SEARCH

SEARCH BY CITATION

Keywords:

  • Ecological risk assessment;
  • Population modeling;
  • Hazard ratio;
  • Daphnia;
  • 91/414/EEC

Abstract

  1. Top of page
  2. Abstract
  3. ECOLOGICAL RISK ASSESSMENT OF PESTICIDES
  4. SAFETY FACTORS
  5. WHAT DO WE WANT TO PROTECT?
  6. POPULATION MODELS IN ERA
  7. ACUTE TOXICITY OF SPINOSAD
  8. POPULATION MODELS
  9. DISCUSSION
  10. CONCLUSIONS
  11. Acknowledgements
  12. REFERENCES

Traditionally, ecological risk assessments (ERA) of pesticides have been based on risk ratios, where the predicted concentration of the chemical is compared to the concentration that causes biological effects. The concentration that causes biological effect is mostly determined from laboratory experiments using endpoints on the level of the individual (e.g., mortality and reproduction). However, the protection goals are mostly defined at the population level. To deal with the uncertainty in the necessary extrapolations, safety factors are used. Major disadvantages with this simplified approach is that it is difficult to relate a risk ratio to the environmental protection goals, and that the use of fixed safety factors can result in over- as well as underprotective assessments. To reduce uncertainty and increase value relevance in ERA, it has been argued that population models should be used more frequently. In the present study, we have used matrix population models for 3 daphnid species (Ceriodaphnia dubia, Daphnia magna, and D. pulex) to reduce uncertainty and increase value relevance in the ERA of a pesticide (spinosad). The survival rates in the models were reduced in accordance with data from traditional acute mortality tests. As no data on reproductive effects were available, the conservative assumption that no reproduction occurred during the exposure period was made. The models were used to calculate the minimum population size and the time to recovery. These endpoints can be related to the European Union (EU) protection goals for aquatic ecosystems in the vicinity of agricultural fields, which state that reversible population level effects are acceptable if there is recovery within an acceptable (undefined) time frame. The results of the population models were compared to the acceptable (according to EU documents) toxicity exposure ratio (TER) that was based on the same data. At the acceptable TER, which was based on the most sensitive species (C. dubia), the maximum reduction in population size was 13% and the maximum time to recovery was 4 d (both for D. magna). This information is clearly more informative for risk management than a risk ratio. For one of the species, D. pulex, a more complex model, which included sublethal effects on reproduction, was set up. The results of this model were in good agreement with a previous microcosm study and indicated that a traditional TER was overprotective. Integr Environ Assess Manag 2012; 8: 262–270. © 2011 SETAC


ECOLOGICAL RISK ASSESSMENT OF PESTICIDES

  1. Top of page
  2. Abstract
  3. ECOLOGICAL RISK ASSESSMENT OF PESTICIDES
  4. SAFETY FACTORS
  5. WHAT DO WE WANT TO PROTECT?
  6. POPULATION MODELS IN ERA
  7. ACUTE TOXICITY OF SPINOSAD
  8. POPULATION MODELS
  9. DISCUSSION
  10. CONCLUSIONS
  11. Acknowledgements
  12. REFERENCES

Pesticides are designed to have biological effects and are intentionally released into the environment. This means that there is an obvious risk for effects on nontarget organisms. Ecological risk assessment (ERA) is carried out to provide information that can be used by decision makers to determine if society needs to intervene (risk management) by restricting the use of a particular pesticide (Suter 1993). Although ERA is a scientific process, risk management can be seen as a cost-benefit analysis where the environmental cost of using a pesticide is weighed against the benefit it brings. For successful management, which maximizes the net benefit for society, it is essential that the output of the ERA (risk characterization) provides relevant information that helps the risk manager to compare and rank different agricultural practices (Costanza 2006). However, current approaches for ERA in Europe, North America, and elsewhere often provide risk characterizations that cannot easily be related to ecological values. Instead, the risk characterization is most often expressed as a risk ratio, where potential environmental concentrations are compared to concentrations that cause biological effects. In the EU, for example, risk characterization for most types of chemicals is expressed as the ratio between the predicted environmental concentration (PEC) and the predicted no effect concentration (PNEC) (EC 2003). For pesticides, a similar approach is used by expressing the risk as the toxicity exposure ratio (TER), where the toxicity endpoint is divided by the PEC (SANCO 2006).

SAFETY FACTORS

  1. Top of page
  2. Abstract
  3. ECOLOGICAL RISK ASSESSMENT OF PESTICIDES
  4. SAFETY FACTORS
  5. WHAT DO WE WANT TO PROTECT?
  6. POPULATION MODELS IN ERA
  7. ACUTE TOXICITY OF SPINOSAD
  8. POPULATION MODELS
  9. DISCUSSION
  10. CONCLUSIONS
  11. Acknowledgements
  12. REFERENCES

To deal with the uncertainty in extrapolating from laboratory toxicity tests to the real environment, safety factors are applied. For risk characterizations that are based on PEC/PNEC ratios, the PNEC value is determined by dividing the endpoints of the toxicity tests with an application factor. If the PEC/PNEC ratio is greater than 1, the chemical is of concern and further testing is needed before authorization (EC 2003). For pesticides, a similar approach is taken by accepting different levels of TERs, depending on the level of uncertainty (SANCO 2002). For example, the technical guidance document (EC 2003) in support of the directive on new notified substances (93/67/EEC), the directive on biocidal products (98/8/EC) and the regulation of existing substances (1488/94), states that the application factor for intermittent exposures should be 100 for PNECs that are derived from LC50 values (the lethal concentration that kills 50% of the animals). This can be compared to the EU guidance document on aquatic ecotoxicology (SANCO 2002), in support of the directive on plant protection products (91/414/EEC), which states that no authorization should be granted for acute exposures if the TER for fish and Daphnia is less than 100. Similar approaches, with risk ratios and safety factors, are also used in North America (USEPA 2004).

Ideally, management decisions would be based on full knowledge about the ecological impact of different scenarios. In such a case, there would be no need for safety factors. However, the amount of data that would be needed to reach this level of certainty is, of course, never possible to retrieve. Instead, dealing with uncertainty always has to be an integrated part of ERA. Presently this is done by applying larger safety factors when the measurement endpoints are further from the protection goals, so that an acute LC50 value is divided by a larger number than a chronic no observed effect concentration (NOEC; the highest tested concentration that is not statistically different from the control) when the PNEC is determined (EC 2003), or that a lower TER is accepted for long-term exposure compared to acute exposures (SANCO 2002). If more complex studies have been carried out, such as mesocosm or field studies, the safety factors can be further reduced. The safety factors that are used in ERA today are set to be protective for the majority of the species and ecosystems. This means that screening-level ERA tends to be overprotective, potentially leading to unnecessary restrictions. The proportion of overprotective assessments can easily be reduced by using lower safety factors. However, this would mean an increased probability for underprotective assessments, which is unlikely to be accepted among regulators dealing with resources protection or the general public (who ultimately decides what level of risk that is acceptable).

WHAT DO WE WANT TO PROTECT?

  1. Top of page
  2. Abstract
  3. ECOLOGICAL RISK ASSESSMENT OF PESTICIDES
  4. SAFETY FACTORS
  5. WHAT DO WE WANT TO PROTECT?
  6. POPULATION MODELS IN ERA
  7. ACUTE TOXICITY OF SPINOSAD
  8. POPULATION MODELS
  9. DISCUSSION
  10. CONCLUSIONS
  11. Acknowledgements
  12. REFERENCES

Aquatic ecosystems may be contaminated by pesticides through processes such as spray-drift, leaching, and runoff, potentially leading to adverse effects on aquatic organisms. To avoid unacceptable damage to ecosystems, regulatory authorities have set up criteria to protect aquatic organisms from pesticide stress. In the EU, these criteria are found in overlapping documents such as the directive on plant protection products (91/414/EEC), which has its focus on small water bodies in the vicinity of agricultural fields, and the water framework directive (2000/60/EC), which has its focus on large water bodies, such as river basins, and retrospective assessments (monitoring). Because of the different aims of the directives, the acceptable level of impact differs. Brock et al. (2006) discussed 4 principles for differentiation of protection goals; pollution prevention, ecological threshold, community recovery, and functional redundancy. Pollution prevention is the most environmentally conservative principle and is adopted to protect the most sensitive populations by keeping the concentrations of toxic chemicals to a minimum. This approach is sometimes used for setting environmental quality standards for persistent and bioaccumulative chemicals. The ban of DDT in agriculture can be said to be in line with this principle. Functional redundancy is the least environmentally conservative approach and accepts structural effects, such as reduced biodiversity, as long as the function of the ecosystem is not affected beyond an unacceptable level. This principle may be most useful for determining the acceptable level of impact in ecosystems that have a main function in food production, and only limited importance for the conservation of biodiversity. The ecological threshold and community recovery principles represent intermediate levels of protection. The ecological threshold principle accepts a certain degree of contamination as long as the most sensitive species are not (or only briefly) affected, and the community recovery principle accepts reversible effects if recovery occurs within an acceptable time frame. Brock et al. (2006) and Hommen et al. (2010) concluded that the EU directive on plant protection products (91/414/EEC) tends toward the community recovery principle. This means that ERAs of pesticides are carried out to protect small water bodies in the vicinity of agricultural fields from irreversible or long term effects. However, effects on sensitive populations are accepted as long as they recover within an acceptable (not defined) time frame. For environmental assessments of pesticides in larger water bodies, i.e., downstream of the small water bodies close to the fields, the water framework directive (2000/60/EC) tends toward the ecological threshold or even the pollution prevention (for priority substances) principles (Brock et al. 2006). Regardless of which principle for determining acceptable impacts that is used, it is clear that PEC/PNEC ratios or TERs tell the risk manager very little about the probability for unacceptable effects.

POPULATION MODELS IN ERA

  1. Top of page
  2. Abstract
  3. ECOLOGICAL RISK ASSESSMENT OF PESTICIDES
  4. SAFETY FACTORS
  5. WHAT DO WE WANT TO PROTECT?
  6. POPULATION MODELS IN ERA
  7. ACUTE TOXICITY OF SPINOSAD
  8. POPULATION MODELS
  9. DISCUSSION
  10. CONCLUSIONS
  11. Acknowledgements
  12. REFERENCES

To reduce uncertainty and increase value relevance in ERA, it has been argued that population models should be used more frequently (Forbes et al. 2010; Galic et al. 2010; Grimm et al. 2009; Stark et al. 2007). In other scientific fields, such as fisheries management and conservation biology, population models have been used extensively to provide information for decision makers, whereas the use of population models in ERA has been minimal. Population models provide a mechanistic link between the individual and the population (Forbes et al. 2008; Galic et al. 2010), which means that traditional data, with individual level endpoints, can be used as input for the models. Furthermore, population models can be used to integrate effects of multiple stressors, such as habitat loss and harvesting, which is impossible in any meaningful way when a PEC/PNEC ratio or TER is used.

A reduction in uncertainty means that the proportion of overprotective management decisions can be reduced without increasing the number of underprotective decisions. A reduction in the proportion of overprotective management decisions may, in turn, allow limited management resources to be more efficiently directed to cases where populations truly are in danger. Furthermore, risk characterizations based on population level endpoints are likely to be more informative for risk management as they can be related to the protection goals.

In the present study, population models were used to show how uncertainty can be reduced and value relevance can be increased in the ERA of a pesticide. The study was based on acute mortality data for 3 species of daphnids (Deardorff and Stark 2009). The population models were used to show what these data actually mean in terms of population level effects, and the level of uncertainty that has to be addressed. For one of the species, a more complex model was set up including sublethal effects on reproduction. The data for reproductive effects were taken from a previously published life table study (Stark and Vargas 2003). The output of the more complex model, which included sublethal effects, was compared to a previously published microcosm study (Duchet et al. 2008). When assumptions had to be made, they have been conservative to avoid underprotective risk estimates. As daphnids are one of the most important groups of test organisms in ERA and play an important role in aquatic ecosystems (Lampert 2006), it is beneficial to develop more ecologically relevant methods for ERA with daphnids.

ACUTE TOXICITY OF SPINOSAD

  1. Top of page
  2. Abstract
  3. ECOLOGICAL RISK ASSESSMENT OF PESTICIDES
  4. SAFETY FACTORS
  5. WHAT DO WE WANT TO PROTECT?
  6. POPULATION MODELS IN ERA
  7. ACUTE TOXICITY OF SPINOSAD
  8. POPULATION MODELS
  9. DISCUSSION
  10. CONCLUSIONS
  11. Acknowledgements
  12. REFERENCES

In response to restrictions in the use of pesticides, which follows from public concerns about health as well as the environment, producers have developed new pesticides that are designed to be selectively toxic for target species and to degrade more quickly after application (Casida and Quistad 1998). Spinosad (DowElanco, Indianapolis, IN) is an example of a selective pesticide that can be used to control a range of pest species and is quickly degraded in the environment (Cleveland, Bormett et al. 2002). Spinosad is a mixture of spinosyn A and D, which are fermentation products of soil bacterium (Crouse et al. 2001).

In a previous study, acute toxicity of spinosad was determined for 3 species of daphnids, Ceriodaphnia dubia Richard, Daphnia magna Straus, and D. pulex Leydig (Deardorff and Stark 2009). The most sensitive species, according to the acute toxicity data was C. dubia (LC50 = 1.8 µg/L), whereas the least sensitive was D. pulex (LC50 = 129 µg/L). According to the guidance document (SANCO 2002), no authorization should be granted if the TER for acute exposure to Daphnia is lower than 100. The maximum acceptable concentration (MAC), based on the most sensitive species (C. dubia), is thus 0.018 µg/L (MACC. dubia). However, if only D. pulex had been tested, the MAC would be 1.29 µg/L (MACD. pulex). An environmental concentration above this will result in a TER that is less than 100, which means that the chemical is of concern, and further action must be taken. However, as mentioned previously, this ratio tells the risk manager very little about the ecological risk in relation to the protection goals and is not very helpful for setting up a cost-benefit analysis where alternative agricultural practices are compared.

POPULATION MODELS

  1. Top of page
  2. Abstract
  3. ECOLOGICAL RISK ASSESSMENT OF PESTICIDES
  4. SAFETY FACTORS
  5. WHAT DO WE WANT TO PROTECT?
  6. POPULATION MODELS IN ERA
  7. ACUTE TOXICITY OF SPINOSAD
  8. POPULATION MODELS
  9. DISCUSSION
  10. CONCLUSIONS
  11. Acknowledgements
  12. REFERENCES

Model parameterization

Logistic models that describe the dose-mortality relationship were set up based on the 48 h acute mortality data from Deardorff and Stark (2009). The logistic model was written as M48 = 1/(1 + e−α×(lnC−β) (Cavallini 1993), where M48 is the 48 h mortality, C is the concentration of spinosad, and α and β are parameters that describes the shape of the signoid curve. The parameters were determined by fitting the model to the data using the method described by Cavallini (1993). The resulting parameters were α = 0.746, 0.577 and 0.730, and β = 0.617, 1.624, and 4.900 for C. dubia, D. magna, and D. pulex, respectively. Figure 1 shows the 48 h LC50 data for the 3 species, the logistic dose-mortality models, and the MACs (based on a minimum TER of 100) as derived from data on C. dubia (MACC. dubia) and D. pulex (MACD. pulex), respectively.

thumbnail image

Figure 1. Forty-eight hours LC50 values for C. dubia, D. magna, and D. pulex including the logistic dose-mortality models for the 3 species. The solid vertical lines show the lowest maximum acceptable concentration (MAC), according to the EU guidance document, based on data from C. dubia and D. pulex, respectively.

Download figure to PowerPoint

Spinosad is removed from the water column relatively quickly as a result of photolysis and sediment partitioning (Cleveland, Bormett et al. 2002; Duchet et al. 2008). Microcosm studies with spinosad have shown a half-life in the aquatic environment in the range of 0.76 d (Duchet et al. 2008) to 1.8 d (Cleveland, Bormett et al. 2002). To simulate different environmental conditions, the half-life of spinosad in the present study was set to 0.5 d and 2 d, i.e., slightly lower and higher than the observed half-lives, respectively. The concentration at each time step was calculated as Ct = C0 × (1/2)t/HL, where Ct is the concentration at time t, C0 is the concentration at the start of the simulation, and HL is the half-life of spinosad.

The population models were parameterized from life tables for the 3 species (Hanson and Stark 2011a). The life tables were based on daily counts of survival and reproduction for the full life cycles. This work is time consuming and can be compared to more advanced ecotoxicological tests such as life table response experiments (Caswell 1996) and microcosm studies. However, once the models are parameterized, they can be reused for an unlimited number of chemicals and exposure scenarios. The resulting daily survival rates and fertility values were used to set up Leslie matrices (M) for the 3 species. The first row of M consists of age specific fertility values and the subdiagonal contains the age specific survival rates. The age class density vector (n) of time t + 1 is given by the matrix multiplication nt+1 = M × nt (Caswell 2001). The vector consists of the number of individuals in each age class and the matrix multiplication results in exponential population growth with a stable population structure. The dominant eigenvalue of M is equal to the population growth rate (λ), which is the multiplication rate of a population that is not restricted by resource limitations. If λ is greater than 1, the population is growing exponentially, if λ = 1, the population is stable, and if λ is less than 1, the population is going toward extinction.

Models based on acute mortality

Population growth rate

To get an estimate of the effect on λ of the mortality at different concentrations, we used the logistic dose-mortality models together with the matrix population models. This was done by reducing the survival rates in the Leslie matrix, based on the logistic dose-mortality models. As the data presented were for 48 h mortality, and the matrix population model had a projection interval of 1 d, the survival rate for each age class was calculated as Stox = S0 × (1 – M48)1/2, where Stox is the survival rate after toxic impact, S0 is the age specific survival rate without toxic impact, and M48 is the 48 h mortality according to the logistic model. The resulting λ values are shown in Figure 2A. The most sensitive species was C. dubia, where the limit for positive population growth (λ = 1) was 5.26 µg/L, i.e., approximately 3 times higher than the LC50. However, this is based on the assumption that there are no sublethal effects, such as reduced fertility. The concentration 5.26 µg/L should therefore only be seen as a theoretical upper limit for what the population could tolerate without risking extinction, if there are no sublethal effects. Figure 2B shows the reduction in fertility that renders λ = 1 at different concentrations of spinosad. It can be noted that for all 3 species, a reduction in fertility of greater than 99% is required for λ to be below 1 at MACC. dubia. At MACD. pulex, a reduction in fertility of 84% is needed to cause population decline for the most sensitive species (C. dubia).

thumbnail image

Figure 2. The reductions in population growth rate (λ) for C. dubia, D. magna, and D. pulex, as functions of spinosad concentration, are shown in (A). The vertical lines show the maximum acceptable concentration (MAC) that is derived from LC50 data for C. dubia (MACC. dubia) and D. pulex (MACD. pulex). The model assumes no impact on reproduction, which means that the estimated λ values are likely to be underprotective. The reduction in fertility that is needed to get λ = 1 at different concentrations of spinosad are shown in (B). At the MACC. dubia, a reduction in fertility of almost 100% would be needed to cause negative population growth.

Download figure to PowerPoint

Minimum population size and time to recovery

Spinosad concentrations can be expected to peak soon after application of the pesticide, and then return to normal after a time that is determined by the degradation and transportation of the chemical. The effect on the population can be simulated by creating a population projection from matrix multiplication (nt+1 × M × nt) of the Leslie matrix. The reduction in survival at each time step was calculated from the chemical concentration using the equations that were described previously (Ct = C0 × (1/2)t/HL equation image M48 = 1/(1 + e−α×(lnC−β) equation image Stox = S0 × (1 − M48)1/2). To deal with the lack of information regarding effects on fertility, it was assumed that there was no reproduction during the exposure period until the concentration was reduced to the level where the survival rate is back to 99% of the normal. The simulation will therefore render an environmentally conservative (overprotective) result. Although the aim is to get ERAs that are neither over- nor underprotective, it is most in agreement with present day approaches to use conservative assumptions when information is missing. The concentration of spinosad was reduced at each time step using the half-lives 0.5 d and 2 d, respectively.

Two endpoints were retrieved from the simulations, the minimum population size and the time from the start of the exposure until the population has recovered to its original size. In all simulations, the difference in sensitivity between D. magna and C. dubia was small whereas D. pulex was significantly less sensitive (Figure 3). At lower concentrations, D. magna was slightly more sensitive than C. dubia, whereas the opposite was true at higher concentrations. When the half-life was set to 2 d, MACC. dubia (0.018 µg/L) rendered a reduction in population size of 7% and 13% for C. dubia and D. magna, respectively, whereas D. pulex was unaffected. Furthermore, the time to full recovery was 2 d and 4 d for C. dubia and D. magna, respectively. For MACD. pulex (1.29 µg/L), the reduction in population size was 84%, 76%, and 13% and the time to full recovery was 15 d, 16 d, and 3 d for C. dubia, D. magna, and D. pulex, respectively.

thumbnail image

Figure 3. Results from matrix model simulations of C. dubia, D. magna, and D. pulex exposed to spinosad. The vertical lines show the maximum acceptable concentration (MAC) based on the LC50 for C. dubia (MACC. dubia) and D. pulex (MACD. pulex). The minimum population size and the time to recovery for the 3 species when the half-life of spinosad is 2 d are shown in (A) and (B), respectively. The minimum population size and the time to recovery for the 3 species when the half-life of spinosad is 0.5 d are shown in (C) and (D), respectively.

Download figure to PowerPoint

If the half-life was 0.5 d, the reduction in population size at MACC. dubia was 4%, and the time to recovery was 1 d for both C. dubia and D. magna (D. pulex was unaffected). At MACD. pulex, the reduction in population size was 44%, 35%, and 5% for C. dubia, D. magna, and D. pulex, respectively, whereas the time to recovery was 4 d for C. dubia and D. magna compared to 1 d for D. pulex.

Daphnia pulex, acute and chronic data

A population model that included effects on fertility was set up for D. pulex to estimate the total effect on λ at different concentrations. The reductions in fertility were taken from a life table response experiment with spinosad (Stark and Vargas 2003, Hanson and Stark 2011b). The life tables were set up for D. pulex exposed to 0 (control), 2, 4, 6, 8, 10, and 11 µg/L. The basic reproductive rate (R0) was calculated from the life tables as R0 = Σ[lxmx], where lx is the proportion remaining adults and mx is the per capita reproduction at age X. The proportional reduction in R0, compared to the control, was used to reduce the entries in the first row (i.e., the fertility values) of the Leslie matrix for D. pulex. To keep the results of the model environmentally protective, it was assumed that the reduction in fertility at 2 µg/L also applied to all concentrations below 2 µg/L, the reduction in fertility at 4 µg/L was applied to all concentrations between 2 and 4 µg/L, and so on. For concentrations above 11 µg/L, fertility was set to zero. The survival rates were reduced as previously described, using the logistic model for D. pulex. Figure 4A shows the reduction in fertility and survival that was used at different concentrations of spinosad. The stepwise reduction in fertility that is seen in Figure 4A is a result of the conservative approach described above.

thumbnail image

Figure 4. The reductions in fertility and survival that were used in the matrix model are shown in (A). The effect on population growth rate (λ), according to the matrix model, is shown in (B). The critical value of λ = 1 is reached at a spinosad concentration of 8 µg/L. This is well in agreement with the findings of Duchet et al. (2008).

Download figure to PowerPoint

The results of the population model for D. pulex were compared to the results of a previously published microcosm study (Duchet et al. 2008), where a natural population of D. pulex were exposed to spinosad in cube shaped enclosures (50 × 50 × 50 cm) that were pushed into the sediment in a shallow marsh. The study used 3 concentrations of spinosad, 8, 17, and 33 µg/L, with 5 replicate microcosms for each concentration (including the unexposed control). At the lowest concentration (8 µg/L), the population growth rate was slowed down, but recovery was observed after a few days. For the higher concentrations (17 and 33 µg/L), the populations crashed, indicating a λ value less than 1.

Figure 4B shows how λ for D. pulex is affected by spinosad concentration, according to the matrix model. The maximum value for λ in this figure is slightly lower than in Figure 2, which is an effect of the conservative assumptions regarding fertility. Because of the stepwise reduction in fertility, there are also distinct steps in λ. The critical value of λ = 1 is passed during one of these steps, when the spinosad concentration passes 8 µg/L, and hence fertility is reduced to the level that was observed in the life table for 10 µg/L. This is well in line with the results of Duchet et al. (2008), who found that the limit for population extinction is somewhere between 8 and 17 µg/L.

DISCUSSION

  1. Top of page
  2. Abstract
  3. ECOLOGICAL RISK ASSESSMENT OF PESTICIDES
  4. SAFETY FACTORS
  5. WHAT DO WE WANT TO PROTECT?
  6. POPULATION MODELS IN ERA
  7. ACUTE TOXICITY OF SPINOSAD
  8. POPULATION MODELS
  9. DISCUSSION
  10. CONCLUSIONS
  11. Acknowledgements
  12. REFERENCES

Agriculture will always have an immense impact on the environment as it requires transformation of natural ecosystems into farmland, with subsequent habitat loss and reductions in biodiversity (Sala et al. 2000). Furthermore, agriculture affects surrounding aquatic ecosystems by nutrient eutrophication, which can affect the structure as well as the function of the ecosystem (Tilman 1999), as well as pesticide pollution, which can cause declines in sensitive populations (McLaughlin and Mineau 1995). Agriculture is essential to humans for food production, so some environmental impact must be accepted. However, other ecosystem services that may be impaired by agriculture are also essential for our well being (fisheries, nutrient turnover, water purification, etc.). Besides the obvious benefit of reducing pests and increasing crop yields, there are secondary benefits of pesticides such as reduced pressure on uncropped land and reduced energy demand for removing weeds by nonchemical means (Cooper and Dobson 2007). This means that a reduction in the use of pesticides will have certain environmental costs, for example transformation of valuable natural habitats into farmland. To maximize the social benefit of agricultural ecosystems, it is important that the choices of agricultural practices are based on well informed management decisions.

To avoid unacceptable impacts of pesticides on the ecosystem, legislative texts (such as the directive on plant protection products and the water frameworks directive) define different protection goals. These protection goals are mostly linked to the population level (Hommen et al. 2010; Brock et al. 2006), e.g., the protection of the most sensitive populations (ecological threshold principle), that populations can be reduced as long as they recover (community recovery principle), or that impact is accepted as long as some populations that perform important ecological functions are protected (functional redundancy principle). However, current methods for ERA do not produce risk characterizations that can easily be related to the population level, regardless of which principle for environmental protection is used.

In the present study, we have shown how population models can be used to extrapolate data from a traditional 48 h LC50 test into more relevant population level endpoints. Furthermore, in 1 model, we included effects on fertility based on data from a life table response experiment (LTRE). The LTRE is a more time consuming and labor intense experiment, where groups of newborns are observed until all individuals are dead. However, a recent study based on the same data showed that good estimates of λ were retrieved also for data from only 4 days of the life tables (Hanson and Stark 2011b). This suggest that traditional tests to estimate reproductive toxicity, such as the 21 d reproductive test on D. magna (OECD 2008), provide adequate information to be used as input in population models. In all cases, the results of the population models in the present study showed that the daphnids tolerated a higher concentration of spinosad than the MACs based on traditional risk ratios and safety factors. This is in agreement with the larger degree of uncertainty that is present when individual level endpoints (LC50, NOEC) are extrapolated to the population level using safety factors. Any reduction in uncertainty by incorporating more ecological theory should, in most cases, lead to a higher MAC. However, in some cases the MAC may be unchanged, or even reduced, when more sophisticated methods are used and uncertainty, hence, is reduced. This is not known a priori, so a general reduction in safety factors could not be an acceptable way to reduce the number of overprotective assessments. Furthermore, it must be remembered that the endpoints of population models can never be used for absolute conclusions about ecological effects as there will always be a certain degree of uncertainty remaining. This is because of errors in data sampling and assumptions, as well as the fact that there is a variation in nature that can never be fully captured in a model. However, we believe that incorporation of ecological theory, through population models, will significantly reduce the uncertainty, and that smaller safety factors therefore can be used. Population models also allow other approaches for dealing with uncertainty, e.g., stochastic models and Monte Carlo simulations (e.g., Hanson 2009; Hanson et al. 2005). Furthermore, it is easier to relate to the uncertainty of a population level endpoint than to the uncertainty of a risk ratio. For example, the results presented in Figure 4 shows that D. pulex will not have negative population growth (λ < 1) unless the concentration of spinosad is above 8 µg/L. A safety factor of 2 would, in this case, render a MAC of 4 µg/L, at which λ = 1.44 (indicating a 44% daily increase in population size). These results actually tell the risk manager something about the margin of safety, compared to a TER of 100 or 50, or a PEC to PNEC ratio of 1 or 2, which means nothing in terms of population level effect.

The population models that were used in the present study were parameterized from life tables carried out in the laboratory. Under such conditions, there is no mortality from predation, no food shortage or other resource limitations, and the environment is stable (e.g., pH, temperature, light conditions). Therefore, it can be assumed that the models in the present study overestimate the λ values in natural populations. Barnthouse (2004) found λ values in the range from 1.11 to 1.65 for 7 species of Daphnia. This indicates that the values found in the present study are at the higher range of possible λ values for daphnids. Therefore, more realistic data need to be used to parameterize models that can be used routinely in ERA. However, to exemplify how population models can be used in ERA, the models employed in the present study proved very useful.

In Figure 2, it was shown that the observed reductions in survival are not a threat to the long-term survival of the populations at the MACC. dubia unless there are simultaneous reductions in fertility close to 100%. However, such reductions in fertility are possible and it is, therefore, not possible to draw any certain conclusion about whether the MAC is over- or underprotective for the long-term survival of the species. However, a quickly degraded pesticide like spinosad has its largest effect on nontarget populations by acute mortality, not by impairing reproduction. By assuming that reproduction was completely stopped during the exposure period, environmentally conservative estimates of population decline and time to recovery could be modeled (Figure 3). Brock et al. (2006) concluded that the protection of surface waters in the vicinity of agricultural fields, according to the directive on plant protection products, is in line with the community recovery principle. This means that a certain degree of impact is acceptable as long as the population recovers within a reasonable time. Although the accepted degree of impact and the reasonable time for recovery are not specified in the documents, the data that are presented in Figure 3 clearly provide a better basis for risk management than does a risk ratio.

Figure 4 shows the reduction in λ for D. pulex when both mortality and reduction in fertility were considered. Again, it is clear that the limit for acceptable population effects (according to the community recovery principle) is higher than the MACs. Although MACD. pulex was 1.29 µg/L, the limit for positive population growth was 8 µg/L. These results were in agreement with the results of the microcosm study by Duchet et al. (2008), where the limit for λ = 1 was somewhere between 8 and 17 µg/L. As chronic data allows lower safety factors, it may be appropriate to also use a chronic MAC for comparison. According to the guidance document, a chronic MAC is derived by dividing the NOEC with 10 (SANCO 2002). Unfortunately, the data used in the present study did not provide a chronic NOEC for D. pulex. However, as there were reductions in fertility at the lowest test concentration (2 µg/L), it can be concluded that the chronic MAC cannot be higher than 0.2 µg/L. This can be compared to the chronic NOEC for D. magna, which is 1.2 µg/L (SANCO 2006). Therefore, the chronic MAC would also be overprotective.

As it is impossible to test all species that potentially could be exposed to a pesticide, results from some species have to be used to protect the rest of the species from unacceptable harm. In the present study, it was seen that C. dubia and D. magna were significantly more sensitive to spinosad than D. pulex. According to the guidance document (SANCO 2002), ERA of a pesticide always require 1 acute Daphnia test, 2 acute fish tests (1 cold water and 1 warm water species), and 1 test on green algae. These tests represent 3 trophic levels that can be considered as essential for all aquatic ecosystems. If the assessed pesticide has certain properties, such as slow dissipation time or a high bioconcentration factor, further tests are needed (SANCO 2002). In the case of spinosad, daphnids are more sensitive than fish and algae (Deardorff and Stark 2009; Cleveland, Mayes et al. 2002b). This means that the MAC for spinosad is based on LC50 for daphnids. However, as C. dubia was 72 times more susceptible to spinosad than D. pulex (Deardorff and Stark 2009), the MAC will differ greatly depending on which species of daphnids that is tested. Using the large safety factor of 100 on the LC50 value from D. pulex may be regarded as sufficiently protective also for more sensitive species, including other groups of organisms than daphnids (e.g., insects, which are not tested). However, if the species that is tested belongs to the more sensitive species in the community, the safety factor of 100 will lead to a clearly overprotective MAC, resulting in unnecessary restrictions in the use of the pesticide. As the TERs and PEC to PNEC ratios generally are applied to the lowest of the available effect endpoints, testing more species will only affect the MAC if they are lower than the data already used (Forbes et al. 2008). However, when more species are tested, it is less likely that the lowest value will be from one of the least sensitive species in the community. Therefore, it would make sense to use lower safety factors when more species have been tested. This is true regardless of whether the assessment is based on population level endpoints from population models or traditional risk ratios based on individual level effects. Differences in sensitivity between species may also be caused by differences in life-history strategy (Spromberg and Birge 2005; Calow et al. 1997). In traditional ERA, this is included in the safety factors that are applied to the daphnia data. When population models are used, it is possible to more explicitly deal with this by using different models that represent a number of ecologically important life-history types. Although interspecies extrapolation was beyond the scope of the present study, this should be mentioned as another possible advantage of population models in ERA.

The models that were used in the present study can be used to replace traditional hazard ratios based on the same data (survival and reproduction in individual daphnids). However, when more data, or data from higher levels of organization, are available, better predictions can be made. Therefore, the models presented here should not be seen as alternatives to species sensitivity distributions (SSDs), mesocosms, field studies, or other higher tier assessments. Instead, they should be seen as cost-effective methods to reach higher levels of ecological relevance without the need to collect more data. Increased model complexity, e.g., incorporating environmental stochasticity, may provide further ecological relevance. However, additional data are needed for such models.

CONCLUSIONS

  1. Top of page
  2. Abstract
  3. ECOLOGICAL RISK ASSESSMENT OF PESTICIDES
  4. SAFETY FACTORS
  5. WHAT DO WE WANT TO PROTECT?
  6. POPULATION MODELS IN ERA
  7. ACUTE TOXICITY OF SPINOSAD
  8. POPULATION MODELS
  9. DISCUSSION
  10. CONCLUSIONS
  11. Acknowledgements
  12. REFERENCES

Numerous articles have argued for the use of population models in ERA to extrapolate from the individual level to the population level, to reduce uncertainty, and to increase value relevance for risk managers. In the present study, we have shown how population models can improve ERA of a pesticide based on standard data on 48 h mortality. The resulting population level endpoints, such as population growth rate, reduction in population size, and time to recovery, are clearly more relevant for risk managers compared to traditional risk ratios, regardless of the desired level of environmental protection.

Acknowledgements

  1. Top of page
  2. Abstract
  3. ECOLOGICAL RISK ASSESSMENT OF PESTICIDES
  4. SAFETY FACTORS
  5. WHAT DO WE WANT TO PROTECT?
  6. POPULATION MODELS IN ERA
  7. ACUTE TOXICITY OF SPINOSAD
  8. POPULATION MODELS
  9. DISCUSSION
  10. CONCLUSIONS
  11. Acknowledgements
  12. REFERENCES

We thank Grace and Oriki Jack for their help with this study. N Hanson's participation was financed by the Swedish Research Council.

REFERENCES

  1. Top of page
  2. Abstract
  3. ECOLOGICAL RISK ASSESSMENT OF PESTICIDES
  4. SAFETY FACTORS
  5. WHAT DO WE WANT TO PROTECT?
  6. POPULATION MODELS IN ERA
  7. ACUTE TOXICITY OF SPINOSAD
  8. POPULATION MODELS
  9. DISCUSSION
  10. CONCLUSIONS
  11. Acknowledgements
  12. REFERENCES
  • Barnthouse LW. 2004. Quantifying population recovery rates for ecological risk assessment. Environ Toxicol Chem 23: 500508.
  • Brock TCM, Arts GHP, Maltby L, van den Brink PJ. 2006. Aquatic risks of pesticides, ecological protection goals, and common aims in European Union legislation. Integr Environ Assess Manag 2: e20e46.
  • Calow P, Sibly RM, Forbes V. 1997. Risk assessment on the basis of simplified life-history scenarios. Environ Toxicol Chem 16: 19831989.
  • Casida JE, Quistad GB. 1998. Golden age of insecticide research: Past, present, or future? Annu Rev Entomol 43: 116.
  • Caswell H. 1996. Analysis of life table response experiments. 2. Alternative parameterizations for size- and stage-structured models. Ecol Model 88: 7382.
  • Caswell H. 2001. Matrix population models. 2nd ed. Sunderland (MA): Sinauer Associates. 712 p.
  • Cavallini F. 1993. Fitting a Logistic Curve to Data. College Math J 24: 247253.
  • Cleveland CB, Bormett GA, Saunders DG, Powers FL, McGibbon AS, Reeves GL, Rutherford L, Balcer JL. 2002. Environmental fate of spinosad. 1. Dissipation and degradation in aqueous systems. J Agric Food Chem 50: 32443256.
  • Cleveland CB, Mayes MA, Cryer SA. 2002. An ecological risk assessment for spinosad use on cotton. Pest Manag Sci 58: 7084.
  • Cooper J, Dobson H. 2007. The benefits of pesticides to mankind and the environment. Crop Prot 26: 13371348.
  • Costanza R. 2006. Thinking broadly about costs and benefits in ecological management. Integr Environ Assess Manag 2: 166173.
  • Crouse GD, Sparks TC, Schoonover J, Gifford J, Dripps J, Bruce T, Larson LL, Garlich J, Hatton C, Hill RL, Worden TV, Martynow JG. 2001. Recent advances in the chemistry of spinosyns. Pest Manag Sci 57: 177185.
  • Deardorff AD, Stark JD. 2009. Acute toxicity and hazard assessment of spinosad and R-11 to three cladoceran species and coho salmon. Bull Environ Contam Toxicol 82: 549553.
  • Duchet C, Larroque M, Caquet T, Franquet E, Lagneau C, Lagadic L. 2008. Effects of spinosad and Bacillus thuringiensis israelensis on a natural population of Daphnia pulex in field microcosms. Chemosphere 74: 7077.
  • [EC] European Commission. 2003. Technical guidance document in support of Commission Directive 93/67 EEC on risk assessment for new notified substances, Commission Regulation (EC) no. 1488/94 on risk assessment for existing substances and Directive 98/8/EC of the European Parliament and of the Council concerning the placing of biocidal products on the market. European Chemical Bureau, Ispra, Italy.
  • Forbes VE, Calow P, Grimm V, Hayashi T, Jager T, Palmqvist A, Pastorok R, Salvito D, Sibly R, Spromberg J., et al. 2010. Integrating population modeling into ecological risk assessment. Integr Environ Assess Manag 6: 191193.
  • Forbes VE, Calow P, Sibly RM. 2008. The extrapolation problem and how population modeling can help. Environ Toxicol Chem 27: 19871994.
  • Galic N, Hommen U, Boveco JM, van den Brink PJ. 2010. Potential application of population models in the European ecological risk assessment of chemicals II: Review of models and their potential to address environmental protection aims. Integr Environ Assess Manag 6: 338360.
  • Grimm V, Ashauer R, Forbes V, Hommen U, Preuss TG, Schmidt A, van den Brink PJ, Wogram J, Thorbek P. 2009. CREAM: a European project on mechanistic effect models for ecological risk assessment of chemicals. Environ Sci Polluti Res 16: 614617.
  • Hanson N. 2009. Population level effects of reduced fecundity in the fish species perch (Perca fluviatilis) and the implications for environmental monitoring. Ecol Model 220: 20512059.
  • Hanson N, Stark JD. 2011a. A comparison of simple and complex population models to reduce uncertainty in ecological risk assessments of chemicals: Example with three species of Daphnia. Ecotoxicology 20: 12681276.
  • Hanson N, Stark JD. 2011b. Extrapolation from individual level responses to population growth rate using population modeling. Hum Ecol Risk Assess 17: 13321347.
  • Hanson N, Åberg P, Sundelöf A. 2005. Population-level effects of male-biased broods in eelpout (Zoarces viviparus). Environ Toxicol Chem 24: 12351241.
  • Hommen U, Baveco J, Galic N, van den Brink PJ. 2010. Potential application of ecological models in the European environmental risk assessment of chemicals I: Review of protection goals in EU directives and regulations. Integr Environ Assess Manag 6: 325337.
  • Lampert W. 2006. Daphnia: Model herbivore, predator and prey. Polish J Ecol 54: 607620.
  • McLaughlin A, Mineau P. 1995. The impact of agricultural practices on biodiversity. Agric Ecosyst Environ 55: 201212.
  • [OECD] Organisation for Economic Co-operation and Development. 2008. Test No. 211: Daphnia magna reproduction test. OECD guidelines for the testing of chemicals/section 2: Effects on biotic systems. Paris, France: OECD Publishing.
  • Sala OE, Chapin FS, Armesto JJ, Berlow E, Bloomfield J, Dirzo R, Huber-Sanwald E, Huenneke LF, Jackson RB, Kinzig A., et al. 2000. Biodiversity—Global biodiversity scenarios for the year 2100. Science 287: 17701774.
  • [SANCO] Santé des Consommateurs. 2002. Guidance Document on Aquatic Ecotoxicology in the context of the Directive 91/414/EEC. European Comission, Health and Consumer Protection Directorate-General, Brussels, Belgium. Report No. SANCO/3268/2001.
  • [SANCO] Santé des Consommateurs. 2006. Review report on the active substance Spinosad. European Comission, Health and Consumer Protection Directorate-General, Brussels, Belgium. Report No. SANCO/1428/2001.
  • Spromberg JA, Birge WJ. 2005. Modeling the effects of chronic toxicity on fish populations: The influence of life-history strategies. Environ Toxicol Chem 24: 15321540.
  • Stark JD, Sugayama RL, Kovaleski A. 2007. Why demographic and modeling approaches should be adopted for estimating the effects of pesticides on biocontrol agents. Biocontrol 52: 365374.
  • Stark JD, Vargas RI. 2003. Demographic changes in Daphnia pulex (Leydig) after exposure to the insecticides spinosad and diazinon. Ecotox Environ Safe 56: 334338.
  • Suter G. 1993. Ecological risk assessment. Boca Raton (FL): Lewis Publishers. 538 p.
  • Tilman D. 1999. Global environmental impacts of agricultural expansion: The need for sustainable and efficient practices. Proc Natl Acad Sci USA 96: 59956000.
  • [USEPA] United States Unvironmental Protection Agency. 2004. Overview of the ecological risk assessment process in the office of pesticide programs. Washington DC: Office of Pesticide Programs. Available from: http://www.epa.gov/espp/consultation/ecorisk-overview.pdf