Endogenous Lifecycle Models for Chemical Risk Assessment.

Despite over 50 years of research on the use of population models in chemical risk assessment, their practical utility has remained elusive. A novel application and interpretation of ecotoxicological models, Endogenous Lifecycle Models (ELM), is proposed that offers some of the benefits sought from population models, at much lower cost of design, parametrization, and verification. ELMs capture the endogenous lifecycle processes of growth, development, survival, and reproduction and integrate these to estimate and predict expected fitness. Two measures of fitness are proposed as natural model predictions in the context of chemical risk assessment, lifetime reproductive success, and the expected annual propagation of genetic descendants, including self (intrinsic fitness). Six characteristics of the ELM approach are reviewed and illustrated with two ELM examples, the first for a general passerine lifecycle and the second for bald eagle (Haliaeetus leucocephalus). Throughout, the focus is on development of robust qualitative model predictions that depend as little as possible on specific parameter values. Thus, ELMs sacrifice precision to optimize generality in understanding the effects of chemicals across the diversity of avian lifecycles. Notably, the ELM approach integrates naturally with the adverse outcome pathway framework; this integration can be employed as a midtier risk assessment tool when lower tier analyses suggest potential risk.

Selecting surrogate endpoints for estimating pesticide effects on avian reproductive success.

A Markov chain nest productivity model (MCnest) has been developed for projecting the effects of a specific pesticide-use scenario on the annual reproductive success of avian species of concern. A critical element in MCnest is the use of surrogate endpoints, defined as measured endpoints from avian toxicity tests that represent specific types of effects possible in field populations at specific phases of a nesting attempt. In this article, we discuss the attributes of surrogate endpoints and provide guidance for selecting surrogates from existing avian laboratory tests as well as other possible sources. We also discuss some of the assumptions and uncertainties related to using surrogate endpoints to represent field effects. The process of explicitly considering how toxicity test results can be used to assess effects in the field helps identify uncertainties and data gaps that could be targeted in higher-tier risk assessments.

Quantifying the effects of pesticide exposure on annual reproductive success of birds.

The Markov chain nest productivity model (MCnest) was developed for quantifying the effects of specific pesticide-use scenarios on the annual reproductive success of simulated populations of birds. Each nesting attempt is divided into a series of discrete phases (e.g., egg laying, incubation, nestling rearing), and results from avian toxicity tests are used to represent the types of effects possible in the field during each breeding phase. The expected exposure dose each day throughout the breeding season can be compared to the toxicity thresholds assigned to each breeding phase to determine whether the nest attempt is at risk. The primary output of the model is an estimate of the number of successful nest attempts per female per year, which is multiplied by the number of fledglings per successful nest to estimate the number of fledglings per female per breeding season (i.e., annual reproductive success). In this article, we present a series of MCnest simulations to demonstrate the extent to which the magnitude of change in annual reproductive success can be affected by considering life history attributes and the timing of pesticide applications relative to a species' typical breeding phenology. For a given pesticide-use scenario, MCnest can identify which species are at greatest risk. By allowing multiple species to be run under a single scenario, it can also help to identify the life-history traits that contribute to a species' vulnerability to a given pesticide-use scenario. It also can determine which application dates have the greatest impact and demonstrate the extent to which pesticide characteristics (e.g., residue half-life, mode of action) affect productivity. MCnest goes beyond the current qualitative screening-level assessments of risks to avian reproduction to provide an approach for quantifying the reduction in annual reproductive success by integrating species life history and timing of pesticide exposures, despite limitations in existing information on species life history and toxicity responses from existing laboratory tests.

Hidden Markov models for estimating animal mortality from anthropogenic hazards.

Carcass searches are a common method for studying the risk of anthropogenic hazards to wildlife, including nontarget poisoning and collisions with anthropogenic structures. Typically, numbers of carcasses found must be corrected for scavenging rates and imperfect detection. Parameters for these processes (scavenging and detection) are often estimated using carcass distribution trials in which researchers place carcasses in the field at known times and locations. In this manuscript I develop a variety of estimators based on multi-event or hidden Markov models for use under different experimental conditions. I apply the estimators to two case studies of avian mortality, one from pesticide exposure and another at wind turbines. The proposed framework for mortality estimation points to a unified framework for estimation of scavenging rates and searcher efficiency in a single trial and also allows estimation based only on accidental kills, obviating the need for carcass distribution trials. Results of the case studies show wide variation in the performance of different estimators, but even wider confidence intervals around estimates of the numbers of animals killed, which are the direct result of small sample size in the carcass distribution trials employed. These results also highlight the importance of a well-formed hypothesis about the temporal nature of mortality at the focal hazard under study.

Adverse outcome pathways and ecological risk assessment: bridging to population-level effects.

Maintaining the viability of populations of plants and animals is a key focus for environmental regulation. Population-level responses integrate the cumulative effects of chemical stressors on individuals as those individuals interact with and are affected by their conspecifics, competitors, predators, prey, habitat, and other biotic and abiotic factors. Models of population-level effects of contaminants can integrate information from lower levels of biological organization and feed that information into higher-level community and ecosystem models. As individual-level endpoints are used to predict population responses, this requires that biological responses at lower levels of organization be translated into a form that is usable by the population modeler. In the current study, we describe how mechanistic data, as captured in adverse outcome pathways (AOPs), can be translated into modeling focused on population-level risk assessments. First, we describe the regulatory context surrounding population modeling, risk assessment and the emerging role of AOPs. Then we present a succinct overview of different approaches to population modeling and discuss the types of data needed for these models. We describe how different key biological processes measured at the level of the individual serve as the linkage, or bridge, between AOPs and predictions of population status, including consideration of community-level interactions and genetic adaptation. Several case examples illustrate the potential for use of AOPs in population modeling and predictive ecotoxicology. Finally, we make recommendations for focusing toxicity studies to produce the quantitative data needed to define AOPs and to facilitate their incorporation into population modeling.

Markov chain estimation of avian seasonal fecundity.

Avian seasonal fecundity is of interest from evolutionary, ecological, and conservation perspectives. However, direct estimation of seasonal fecundity is difficult, especially with multi-brooded birds, and models representing the renesting and quitting processes are usually required. To explore the consequences of modeling decisions on inference about avian seasonal fecundity, we generalize previous Markov chain (MC) models of avian nest success to formulate two different MC models of avian seasonal fecundity that represent two different ways to model renesting decisions and breeding cessation. We parameterize both Markov chains (regular and absorbing) for two species (Eastern Meadowlark, Sturnella magna, and Dickcissel, Spiza americana) and compare the results using mean-square error of the estimated number of successful broods per breeding female. We also provide formulae for estimating the expected variation in female breeding success. The absorbing MC performed better for both species, although the regular MC performed almost as well when the duration of the breeding season was estimated by taking the 95th percentile of a negative binomial distribution fit to the observed durations among all females. In their simplest form the models contain very few parameters (four or five) and should also prove useful as a foundation for more complex models of avian seasonal fecundity and demography.