Assessing the Environmental Impact of Wastewaters

ECOTOXICOLOGY AND ENVIRONMENTAL SAFETY

40, 88—93 (1998)

ENVIRONMENTAL RESEARCH, SECTION B ARTICLE NO.

ES981647

Assessing the Environmental Impact of Wastewaters P. Isnard RhoL ne-Poulenc Industrialisation, 24 avenue Jean Jaure% s, 69153 Decines Cedex, France Received November 14, 1996

tem must be assessed, and this can be done by taxonomic analyses (biological indices) or biomarkers. Second, the wastewater itself has to be characterized and several approaches may be used: f Conventional approaches based on the measurement of physicochemical and ecotoxicological global parameters f Chemical-specific approaches based on the determination of the ratio of the predicted environmental concentration (PEC) to the predicted no-effect concentration (PNEC) for compounds that may be particularly harmful for the environment f Integrated approaches trying to take into account the fact that a wastewater is a mixture of compounds This paper discusses these three approaches with a special emphasis on the last two.

The conventional approach for assessing the environmental impact of wastewaters uses a set of global physicochemical and ecotoxicological parameters and is well adapted to the vast majority of wastewaters. When some chemicals may be particularly harmful for the environment, a specific approach based on a comparison between the predicted environmental concentration (PEC) and the predicted no-effect concentration (PNEC) may be used. The four steps of extrapolation required for PNEC evaluation are discussed and the importance of the interspecies extrapolation is highlighted. It may also be useful to use an integrated approach relating the characteristics of the wastewater to that of specific compounds. For physicochemical parameters, a simple addition is adequate, whereas, for ecotoxicity, the problem is more complex. The toxicity of a mixture of compounds acting by the same mechanism is often described by the concentration addition model. Although this model is very useful for practical applications owing to its simplicity, a statistical evaluation of its performance indicates that it slightly overpredicts the toxicity of mixtures. A new model derived from the statistically sound ‘‘independence action’’ principle and based on a precise mathematical description of the dose–response relationship is proposed. Applications of this model to mixtures tested with algae demonstrate the accuracy of this model with the experimental data. ( 1998 Academic Press

CONVENTIONAL APPROACH

Wastewaters are primarily controlled by a set of global physicochemical and ecotoxicological parameters ensuring that release limits are not exceeded. Global physicochemical parameters, such as pH, temperature, suspended solids, and chemical oxygen demand, are frequently used but others may also be used, depending on the nature of the effluent. A few ecotoxicological endpoints derived from bioassays such as the daphnia test or the microtox test also are in current use. In France, for example, the 24-h acute daphnia test serves as a basis for the taxes paid to the Water Agencies. This approach is well adapted to the vast majority of wastewaters, especially those that do not contain harmful chemicals with high persistency, toxicity, or bioaccumulation potential, or all three.

INTRODUCTION

Municipal and industrial wastewaters are of major concern for the aquatic environment and have been subjected to regulations by competent authorities for several decades and others for centuries (Charles VI, 1415). In the past, the main problem was the deoxygenation of surface waters and, thus, wastewater releases were mainly monitored and regulated on the basis of their organic matter content. Today, owing to scientific progress and to the fact that organic matter loads have greatly decreased, other types of impact have been highlighted, especially toxic impacts. The main problem remains the quantification of the impact of a wastewater and the identification of its harmful components. The answer lies in two complementary approaches. First, the quality of the receiving aquatic ecosys-

CHEMICAL-SPECIFIC APPROACH

When some chemical may be particularly harmful for the aquatic environment, it should be monitored and specific release limits must be defined. For example, in France, the ‘‘arreˆte´ du 1er mars 1993’’ lists threshold concentrations for the chemicals listed in annex I of Directive EEC/76/464. However, the definition of this threshold concentration is 88

0147-6513/98 $25.00 Copyright ( 1998 by Academic Press All rights of reproduction in any form reserved.

89

ENVIRONMENTAL IMPACT OF WASTEWATERS

not well based scientifically, and, moreover, this approach does not take into account the nature, especially the size, of the receiving river. Thus, assessing the impact of these specific chemicals could be improved by the use of the PEC/PNEC approach, originally developed for the assessment of new or existing substances (e.g., Directive EEC/93/67). In regard to wastewaters, the PEC at the end of the pipe may be easily determined by a simple dilution calculation: dividing the load of the chemical by the flow of the river. Taking into account the temporal distribution of this last parameter can lead to the development of a statistical approach. In contrast, assessing the PNEC is more difficult. This assessment is usually made by an extrapolation of experimental ecotoxicity data with the use of assessment factors that depend on their quantity and nature (e.g., Directive EEC/93/67 and associated technical guidance documents). However, the current assessment factors suffer from a lack of scientific basis and the current extrapolation methods could be criticized [e.g., Roman (1996)]. Thus, there is a need for the development of new methods and a better knowledge of the gaps between the results of classic ecotoxicity tests and the no-effect concentration in a real ecosystem. Four steps of extrapolation may be distinguished, depending on the nature of the available ecotoxicity data: (1) Effect/no-effect (2) Short term/long term (3) Interspecies (4) Laboratory/environment Numerous studies have been published on these different extrapolation steps, but the relative magnitude of these different steps is still debated. Because it is of crucial importance for the assessment of PNEC, several studies were conducted, with the support of the French ministry of the environment and Elf-Atochem, to get insight into these phenomena and be able to improve the current PNEC assessment methods. Results obtained in the Rhoˆne-Poulenc ecotoxicology laboratory are briefly reported here; detailed publications are in preparation. Effect/No-Effect Extrapolation The effect/no-effect extrapolation step was studied in an experimental study based on the 48-h acute daphnia tests. 4-Chlorophenol, tributylphosphate, dinitro-ocresol, phenol, sodium chloride, and cadmium chloride were tested according to normalized protocols. However, to obtain sufficient data to apply regression and ANOVA analyses accurately, tests were performed with a large number of concentrations and five replicates (e.g., 100 daphnias per concentration). The main results are summarized in Table 1.

TABLE 1 Effect/No-Effect Ratios for 4-Chlorophenol, Tributylphosphate, Dinitro-o-cresol, Phenol, Sodium Chloride and Cadmium Choride According to Regression and ANOVA Analyses of 48-h Acute Daphnia Tests

LC /NOEC 50 LC /NOEC 10 Modeled effect at the NOEC

Mean

90th percentile

3.44 1.25 7.5%

5.05 1.97 15.8%

These results quite clearly indicate that there is no great differences between the LC and the NOEC, because, in 50 90% of cases, the ratio is less than 5.05. Of course, this ratio is still lower for the LC , and this endpoint could be 10 considered a good substitute for the NOEC. More precisely, these data show that the mean effect occurring at the NOEC is 7.5%, whereas this effect is less than 15.8% in 90% of cases. In other words, on average, the NOEC is equal to an LC . Therefore, it seems that this effect/no-effect extra07.5 polation step is a rather low uncertainty extrapolation step. Short-Term/Long-Term Extrapolation To assess the relative uncertainty associated with the short-term/long-term extrapolation, an experimental study based on the acute and 21-day chronic daphnia tests was performed, using the same chemicals previously tested. Table 2 presents the acute/chronic ratios derived from both the acute and the chronic tests. Considering the same endpoints, it appears that there is little difference between the results obtained after 48 h and 21 days. The ratio between the LC s or the NOECs varies with the chemical but, on 50 the average, is not very different from unity and is less than 5 in 90% of cases. If the results for the two first steps are combined by computing the ratio between the short-term LC and the long-term NOEC, it appears that the ratio is 50 close to one order of magnitude. It should be recognized that, in this study, the number of chemicals tested being rather limited, the sample is probably not representative of all the existing chemicals. In particular, TABLE 2 Acute/Chronic Ratios for 4-Chlorophenol, Tributylphosphate, Dinitro-o-cresol, and Sodium Chloride Derived from the 48-h Acute and 21-Day Chronic Daphnia Tests

LC 48h/LC 21 day 50 50 NOEC 48 h/NOEC 21 day LC 48 h/NOEC 21 day 50

Mean

90th percentile

2.35 1.85 5.62

4.18 3.56 11.53

90

P. ISNARD

compounds with specific modes of action or having a bioaccumulation potential may behave differently. Indeed, it is possible to find in the literature greater acute—chronic ratios [e.g., Kenaga (1982) and Heger et al., (1995)]. Nevertheless, this second step can be considered a low-to-medium uncertainty step.

TABLE 4 Chronic/Microcosm No-Effect Ratios for 4-Chlorophenol, Tributylphosphate, Dinitro-o-cresol and Sodium Chloride Derived from 21-Day Chronic Daphnia and Algae/Daphnia Continuous Flow Microcosm Tests

Interspecies Extrapolation

Chronic/microcosm

The third step is the interspecies extrapolation and was studied through a statistical study based on bibliographic data. The objective was to calculate, for a great number of chemicals, the ratio between the lowest and the highest toxicity values. Two databases were used, based on the acute and chronic data published by Sloof et al. (1983) and Sloof and Canton (1983). The first comprises 14 chemicals tested on 19 species in acute tests: the lowest/highest toxicity ratios range between 33 and 8971. The second contains 11 chronic endpoints for 8 chemicals, and the ratio ranges between 31 and 10,000. A third database with only algae, daphnia, and fish acute data (i.e., the base-set data sheet) for 55 chemicals also was used and reflects a ratio ranging between 1.5 and 2,000 (see Table 3). From all these data, it clearly appears that the interspecies extrapolation step is of higher uncertainty than the other extrapolation steps. Laboratory/Environment Extrapolation Although, it is not possible to determine the actual threshold concentration in a natural ecosystem, it is possible to use a micro- or mesocosm approach to derive more environmentally relevant endpoints than can be derived from monospecific tests. It seems reasonable to think that the more complex the microcosm, the more environmentally relevant the data. Nevertheless, it also appears from published studies that the more complex the microcosm, the less reproducible it will be. Furthermore, financial and timesaving reasons argue for simpler systems. Thus, to get insight into this fourth extrapolation step, a validation study was conducted on the basis of a small continuous flow microcosm based on the algae/daphnia couple. TABLE 3 Lowest/Highest Toxicity Values Ratios for Three Databases Database 14 chemicals/19 acute tests 8 chemicals/11 chronic tests 55 chemicals/ADF tests

Lowest/highest toxicity 33—8971 31—10000 1.5—2000

Note. Data Published by Sloof et al., (1983), Sloof and Canton (1983), and others.

Mean

90th percentile

0.79

1.15

Comparison was made of the NOEC derived from the chronic daphnia tests and the NOEC derived from the microcosms studies. Table 4 gives the ratio between these two no-effect values and clearly indicates a very slight difference. These results are in accordance with published studies. An analysis of literature data indicates that the lowest NOEC derived from micro/mesocosm experiments correlates well with the lowest NOEC derived from standard single-species tests. Log (NOEC microcosms) "1.07]log (NOEC monosp.)!0.25 On the basis of 34 relevant references (24 different chemicals), this correlation is highly significant (r"0.93) and statistically not different from the equation ½"X. Moreover, the 95% confidence interval is a factor of 25.5, which is quite low compared with interspecies differences. Synthesis As mentionned previously, the objective of this study was not to get absolute values that would require testing a great number of chemicals, but rather to compare the four extrapolation steps. From these results, it clearly appears that the interspecies extrapolation step is the highest uncertainty step. This has to be taken into account when defining an extrapolation method for deriving the PNEC from laboratory ecotoxicity data (Roman et al., 1996). INTEGRATED APPROACH

It may also be useful to relate the characteristics of the whole effluent to the specific chemicals. Indeed, this allows the possibility of balances and the identification of major chemicals at risk (knowledge of which will help to design a specific wastewater treatment process). These balances may easily be done for physicochemical parameters (e.g., organic carbon, organic nitrogen, etc.), but this is not so easy for ecotoxicity. The concentration addition model is widely used to evaluate the toxicity of a mixture of compounds. This model is based on the toxic unit concept defined as the ratio of the

91

ENVIRONMENTAL IMPACT OF WASTEWATERS

concentration of the compound in the mixture to its toxic value, such as the acute daphnia 48-h LC (Sprague and 50 Ramsay, 1965; Sprague, 1970). Toxic unit"C /LC * 50* It has been quite often reported that the toxicity of mixtures is additive; that is, the toxicity of the mixture is equal to the sum of the toxic units. In published studies, the tested mixtures are normally equitoxic, which means that the toxic units are equal for each component, their sum being equal to unity. Thus, when such a mixture is tested, it is quite easy to see if the expected endpoint is obtained (i.e., 50% immobilization/death). Nevertheless, this concept can also be applied to any kind of mixture. In this case, the toxicity of the mixture can be defined as the dilution factor necessary to get the expected endpoint (e.g., the LC ). In France, 50 a unit has been given to this dilution factor: e´quitox/m3. Toxicity (dilution factor)"+ toxic units However, although many researchers agree to think that chemicals acting by the same mode of action are additive, some authors have observed a ‘‘less than additivity’’ response (Hermens and Leeuwangh, 1982; Deneuvy, 1987; Enserink et al., 1991; Munoz and Tarazona, 1993; Chen et al., 1996). Validation of the **Concentration Addition++ Model A decision was made to check this point by applying this method to a great series of mixed chemicals. A quite important number of tests were performed, because, if the deviation from the perfect additivity law is small, it may not be observed from a small number of tests since the biological uncertainty may mask the difference between the experimental result and the theoretical value. The acute daphnia test was chosen for this validation study and 82 tests were conducted for 24 h, because this exposure period is used for the control of wastewaters in France. However, 32 tests were also prolonged to 48 h, because this is the normalized period in the European regulation for existing and new substances. In contrast with many publications that reported studies of chemicals acting by different modes of action (e.g., pesticides or heavy metals or surfactants), a voluntary choice was made of basic organic chemicals or common salts that are frequently encountered in industrial effluents. For each test, the experimental toxicity was measured (expressed as the dilution factor necessary to get a 50% effect) and the corresponding toxicity was computed according to the concentration—addition model from the concentration of each chemical in the mixture related to its LC . The ratio 50 between these two values was then determined, and Table 5

TABLE 5 Statistics of Measured and Concentration–Addition-Computed Toxicity Values Ratios for Mixtures Derived from Acute Daphnia Tests Statistical parameters Number of tests Measured/Computed ratio Mean Standard deviation Probability('1)

24-h tests 82

48-h tests 32

0.80 0.28 0.11

0.87 0.29 0.125

presents the characteristics of the statistical distribution of these ratios for all the tests. From these results, it appears that the concentration—addition model is obeyed quite well while a slightly ‘‘less than additive’’ effect is observed. The measured toxicities of the mixtures are about 20% less than the computed toxicities. This conclusion is, in fact, not really surprising, because it must be recognized that the concentration—addition model has no real scientific basis. Indeed, it suggests first, that effects of chemicals are additive respective to their concentration, which is not possible, because, for example, a mixture of three chemicals at concentrations producing a 40% effect would, of course, not produce a 120% effect. Second, the model suggests that the effect of a chemical is proportional to its concentration, whereas measurements reveal that a sigmoı¨ d model is more appropriate. To overcome these two major drawbacks, an effort was made to develop a new model derived from the ‘‘independence action’’ principle. Development of a Model Based on the **Independence Action++ Principle From a statistical point of view, the combined probability of two independant events can be calculated from the probability of each event (Bliss, 1939): P(A#B)"P(A)#P(B)!P(A)]P(B) The same statistically sound principle can be applied to the probability of the effects of two chemicals acting independantly, providing the probability of the effect of each component can be adequately described. For this purpose, the ‘‘logit’’ model may be applied (or any other model; e.g., probit, weibull). P(A)"Effect(C )"1/(1#exp(!l)) with l"a#b ) log C a a Thus, if a mixture of two chemicals of concentrations C and a C is considered, it is possible to determine its toxicity b expressed as the dilution factor leading to a 50% effect. The

92

P. ISNARD

problem is to find the dilution factor, D, which can be applied to C and C in such a way that the combined a b probability of effect is 0.5: 0.5"Effect(C /D)#Effect(C /D) a b !Effect(C /D)]Effect(C /D) a b

CONCLUSION

Solving this equation for D is possible with the help of numerical methods (e.g., iterative approach) if the coefficients of the logit equations are known. To validate this model, experimental bioassays were conducted by using the 72-h algal growth-inhibition test. This test was selected instead of the acute daphnia test previously used, because it allows an easier and more accurate estimation of the logit dose—response curve. The test was conducted with several mixtures of the same chemicals used for the previous study. A stock solution of each chemical (at a concentration of about 10 e´quitox/m3) was prepared and tested, and the dose—response curves were determined. Then, mixtures were prepared by mixing the two stock solutions in different proportions (e.g., 25/75, 50/50, and 75/25). Table 6 presents the results obtained with a sodium chloride/tributylphosphate mixture (the same toxicity for the two stock solutions being purely coincidental). These data indicate a good agreement between the ‘‘independence action’’ model and the measured data, whereas the ‘‘concentration addition’’ model fails to reproduce the concave form of the experimental response. In conclusion, the model based on the statistically sound ‘‘independence action’’ principle is best suited to a correct description of the toxicity of mixtures. However, it should be recognized that it is not easy to use routinely: first, it requires the knowledge of the dose—response equations and, second, the equation becomes quite complex for a mixture of a great number of compounds.

TABLE 6 Comparison of Measured, Independence-Action-Computed, and Concentration–Addition-Computed Toxicity Values for Sodium Chloride/Tributylphosphate Mixtures Derived from 72h Algal Growth Inhibition Test Ecotoxicity (e´quitox/m3)a Mixture NaCl/TBP 0/100 25/75 50/50 75/25 100/0

The ‘‘concentration addition’’ model, which suffers from a lack of scientific basis, leads to a slight overestimation of the toxicity of a mixture but is rather simple to use. Thus, for practical applications, such as toxicity balances, the ‘‘concentration addition’’ model should be preferred.

Measured

Independence action

Concentration addition

10.8 7.6 7.3 7.8 10.8

10.8 8.3 7.5 8.5 10.8

10.8 10.8 10.8 10.8 10.8

a This French unit corresponds to the dilution factor that has to be applied to get the EC . 50

Conventional approaches based on the measurement of several physicochemical and ecotoxicological parameters are well adapted to the vast majority of wastewaters. When some specific compounds may be particularly harmful to the environment, a PEC/PNEC approach may be used. Although the estimation of the PEC from monitoring data is easy, the assessment methods for the PNEC should be improved. Results regarding the four extrapolation steps from the laboratory data should help in the refinement of these methods. The possibility of performing toxicity balances would be of great help in identifying the major chemical at risk because this knowledge will help to design a specific wastewater treatment process. Although the model described in this work, based on the ‘‘independence action’’ principle, is statistically more sound and more correctly describes the toxicity of a mixture of compounds, the ‘‘concentration addition’’ model can be used to integrate the joint effect of several substances. This last model leads to only a slight overestimation of the toxicity of a mixture and is simpler to use. ACKNOWLEDGMENTS Professor J.-M. Jouany of the University of Rouen (France) is greatly acknowledged for his involvement in the first part of this work, which would not have been possible without the financial cosupport of the French Ministe`re de l’Environnement and ELF-ATOCHEM. The author also thanks G. Roman, P. Cellier, and all the staff of the laboratory of Ecotoxicology of Rhoˆne-Poulenc Industrialisation for their major contributions to the work described in this paper.

REFERENCES Bliss, C. I. (1939). The toxicity of poison applied jointly. Ann. Appl. Biol. 26, 585—615. Charles, VI (1415). Ordonnance royale relative a` la pollution de la Seine. Chen, J., Liao, Y., Zhao, Y., Wang, L., Lu, G., and Zhao, T. (1996). Quantitative structure—activity relationships and mixture toxicity studies of heterocyclic nitrogen compounds. Bull. Environ. Contam. ¹oxicol. 57, 77—83. Deneuvy, J. P. (1987). ¸es lixiviats de de´ charge: Approche me´ thodologique de leur toxicite´ aigue¨ en fonction de diffe´ rents modes de traites. The`se doct. ing. INSA de Lyon et ENTPE de Vaulx en Velin. Enserink, E. L., Maas-Diepeveen, J. L., and Van Leeuwen, C. J. (1991). Combined effects of metals: An ecotoxicological evaluation. ¼ater Res. 25(6), 679—687.

ENVIRONMENTAL IMPACT OF WASTEWATERS Heger, W., Jung, S. J., Martin, S., and Peter, H. (1995). Acute and prolonged toxicity to aquatic organisms of new and existing chemicals and pesticides. Chemosphere 31(2), 2707—2726. Hermens, J., and Leeuwangh, P. (1982). Joint toxicity of 8 and 24 chemicals to the guppy. Ecotoxicol. Environ. Saf. 6, 302—310. Kenaga, E. E. (1982). Predictability of chronic toxicity from acute toxicity of chemicals in fish and aquatic invertebrates. Environ. ¹oxicol. Chem. 1, 347—358. Munoz, M. J., and Tarazona, J. V. (1993). Synergistic effect of two- and four- component combinations of the polycyclic aromatic hydrocarbons: Phenanthrene, anthracene, naphthalene and acenaphthene to Daphnia magna. Bull. Environ. Contam. ¹oxicol. 50, 363—368. Roman, G. (1996). Mise au point d’une me´ thodologie d’e´ valuation des concentrations sans effect des substances sur l’environnement aquatique. The`se

93

de Doctorat es-Sciences ‘‘Toxicologie de l’Environnement.’’ Universite´ de Rouen. Roman, G., Isnard, P., and Jouany, J.-M. (1996). Improved methods for PNEC assessment. [in preparation] Sloof, W., and Canton, J. H. (1983). Comparison of the susceptibility of 11 freshwater species to 8 chemical compounds II: Semichronic toxicity tests. Aquat. ¹oxicol. 4, 271—282. Sloof, W., Canton, J. H., and Hermens, J. L. M. (1983). Comparison of the susceptibility of 22 freshwater species to 15 chemical compounds I: Subacute toxicity tests. Aquat. ¹oxicol. 4, 113—128. Sprague, J. B. (1970). Measurement of pollutant toxicity to fish II: Utilizing and applying bioassay results. ¼ater Res. 4, 3—32. Sprague, J. B., and Ramsay, B. A. (1965). Lethal levels of mixed copper-zinc solutions for juvenile salmon. J. Fish. Res. Board. Can. 22, 425—432.