced by multiple anthropogenic and natural stressors. It can certainly not be expected that all taxa are able to occur on all sites. An alternative for soft sediments could be the use of the nematode community as this is a good indicator also for sediment quality (de Deckere et al. 2002; Heininger et al. 2007), but monitoring data of nematode communities are presently too limited. When only taking into account macroinvertebrate taxa found at least at 30 sites, 99 taxa out of 127 should have been excluded from the calculations. However, the rare taxa that only occurred at five to 10 sites were already widely spread over Flanders and not restricted to one specific subbasin. These taxa consisted of sensitive as well as tolerant taxa of all macrobenthic groups. It is important to note that the relative number of sensitive taxa is higher when considering the 51 taxa that only occurred on less than five sites, even if 60% of these taxa do not usually live on or in the sediment (De Pauw and Heylen 2001). These pelagic taxa might also be absent in the samples due to the sampling methodology used. Furthermore, the overall bad and moderate status of Flemish water courses (Gevrey et al. 2010) might have resulted in the overall absence of a number of sensitive taxa. This means that the calculated LEL based on the 76 taxa, used in this study, might be an underestimation due to the absence of a number of sensitive taxa.

The ecotoxicological SECs calculated in this study based on the results of the test with H. azteca are TEL and PEL. At concentrations above PEL, toxic effects are frequently observed. At concentrations below TEL, no acute toxic effects are anticipated linked to the compound for which the TEL is calculated. According to Smith et al. (1996), both the “effect” and “no effect” groups require at least 20 sites for each individual chemical. The data in the “effect” group of which the concentrations were lower than the average value of the “no effect” group were not taken into account in further calculations, as the trend was ambiguous. At the “no effect” sites, toxic effects were measured although the concentration of the chemical was significantly lower than the average “no effect” value. Hence, these compounds were most likely not responsible for the observed effects in the measured concentration range. Since the data are based on field samples instead of tests with spiked sediments, the observed effects are the result of a whole mixture of chemicals and thus the toxicity measured could be caused by other compounds than the compound for which the values are calculated. The “no effect” data with higher concentrations than the average of the “effect” group was also not included. High concentrations of pollutants were found on those sites even though no toxicity was observed. However, bioavailability is not taken into account as the chemical data within the monitoring programme are total concentrations.

The results of SEL and PEL showed, except for some compounds, a general trend with SEL being always at least two times the PEL value. PEL is based on the sensitivity of one single organism, H. azteca, used in a standardised toxicity test while the SEL value is based on the in situ sensitivity of the considered species. The higher SEL values indicate that the in situ community is less sensitive for single compounds than the toxicity test with H. azteca. Hence, sensitive species might have already become extinct.

Selection of the most appropriate SQGs for specific applications can be a difficult task for sediment assessors. One of the advantages of the co-occurrence approach to set guidelines is that this method can be used to develop guidelines for any contaminant that is analysed (MacDonald et al. 1992). However, the observed effects, both in the bioassay as well as for the in situ community, might be a result of combined toxicity of different contaminants, which means that the obtained results can differ from single compound toxicity tests (Leung et al. 2005). A disadvantage of this method is that adverse effects of pollution are only manifested when a taxon is already missing on a site, while taxa that are present can also suffer severely from sublethal stress and indirect effects due to pollutants (Preston 2002). On the other hand, the absence of a taxon is, in the SQG theory, directly linked to toxicity while in reality other factors like biological interactions and habitat characteristics also influence the occurrence of macrobenthic species (Leung et al. 2005). Species might disappear due to increasing competition with another more tolerant species (Preston 2002). These factors do not have to be considered in ecotoxicological tests. The conditions under which the toxicity information is derived in ecotoxicological tests differ from those in the field. In ecotoxicological risk assessment, this is partly overcome by the use of uncertainty factors. Uncertainty factors, also called safety factors, are factors by which the quality guideline is lowered to account for uncertainties in terms of environmental conditions, like for example higher temperature or salinity as well as low dissolved oxygen levels (Connell et al. 1999), or to extrapolate from acute to chronic effects and/or from single to multiple species. The ideal situation is that sufficient data are available so that safety factors can be relatively small (Chapman et al. 1998).

4.1 Predictability and sensitivity

A critical component in the application of SQGs is their ability to correctly predict the presence or absence of toxicity in field collected sediments (Ingersoll et al. 2000). The predictability of the SQGs when all 36 compounds are considered is 97%, but also the guidelines for the groups of metals, PAHs and PCBs proved to be highly predictive. Hence, the sediments are usually polluted by a mixture of pollutants. However, the toxic effect observed on 51% of the samples could not be explained by compounds exceeding the guidelines. It can be expected that there is a great number of pollutants present in sediments, other than the 36 included in our study which might explain the observed toxicity. This could explain why the ability to predict non-toxic samples is so poor. Only 5% of the 299 samples were predicted to be non-toxic of which none were actually non-toxic. However, only 4% of the 1,027 sites of the database used to calculate the SECs was actually unpolluted according to the TRIAD evaluation. Thus, the data used to calculate the SQGs are very much biased towards polluted samples. Another disguising factor can be the use of the H. azteca test to predict toxicity. Vangheluwe et al. (2000) reported that the H. azteca sediment contact test exhibited “all or nothing” responses. Besides this, the test with H. azteca was more sensitive compared to the in situ community, as shown by the higher SEL values compared to the PEL values. On the other hand, again, the acute contact test considers only direct toxic effects. Several studies have shown indirect effects on the macrobenthos, at levels as low as 1,000 times below the acute LC50 for Daphnia magna (Liess and von der Ohe 2005; Schäfer et al. 2007; von der Ohe et al. 2009). However, the test is very useful for classifying samples as very toxic or rather non-toxic, but it was not successful to further discriminate sediments in the grey zone between very toxic and non-toxic. More toxicity tests should be used to validate the obtained SQGs. However, for the moment, the test with H. azteca is the only sediment contact tests conducted in the Flemish monitoring programme.

4.2 International guidelines

Many countries have developed SQGs. Guidelines derived in one region will not be relevant for all regions, because, for example, biochemical reaction rates and biological activity increase exponentially with temperature (Chapman et al. 2006). Also the in situ communities can vary between different regions. Regionally developed SQGs may be less relevant in other regions with different contaminant mixtures (Apitz et al. 2007). Further, the derivation methods, purposes and applicability can vary between different SQGs resulting in different SQGs as shown in Table 4 and 5. The LEL, TEL, SEL, and PEL values calculated in this study and for Florida inland waters are empirically derived. In contrast, the Venice Limit Class A and B values are consensus-based, as is the Elbe Class II. The Dutch target and QO values are equilibrium partitioning derived. Using the equilibrium partitioning methodology to transfer the SQGs derived in this study to concentrations in the water phase shows that this results in concentrations in the same order of magnitude as the AA-EQS for the priority pollutants (Commission 2008) and for the PNEC values derived for the update of this list of priority pollutants (James et al. 2009). In most cases, the SQGs are actually more stringent, which leads to the question if the AA-EQS values are really protective in all cases, considering the more field relevant approach used for the SQGs in an already heavily polluted river basin.

As mentioned before, next to the derivation method, other factors can possibly also explain the differences between the international SQGs. The Venice Lagoon is a marine study area. This may be one of the main reasons explaining the differences between the Limit Class A and B and the other international guidelines, derived for freshwater ecosystems. Moreover, the species considered can also differ between Europe and, for example, the Great Lakes of the USA, which may not be representative for all freshwater ecosystems (Washington Department of Ecology 2002). In contrast to the lakes of the USA and Venice Lagoon, the Dutch rivers are very similar to the Flemish lowland rivers. The differences between the Dutch and Flemish values are unlikely to be due to geomorphologic differences between the aquatic ecosystems or composition of the macrobenthic community.

The authors proposed the use of consensus value 1 as a long term objective and consensus value 2 as a short term objective. Ideally, all sites should reach chemical concentrations below consensus value 1 and the question can be posed in what context it is a good short term objective if you lose 95% of the benthic taxa. As mentioned before, only 4% of the 1,027 sites in Flanders are not polluted at all. If the first aim would be to obtain consensus value 1 for all substances on all sites, 96% of all sites in Flanders need to be remediated dramatically. From the management point of view, this is not feasible in a short time and therefore the first aim is to have all concentrations below consensus value 2 as soon as possible. In this way, the consensus value 2 could be used as a trigger for immediate remediation action. Over a longer and more realistic time period, it is proposed that all sites will have to achieve consensus values 1.

5 Conclusions

This study showed that, using a large dataset containing both ecological and ecotoxicological data for sediment samples, SQGs that are based on multiple types of data can be set, resulting in values that are in the same order of magnitude as the AA-EQS for priority substances in water. Using two different types of data and methods, the obtained ecological and ecotoxicological values were mostly in the same order of magnitude and the respective consensus value 2 correctly predicted the toxicity in 97% of the samples. On the other hand, observed toxicity could only be explained by the measured compounds for 49% of the samples. The consensus values 1 have been used as the basis for the SQGs that were implemented in the Flemish legislation on July 9th 2010, as target values that have to be maintained or reached (see Table 4). However as the number of assessed compounds will always be a limited fraction of the total number of compounds present in the environment, a multiple lines of evidence approach, such as the TRIAD method, should be applied to assess the overall sediment quality.

Acknowledgements

This study was financially supported by the Flemish Environmental Agency (Flanders Environment Report, MIRA) and the European Commission (MODELKEY, Contract No 511237-GOCE). P. C. von der Ohe received financial support through a Deutsche Forschungsgemeinschaft fellowship (PAK 406/1).

Copyright information

© Springer-Verlag 2011