|STARK, JOHN - Washington State University|
|BANKS, JOHN - University Of Washington|
Submitted to: Environmental Toxicology
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 3/5/2015
Publication Date: 3/11/2015
Citation: Stark, J.D., Vargas, R.I., Banks, J.E. 2015. Incorporating variability in point estimates in risk assessment: bridging the gap between LC50 and population endpoints. Environmental Toxicology. DOI:10.1002/ETC.2978.
Interpretive Summary: Gauging the effects of pesticides on target pest and non-target organisms has become an increasingly critical enterprise in both basic and applied ecology. In agriculture, pesticide use must strike a balance between pest suppression and preserving beneficial species. A risk assessment is used to better understand and predict impacts to beneficial species by understanding their respective responses to pesticides. However, shortcomings in the current risk assessment process has been identified. For example, in the current risk assessment approach, effects are determined on individuals, not populations -- and there is little proof that surrogate species are good predictors of what may happen to other species. Population level approaches, including interactions between species (population modeling), has been suggested as a means of improving the risk assessment process. In addition, it is especially difficult to extrapolate effects from one species to another, even those that are closely related and have similar acute mortality responses to the same pesticides. In this study, we use a matrix model approach to compare the responses of two pairs of closely related species (target pest and beneficial; and two target pests) in order to explore how the variability in their responses to toxicants may yield misleading results in risk assessment. Our results illustrate that two groups of closely related species that exhibit no significant differences in acute mortality to pesticides may in fact respond very differently at the population level after exposure to these pesticides. Our results further highlight the necessity of population level assessments and modeling in ecological risk assessment of pesticides and further enforces the fact that even closely related species are often poor surrogates for one another.
Technical Abstract: Historically, the use of point estimates such as the LC50 has been instrumental in assessing the risk associated with toxicants to rare or economically important species. In recent years, growing awareness of the shortcomings of this approach has led to an increased focus on analyses using population endpoints. Risk assessment of arthropod species in agroecosystems, however, still relies heavily upon large amounts of LC50 data from concentration-response data amassed over decades in the laboratory. Despite the fact that these data are generally well-replicated, little or no attention has been given to the sometime high levels of variability associated with the generation of point estimates. This is especially important in agroecosystems, where arthropod predator-prey interactions are often disrupted by the use of pesticides. Using lab-derived data of four economically-important species (two fruit fly pest species and two braconid parasitoid species) and matrix-based population models, we demonstrate here a method for bridging traditional point-estimate risk assessments with population outcomes. Our results illustrate that even closely-related species can show strikingly divergent responses to the same exposures to pesticides. Furthermore, we show that using different values within the 95% confidence intervals of LC50 values can result in very different population outcomes – ranging from quick recovery to complete extinction -- for both pest and parasitoid species. We discuss the implications of these results for agroecosystem management, and emphasize the need to incorporate variability/uncertainty in point estimates for use in risk assessment.