SCOPE 49 - Methods to Assess Adverse Effects of Pesticides on Non-target Organisms

 9.1 INTRODUCTION

The critical step in establishing the impacts of toxicants is the characterization of the exposure encountered by non-target organisms such as humans. This is equally true whether approaching the problem through epidemiological or classical toxicity studies. In either case, it is inadequate to show only that an exposure exists; it is also necessary to establish the nature and magnitude of the exposure and how this correlates with the response of organisms to such exposures. Often, so many toxicants are present in the environment and they have such similar effects that valid estimations of the types of potential hazards cannot be made from the simple demonstration that a toxicant is present. Evaluations of the magnitude of the impacts, if any, are even more remote.

 The conundrum faced by the environmental health scientist is in obtaining valid exposure profiles for the thousands of chemicals currently used in society. Each compound is likely to have hundreds of distinct exposure profiles related to a multiplicity of uses. It is impractical to establish these profiles directly through studies of ambient pollutant concentration except for the most important cases, particularly those which appear to have major human health consequences. Even in these cases, retrospective exposure profiles cannot be established from the direct measurement of ambient pollutant concentrations, a condition that significantly hampers any evaluation. 

Over the last decade, environmental scientists and toxicologists have turned to predictive models to obtain the needed information on exposure profiles. Regulatory agencies and industry are developing environmental fate and exposure models to be used in support of regulated activities. The objective of all such exercises is similarthe integration of mathematical descriptions of the major processes affecting the fate of chemicals through the use of computer technology so that the user can easily examine how a chemical will behave in various environmental situations (e.g., Neely and Mackay, 1982; Burns et al., 1981; Di Toro et al., 1982; Neely and Blau, 1985; Neely and Oliver, 1986; Oliver and Laskowski, 1986; Wood, 1986; NRCC, 1981a; Asher et al., 1985). The agencies hope eventually to make decisions based on predictions of the persistence and fate of chemicals in the environment.

Predictions by these models are based on the physical and chemical properties of a chemical and on measured parameters describing the specific environment of interest. The proponents of their use argue that models provide a feasible alternative to costly field monitoring programs. These models range from relatively simple ones such as the `fugacity' and `persistence' models (NRCC 1981a) to highly complex ones such as the `unified transport', or `UTM-tox', model (Patterson, 1986) or the exposure model incorporated into NRCC's general model of Roberts and co-workers describing the economic impacts of pesticides (Mitchell et al., 1987). The reader may refer to the articles of Imboden (1988), Mackay (1979), and NRCC (1981a) for general introductions to modelling concepts.

The problem today is not the lack of models, computing capacity, or an understanding of the mathematics of programming techniques; the software exists to permit any scientist with a rudimentary understanding of computers to generate exposure profiles.

The issue is the validity and usefulness of the generated predictions. Functionally different models can give very similar results in some cases and drastically different predictions in others. For example, both the fugacity and persistence models can give either very similar or dissimilar predictions about the fate of chlorobenzenes, depending on the constraints and conditions employed (Asher et al., 1985). Sabijic (1987) has expressed the problem most directly:

... In general, models are used in sciences to represent, explain, predict or estimate phenomena of interest. They are simply approximations to reality primarily developed for the purpose of prediction. Thus the real value of a model is neither related to the type and size of the model nor to its results for the training data set, but rather its ability to correctly handle new situations ... bad models are as readily available as the good ones. Moreover, desktop computers and modern software allow people with modest or no statistical backgrounds to construct and `test' elaborate formal models. In addition, microcomputers enable us to abuse models at superhuman speed and to produce enormous volumes of questionable numerical results....

To understand the source of the problem, one must understand how exposure models are constructed; their sensitivity depends on the nature of the construction; and the basic underlying assumptions used to approximate the environmental processes that govern the fate of pollutants released into the environment.

Specifically, the exposure models currently used are constructed of individual models of the input, transport, and removal processes impacting on the fate of a chemical, as well as a model of the environment in which the chemical is ultimately to be distributed. The usefulness of each of these models is dependent on our ability to develop correlations between the nature of chemicals, their mobility, and their persistence in a given matrix. The hypothetical model of the environment, sometimes called the `environmental system', is first created by describing several environmental `compartments' considered to have distinct, uniform characteristics, such as pH, percentage of organic matter, or temperature (Figure 9.1). The `discharge', `loading', or `emission' rates of a chemical are then described by discrete models. Finally, mathematical relationships are developed to describe the intercompartmental transfer and removal processes that influence the dynamics of a chemical. If the description of the processes is accurate, only then is the construct a good model. This is true regardless of the appearance of the output. This is the central most important factor often overlooked by many inexperienced users of exposure models. Complex models can almost always be adjusted to mimic observed pollution patterns. In spite of the apparent success of a model, the predictions are only as valid as the submodels that are incorporated into the construct used to predict exposure profiles. Additionally, since each submodel must be calibrated for the characteristics of a pollutant and an ecosystem, predictions will be inadequate unless the input parameters (e.g., Table 9.1) utilized to calibrate the submodels are themselves adequate. 

Figure 9.1 Basic design of a simple environment model

This paper examines the level of validation and usefulness of key submodels of the processes, the environmental systems, and ultimately, the macromodels constructed on the basis of this knowledge.

A critical point to be examined in evaluating the usefulness of models is the degree to which submodels permit extrapolations from one chemical to another or from one environmental system to another. Additionally, these submodels are only useful to the regulator if the properties characterizing the chemicals and the environmental systems are available. With respect to this issue, the availability of the appropriate information on chlorobenzenes will be examined. Chlorobenzenes are chemicals that have received the attention of industrial chemists for decades, and their usage is common. Their properties have been extensively catalogued, and the usefulness of these data for modelling indicates that the data required are available for use in establishing regulations.

Table 9.1. Common parameters used in exposure models of a chemical's fate in the environment

Purpose	Parameter
Characterization of environment	Month
	Latitude
	Water temperature
	Water depth
	Water volume
	Light attenuation in water
	Sediment weight
	Organic content of sediment
	Concentration of suspended solids
	Number of fish
	Average fish weight
	Chemical emissions
Pollutant characterization	Melting point
	Molecular weight
	Vapour pressure
	Water solubility
	Octanolwater partition coefficient
	Fraction of chemical biodegradable by fish
	Quantum yield and extinction coefficients for
	290400 nm
Process description	Hydrolysis rate constant
	Photodegradation rate constant
	Volatilization rate constant
	Biodegradation rate constants for sediments and
	suspended solids
	Partition coefficients between water and each
	compartment
	Transfer rate constants between water and each
	compartment

1. Are the processes that govern a chemical's fate adequately characterized in the models?
2. Does the integration of the processes provide realistic predictions of a chemical's fate?
3. Does sufficient information exist on the physical and chemical properties of the chemicals to provide input for the models?

9.2 VALIDATION OF KEY SUBMODELS OF PROCESSES 

Once a pollutant enters an aquatic environment, its concentration is influenced by the rate at which it is lost to the atmosphere through volatilization and to sorptive pools such as suspended organic matter, flora, and sediments. Additionally, advective and degradative processes can attenuate the concentrations. The accuracy of all aquatic models is dependent on the accuracy of the submodels of these processes. 

9.2.1 VOLATILIZATION

Extensive work has been carried out to develop a model of the volatilization process for dissolved organic pollutants by Mill et al. (1982), Spacie et al. (1977), Liss and Slater (1974), and Mackay and Yeun (1983). While the model is conceptually well established, its usefulness is dependent on the availability of accurate determinations of the Henry's law constants of the chemicals, as well as the exchange coefficients at the air-water interface. These latter coefficients reflect the depth of the heterogeneous layers at this interface and hence the level of turbulence and the velocity of the wind. The variability induced in the predictions of exposure profiles by these factors may be between one and two orders of magnitude (NRCC 1981a).

In the case of the Henry's law constants, there is a paucity of measurements for most pollutants of interest. Measured constants are not available for the chlorobenzenes. Thus, the constants are conventionally estimated from measurements of a chemical's liquid vapour pressure and solubility in water. In the case of solids, the liquid vapour pressure must be estimated from the solid's vapour pressure, thereby compounding errors. The accuracy of extrapolations from the crystal state to the liquid state for the solids is not established. Another problem in the case of chlorobenzenes is that there are few determinations of their vapour pressures in the 1020° range, and further extrapolations are required. Asher et al. (1985) found that vapour pressures reported for chlorobenzenes normally vary by less than 2050 per cent and that errors in the reported solubilities may generally be a greater limitation on the  use of the submodel. For the highly lipophilic chemicals such as the tetra- and penta-chlorobenzenes, the variation in water solubility can be in the range of 5001000 per cent (Table 9.2), significantly hampering any attempt to predict distributional patterns (Asher et al., 1985).

Table 9.2. Variation in physical-chemical properties among chlorobenzenes (Asher et al., 1985) 

	Dichlorobenzenes	Trichlorobenzenes	Tetrachlorobenzenes
	range (% increase)	range (% increase)	range (% increase)
Vapour pressure	1.4781.82 (23.1)	0.3230.389 (19.7)	0.02720.071 (161)
(mm Hg)
Water solubility	89.67124.1 (38.4)	6.5934.57 (424)	0.5954.31 (624)
(mg/1)

Besides the problems presented by the lack of precision in the parameters used to characterize the chemicals, the influence of natural surfactants on volatilization rates has not been established. Mackay and Shiu (1977) have shown that even low concentrations of dissolved surfactants can significantly increase the solubilityand hence potentially decrease volatilization ratesin natural systems as compared with pure water. Today's models assume that the rate of volatilization is unaffected. Thus, while the submodel for volatilization is good at the conceptual level, its use to produce accurate predictions is hampered by very fundamental problems with the data needed for its accurate use.

9.2.2 SORPTION

The major sinks for the highly chlorinated hydrocarbons are the sorptive matrices in aquatic environmental systems. Algae, suspended solids, and sediment serve as major pools in such sinks. The validity of predicted time-dependent exposure profiles of such chemicals can be highly dependent on the accuracy of estimates of the rate of transfer of a chemical in and out of these compartments. 

Assuming that the processes are diffusion controlled and surface-area dependent, as in the persistence model (NRCC, 1981a), one concludes that the time it takes for equilibrium to be achieved is solely dependent on the desorption soil rate constant and inversely related to the watersoil sorption equilibrium constant. If this is correct, situations will occur in which it takes years for equilibrium to be reached in the sediments and suspended solids. For example, using the persistence model, 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) is predicted to take years to reach equilibrium in these matrices (NRCC, 1981b). On the other hand, the fugacity model of Mackay assumes that equilibrium will be reached in hours (Asher et al., 1985). Significantly different exposure profiles (Figure 9.2) will thus be produced depending on which model is used. There is no clear choice between the models. The error that might be introduced in the exposure profiles by use of the incorrect submodel for volatilization can invalidate use of the model in non-equilibrium situations.

Figure 9.2. Predicted concentrations of a hypothetical compound in sediments, suspended solids, and fish when varying log K_ow (Asher et al., 1985)

Models for the desorption process are not validated. However, when equilibrium is assumed in the models, the tacit assumption today is usually that both sorption and desorption processes are described by simple first-order kinetics and that equilibrium is rapidly achieved. For example, this is the assumption used when relations between a chemical's octanolwater partition coefficient and its partition coefficient with organic matter (e.g., Karickhoff et al., 1979; Sabijic,1987) are utilized to describe the partitioning of pollutants between water, sediment, and suspended solids of an environmental system. This is clearly not the case. Empirically determined isotherms for sorption demonstrate that other more complex relations are needed to describe the process. Explicitly, sorptive-desorptive processes can involve multiple phases, some of which may be `bound' (e.g., Khan, 1980; Selvakumar et al., 1988). An additional question raised by the existence of multiple phases is whether each phase is equally available to organisms that ingest contaminated sediments or suspended solids. Today's models cannot provide valid exposure profiles in such situations.

The octanolwater partition coefficient (K_ow)/soil sorption coefficient relationship is often used to provide gross predictions of the influence of these sinks on the concentration of organic pollutants in the water phase at equilibrium. The field validation of these relationships is often difficult, and in the case of chlorobenzenes, not possible (Asher et al., 1985). Asher et al. concluded in the study of chlorobenzenes that the variation in the empirically determined octanolwater partition coefficients was sufficient to mask any variation among isomers in exposure profiles (Table 9.3). In the case of these chemicals, this fact severely limits the usefulness of the predictions, since the toxicity of chlorinated aromatics reflects the specific isomeric structure of the compound ingested. In general, these relations provide the basis for only gross submodels of the sorption phenomenon. Their utilization in models of environmental systems has to be done with great care to avoid overestimation of the results. This is in part due to the fact that few studies report the organic content of the suspended solids and sediments and in part because the size of the sediment layer in equilibrium with the water column is unknown.

Table 9.3. Range of reported experimental 1-octanol water partition coefficients for some industrial organic chemicals

Compound	Range of Kow
Toluene	2.112.80
3,4,5-trichlorophenol	3.604.41
Pentachlorophenol	3.325.86
Trichloroethene	2.293.30
Benzene	1.562.15
Naphthalene	3.014.70
Chlorobenzene	2.183.79
Pentachlorobenzene	4.885.69
Hexachlorobenzene	4.137.42
2,2'-dichlorobiphenyl	4.045.00
2,4,5,2' ,4' ,5' -hexachlorobiphenyl	6.348.11
p,p'-DDT	3.986.36
Aldrin	5.527.40

The values given to these parameters can significantly influence the results (Asher et al., 1985). For example, the sediment compartments of the fugacity and persistence models vary by as much as 1000 times in commonly used configurations, and the variation in the predicted exposure patterns for the two models can be extremely wide (Figure 9.2). The variation reflects the fundamental differences in the way the sorptive processes are treated and not in the way the characteristics of the chemicals are treated. Presently, the correct choice cannot be made.

One major issue faced in developing exposure profiles is the validity of models for the bioconcentration of organic pollutants by aquatic plants. Under some conditions, aquatic plants can serve as major sorptive pools for industrial pollutants; and these pools can represent significant exposure vectors for benthic organisms, fish, and other animals that feed upon them. Today, predictions can usually be based only on empirically determined evaluations of accumulation rates that are valid only under the specific conditions encountered in the environmental system in which the measurements were made. Hence, the handling of these compartments with current models requires decisions about their importance.

9.2.3 BIOCONCENTRATION IN FISH

Possibly the best models for any of the processes are those for the accumulation of organic pollutants in fish. These can be either highly sophisticated models based on the energy requirement of the fish (Norstrom et al., 1976) or relatively simple correlative relations of bioconcentration factors with the octanolwater partition coefficient of the chemicals (e.g., Veith et al., 1979). The level of  laboratory validation of these correlations is good for compounds with octanolwater coefficients of less than about 10. The difficulty is the magnitude of the variation in the octanolwater partition coefficients (Table 9.4), which can lead to considerable uncertainty in the predictions. In the case of chlorobenzenes, this factor can range from a low of at least 160 per cent to a high of 10 000 per cent for hexachlorobenzene. The variation completely masks the prediction of isomeric patterns of bioaccumulation. The actual bioaccumulation factors observed in field studies and those predicted by these correlations can be surprisingly divergent in practice, as evidenced by measurements of chlorobenzenes in water and fish in Lake Ontario (Table 9.5). The divergence is far in excess of that expected to arise from the variation in octanolwater partition coefficients and seems likely to reflect analytical problems in the measurement of the contaminants in the field.

Table 9.4. Variations in predicted bioconcentration factors of chlorobenzenes due to the range in reported octanolwater partition coefficients

	Range of	Range of
	maximum BCF,	minimum BCF,
Chlorobenzene	from log Kow	predicted BCF
Mono-	2.462.84	25522
1,2-di-	3.343.59	1382251
1,3-di-	3.383.62	1492381
1,4-di-	3.353.59	1402251
1,2,3-tri-	3.884.27	3968502
1,2,4-tri-	3.934.27	4378502
1,3,5-tri-	4.154.49	67213082
1,2,3,4-tetra-	4.415.05	111839133
1,2,3,5-tetra-	4.465.05	123339133
1,2,4,5-tetra-	4.525.52	138798177
Penta-	4.885.79	2805166535
Hexa-	5.007.42	3548404600114

9.2.4 DEGRADATION

Ultimately, the exposure profile reflects not only the tendency of the chemical to migrate from compartment to compartment, but also its susceptibility to degrade in the various matrices in which it is sequestered. Important degradative processes include photolysis, hydrolysis, and microbial degradation. Unless the impacts of these processes can realistically be included in the models, the validity of the prediction is suspect.

The removal rate constants used in models today may be drawn either from values measured directly in the laboratory or the field, or they may be calculated using published mathematical descriptions and correlations that may be part of the computer programs. Problems exist with both approaches. Using empirical data is a problem because of the sparseness of reliable data on the kinetics of chemicals in the environment and the variability of the data when they are available. Furthermore, when data are collected, the rate constants are derived from widely differing environmental or laboratory conditions. The problem is that such data are often valid only for the particular environmental system or laboratory study in which they were determined. Thus, the data cannot be used to describe other environmental situations.

The problem with using predictive equations is that, with the exception of the equations for estimating primary photodegradation and hydrolysis (NRCC, 1981a; Smith et al., 1979), none of the relationships have been adequately validated. Even in the cases of photolysis and hydrolysis, the role of sensitizers, such as humic compounds, and natural catalysts lessen the value of any predictive relation. Furthermore, in the case of photolysis, key parameters such as quantum yields and spectra relevant to the aqueous environment are frequently missing. For example, measurements of quantum yields of primary photodegradation of chlorobenzenes under conditions relevant to an environmental analysis could not be found. A reliable estimate of the quantum yield of 1,2,4-trichlorobenzene in 2-propanol at 300 hm and in the presence of oxygen was reported by Akermark et al. (1976); therefore, it is necessary to start arbitrarily with this value to model other polychlorobenzenes (Asher et al., 1985). 

Table 9.5. Measured and predicted bioconcentration factors (BCF) for several chlorobenzenes from Lake Ontario 

	BCF
Chlorobenzene	Predicted	Measured
Mono
1,2-di-	138	143
1,3-di-	17
1,4-di-	158	63
1,2,3-tri-	396	5000
1,2,4-tri-	463	5000
1,3,5-tri-	699	10000
1,2,3,4-tetra-	1118	24000
1,2,3,5-tetra-	1414
1,2,4,5-tetra-	1387	17000
Penta-	3155	27000
Hexa-	5248	1270000

The possible role of sensitizers in their own photodegradation has not been examined and, even for these relatively well-characterized chemicals, cannot be quantified. All the predictions reflect arbitrary choices. The rate of biodegradation is determined by many environmental and chemical variables, which are little understood for quantitative prediction (e.g., Lyman et al., 1982). While anaerobic habitats generally retard biodegradation, the degradation of certain compounds, such as some chlorinated hydrocarbon pesticides (e.g., DDT), is enhanced (Hill and McCarty, 1967; Lyman et al., 1982).

Variations in the rate may also be caused by the rate of growth, rate of death, or debilitation of the microbial population. The viability of the population is difficult to determine unless respirometry or similar techniques are employed (Lyman et al., 1982). The use of such techniques is the exception in field studies, but not in laboratory studies. Other environmental variables not already mentioned that affect the biodegradation rate include light, pH, temperature, moisture, oxygen availability, and salinity. Variability in rates may also be due to inter- and intra-species interactions and to previous history, spatial distribution, and enzymatic make-up of the microbial population (Lyman et al., 1982; Bartholomew and Pfaender, 1983).

Comprehensive integrated predictive models of these processes do not exist. Consideration of microbial processes in current models requires the acceptance of highly arbitrary decisions about the rates appropriate for a specific environmental situation. The empirical studies available commonly relate specifically to the conditions of the study; the correctness of the use of such results in other situations is not established. For example, the biodegradation  rate constants used by Asher et al. (1985) for the chlorobenzenes were estimated on the basis of the values from Bartholomew and Pfaender (1983) for monochlorobenzene and 1,2,4-trichlorobenzene in water. Based on the report of Marinucci and Bartha (1979) that the degradation rate of 1,2,3-trichlorobenzene was one-half to one-third that of 1,2,4-trichlorobenzene, the value for 1,2,3-trichlorobenzene was then arbitrarily set at one-third of the reported value of 1,2,4-trichlorobenzene by Asher et al. (1985).

More controlled and standardized research using direct analytical techniques (e.g., chromatography, radiotracers, spectrometry) is needed to obtain reliable degradation rate constants and to establish patterns of degradation. Recommendations for the standardization of methodology and analytical techniques for basic biodegradation studies are outlined in Lyman et al. (1982). Examination of the degradation of a mixture of compounds would be helpful in establishing relative rates and patterns of microbial degradation. 

9.3 VALIDATION OF EXPOSURE MODELS FOR ECOLOGICAL SYSTEMS

Validation has been defined as the testing of the `agreement between the behaviour of the model and the real system' (Mihram, 1973). It is thus a functional aspect of modelling, because it describes how well the model provides a useful level of convergence between predicted and observed patterns of behaviour. The ease with which this analysis can be done depends upon the system or real world being modelled. If considerable theory exists on how a system's behaviour should vary under extreme conditions, experiments can be designed to test the theory, and validation is relatively straightforward. It becomes far more difficult when knowledge of the system is limited or when there appear to be several feasible explanations for the observed behaviour. In such cases, apparent convergence between field observations and predictions can be artifacts of the choice of input parameters.

The difficulty is that the sensitivity of the models to inaccuracies in the submodels and to inaccuracies in their integration is highly dependent on the environmental systems themselves. For example, if photolysis is the dominant removal process postulated in a model, the accuracy of the submodel and hence the full model will only be tested when significant quantities of a chemical partition in favour of other compartments. At the process level, field measurements in any two connected compartments may be sufficient to test model predictions of partitioning or degradation. This is true when two processes dominate.

To validate the integration of the processes at the environmental system level, measurements in all major compartments of the model may conceivably be necessary. It cannot be assumed that if the field data fit the predicted results for two compartments, they will fit the predictions for other compartments. In addition, to prove that the fit of one field test to the predicted results is not likely to change, several independent tests are required. If the predicted behaviour consistently fits results from several field studies, the model may be an adequate description of the real world for predictive purposes. This does not necessarily mean that the model is correct; however, such models are useful within the defined limits. 

Good validation studies require a thorough understanding of the sensitivity of the model's prediction to the character of the environmental system. Only from this knowledge can true tests be designed that actually estimate whether the model is accurate under sensitive conditions. In the development of the fugacity and persistence models (particularly the latter), numerous field studies have been examined to determine the fit of predictions (NRCC, 1981 a; Asher et al., 1985). Fits could normally be achieved by adjusting parameters within the legitimate scope, given the accuracy of the measurements and the descriptions provided of the systems. However, this does not mean that the models are valid. The process models themselves are not validated, and a model's validation cannot be better than this level of validation. The `fits' often reflect arbitrary but judicious choices of characteristics of the system, such as the depth and nature of the sediment layer, and the role of aquatic flora or sinks.

9.4 CONCLUSIONS

The implications of the above observations in the assessment of exposure are the following.

1. It is unlikely that generic models of complex environmental systems will have the precision necessary for the prediction of exposure patterns in widely different ecosystems and for chemicals with widely differing properties.
2. Models of environmental systems must depend largely on empirical determinations of the descriptive processes. Hence, the results will be highly system-specific. This makes their use unreliable in predicting patterns in highly differing ecosystems. On the other hand, good models can be established for a specific ecosystem, if sufficient effort is focused on characterizing the processes.
3. Studies on models of the processes rather than models of environmental systems are undoubtedly the most important step in expanding our predictive capabilities.
4. Model validation is dependent on carefully designed field experiments which examine pollutant levels under conditions producing significantly different exposure patterns.
5. There is a role for standard models of troublesome situations for comparative purposes. Such models can provide tests of the usefulness of the databases and provide a basis for the prediction of patterns from one chemical to another.
6. Models have roles as tools for assessing the limits and the implications of hypotheses about the impact of ecosystem structure and process descriptions on exposure profiles.

Before modelling techniques for predicting exposure profiles can be of general use to regulators, more attention will need to be paid to developing good models of the critical processes. Explicitly, the current focus on the concepts of modelling rather than the details of the models is relatively unproductive. 

9.5 REFERENCES

Akermark, B., Baeckstrom, P., Westlin, U. E., Gothe, R. and Wachtmeister, C. A. (1976) Photochemical dechlorination of 1,2,4-trichlorobenzene. Acta Chem. Scand. B30(1), 49-52.

Anders, W., Doucette, W. and Deckhut, R. M. (1987) Methods for estimating solubilities of hydrophobic organic compounds: environmental modeling efforts. Adv. Chem. Series 216, 3-26.

Asher, S. C., Lloyd, K. M., Mackay, D., Paterson, S. and Roberts, J. R. (1985) A Critical Examination of Environmental ModelingModeling the Environmental Fate of Chlorobenzenes Using the Persistence and Fugacity Models, NRCC Report No. 23990, National Research Council of Canada, Ottawa, 203pp.

Bartholomew, G. W. and Pfaender, F. (1983) Influence of spatial and temporal variations on organic pollutant biodegradation rates in an estuarine environment. Appl. Environ. Microbiol. 45(1), 102-109.

Burns, L. A., Cline, D. M. and Lassiter, R. R. (1981) Exposure Analysis Modeling Systems (EXAMS): User Manual and System Documentation, US Environmental Protection Agency, Environmental Research Laboratory, Athens, Georgia.

Di Toro, D. M., O'Connor, D. J., Thomann, R. V. and St. John, J. P. (1982) Simplified model of the fate of partitioning chemicals in rivers and streams. In: Dickson, K. L., Maki, A. W. and Cairns, J. (Eds) Modeling the Fate of Chemicals in the Aquatic Environment, Ann Arbor Science Publishers, Ann Arbor, Michigan, p. 165.

Hill, D. W. and McCarty, P. L. (1967) Anaerobic degradation of selected chlorinated hydrocarbon pesticides. J. Water Pollut. Control Fed. 39, 1259-1277.

Imboden, D. M. (1986) Mathematical modeling of the behavior of organic micropollutants in the aquatic environment. Commission of European Communities 1986, 460-464. 

Karickhoff, S. W., Brown, D. S. and Scott, T. A. (1979) Sorption of hydrophobic pollutants on natural sediments. Water Res 13, 241-248.

Khan, S. U. (1980) Pesticides in the Soil Environment, Elsevier, New York, 140pp.

Liss, P. S. and Slater, P. G. (1974) Flux of gases across the airsea interface. Nature (Lond.) 247, 181-184.

Lyman, W. J., Reehl, W. F. and Rosenblatt D. H. (Eds) (1982) Handbook of Chemical Property Estimation Methods, Environmental Behavior of Organic Compounds, McGraw-Hill, Toronto, 977pp.

Mackay, D. and Shiu, W. Y. (1977) Aqueous solubility of polynuclear aromatic hydrocarbons. J. Chem. Eng. Data 22, 399-402.

Mackay, D. and Paterson, S. (1981) Calculating fugacity. Environ. Sci. Techno1. 15(9), 1006-1014.

Mackay, D. and Yeun, A. T. K. (1983) Mass transfer coefficient correlations for volatilization of organic solutes from water. Environ. Sci. Technol. 17, 211-217. 

Mackay, D. (1979) Finding fugacity feasible. Environ. Sci. Technol. 13(10), 1218-1223. 

Marinucci, A. C. and Bartha, R. (1979) Biodegradation of 1,2,3- and 1,2,4-trichlorobenzene in soil and in liquid enrichment culture. Appl. Environ. Microbiol. 38(5), 811-817. 

Mihram, G. A. (1973) Some practical aspects of the verification of validation of simulation models. Oper. Res. Quart. 23(1), 17-29.

Mill, T., Mabey, W. R., Bomberger, D. C., Chou, T. W., Hendrey, D. G. and Smith, J. H. (1982) Laboratory Protocols for Evaluating the Fate of Organic Chemicals in Air and Water, Report No. EPA-600/3-82-022, US Environmental Protection Agency, Environmental Research Laboratory, Athens, Georgia, 338pp.

Mitchell, M. F., Vavasour, E., Curry, P. B. and Roberts, J. R. (1987) Quantitative estimation of the risk to humans resulting from exposure to chemicals. Monograph VI, Strengths and Limitations of BenefitCost Analyses Applied to the Assessment: Industrial Organic Chemicals Including Pesticides, NRCC Report No. 24465, National Research Council of Canada, Ottawa, 90pp.

Neely, W. B. and Oliver, G. R. (1986) A chemical runoff model. In: Cohen, Y. (Ed.) Pollutants in a Multimedia Environment. Proceedings of a Workshop on Pollutant Transport and Accumulation in a Multimedia Environment, Plenum, New York, pp. 133-148.

Neely, W. B. and Mackay, D. (1982) Evaluative models for estimating environmental fate. In: Dickson, K. L., Maki, A. W. and Cairns, J. (Eds) Modeling the Fate of Chemicals in the Aquatic Environment, Ann Arbor Science, Ann Arbor, Michigan, p. 127. 

Neely, W. B. and Blau, G. E. (1985) Introduction to environmental exposure from chemicals. In: Neely, W. B. and Blau, G. E. (Eds) Environmental Exposure from Chemicals, CRC Press, Boca Raton, Florida, pp. 1-11.

Norstrom, R. J., McKinnon, A. E. and deFreitas, A. S. W. (1976) A bioenergetics-based model for pollutant accumulation by fish. Simulation of PCB and methylmercury residue levels in Ottawa River yellow perch (Perca flavescens). J. Fish. Res. Board Can. 33, 248-267.

NRCC (1981a) A Screen for the Relative Persistence of Liphophilic Organic Chemicals In Aquatic Ecosystems. An Analysis of the Role of a Simple Computer Model in Screening, NRCC Report No. 18570, National Research Council of Canada, Ottawa, 300pp.

NRCC (1981b) Polychlorinated Dibenzo-p-Dioxins: Criteria for Their Effects on Man and His Environment, NRCC Report No. 18574, National Research Council of Canada, Ottawa, 248pp.

Oliver, R. G. and Laskowski, D. A. (1986) Development of environmental scenarios for modeling the fate of agricultural chemicals in soil. Environ. Toxicol. Chem. 5, 225-232.

 Patterson, M. R. (1986) Unified transport model for organics. In: Patterson, M. R. (Ed.) Pollutants in a Multimedia Environment. Proceedings of a Workshop on Pollutant Transport and Accumulation in a Multimedia Environment, Plenum, New York, pp. 93-115.

Sabijic, A. (1987) Nonempirical modeling of environmental distribution and toxicity of major organic pollutants. In: Kaiser, K. L. (Ed.) QSAR in Environmental Toxicity, Vol. II, Reidel Holland, Dordrecht, and Kluwer Academic Publications, Norwell, Massachusetts, pp. 309-332.

Selvakumar, G. A., Rodrigo, A. M. and Chan, P. (1988) Binding characteristics of hazardous organic substances to clay soils. In: Bell, J. M. (Ed.) Proceedings of the 42nd Industrial Waste Conference, 1987, Purdue University, Lewis Publishers, Boca Raton, Florida, pp. 775-780. 

Smith, J. H., Mabey, W. R., Bohonos, N., Hold, B. R., Lee, S. S., Chou, T. W., Bomberger, D. C. and Mill, T. (1979) Environmental Pathways of Selected Chemicals in Freshwater Systems. III. Laboratory Studies, Report No. EPA-600/7-78-074, US, Environmental Protection Agency, Environmental Research Laboratory, Athens, Georgia.

Spacie, A., Hamelink, J. L. and Waybrant, R. C. (1977) Application of an evaporative loss model to estimate the persistence of contaminants in lentic environments. In: Mayer, F. L. and Hamelink, J. L. (Eds) Aquatic Toxicology and Hazard Evaluation, ASTM Special Technical Publication 634, American Society for Testing and Materials, Philadelphia, pp. 214-227.

Veith, G. C., Defoe, D. L. and Bergstedt, B. V. (1979) Measuring and estimating the bioconcentration factor of chemicals in fish. J. Fish. Res. Board Can. 36, 1040-1048. 

Wood, W. (1986) Use of exposure models to support TSCA priority setting. In: Environmental Modeling for Priority Setting among Existing Chemicals, Workshop, 11-13 November 1985, Gesellschaft fur Strahlen und Umweltforschung MbH, Munchen, pp. 429-436.