Spatially Explicit Wildlife Exposure Models: Moving Towards Their Increased Acceptance and Use
Deborah Rudnick, Integral Consulting Inc.
The session Spatially Explicit Wildlife Exposure Models held during the SETAC North America 33rd Annual Meeting in Long Beach, CA offered a wealth of information and examples of how researchers and practitioners are working to improve the accuracy and realism of risk assessments by recognizing the value of spatial information and modeling. Ted Wickwire of Exponent moderated the session, substituting for the original moderator Mark Johnson, who was unable to attend. Wickwire started us off with some fundamental questions, including:
- What are the values and roles of explicit models and what are some of the limitations in their use?
- What are some of the impediments to acceptance by regulators?
- Are there regulatory examples confirming appropriate applications of spatial models?
In traditional risk assessment, we often describe sites as either contaminated or not, with static exposures and with what can be a highly conservative assumption that receptors of concern are consistently exposed to the highest site concentrations. This approach does not capture behavioral dynamics of ecological receptors or real variability in the exposures they might encounter. Multiple talks addressed the advantages of spatially explicit models in helping to overcome some of the limitations of deterministic risk assessment approaches by more accurately evaluating spatial and temporal scales, and quantifying uncertainties and variability across management scenarios. Several of the spatially explicit modeling approaches that were discussed in this session also have the capacity to incorporate probabilistic aspects into the model so that we can ask questions such as "What fraction of the population might be affected by a stressor?" or conversely "What fraction of a population might we be able to protect?"
Tomasz Kulakowski, a doctoral student at Reading University, described the model he has built working with the European Chemical Risk Effects Assessment Models (CREAM) program to evaluate pesticide exposure to two bird species with different dietary patterns and foraging strategies. Kulakowski used an individual-based model which uses decisions about where and when an organism might feed and consequently be exposed to a stressor, and incorporated an explicit GIS landscape with a one-day timestep to model the breeding cycle throughout the season and look at changes in effects on adults and nestlings under different application timings in a spatially and temporally explicit manner. The model showed a surprising amount of early-life stage sensitivity to and variability around chemical exposure and uptake, as well as sensitivity to timing of pesticide application, results that could be used to inform pesticide application decisions. These results illustrate how spatially explicit models that capture the spatial and temporal foraging patterns of the individuals can yield population exposures that may be quite different than the types of exposures that might be predicted via individual, static models.
Multiple presentations in this session by Wickwire and by John Isanhart from the US Fish and Wildlife Service addressed the use of the Spatially Explicit Exposure Model (SEEM), developed by the US Army (Mark Johnson project leader) and Charlie Menzie and Wickwire (of Exponent) and Dmitriy Burmistrov (original programmer). Recent updates have been funded by SERDP/ESTCP. SEEM integrates chemical distribution data with ecological information including habitat suitability indices, species-specific dietary information and a free-range foraging model in order to create spatially explicit estimates of exposure and risks. Case studies were presented with implementations for both bird and mammal species. For birds evaluated at a former mining site and for a separate study for two former small arms firing ranges, SEEM predicted risks that were consistent with empirical data collected for chemical concentrations in avian blood and feathers, but that were substantively lower than risks predicted using a deterministic risk assessment. By contrast, small mammal outcomes were consistent with the deterministic model but well above risk predictions based on field data, which may indicate that SEEM does not currently work as well for species with small home ranges. Both authors identified the need for additional testing of this model to further evaluate and validate outcomes for a variety of species and ecosystems.
Additional spatially explicit models discussed included the Los Alamos National Lab’s (LANL) ECORSK model and the FishRAND model. Randall Ryti of Neptune and Company, Inc. described his use of ECORSK for evaluating risk to Mexican spotted owls at the LANL property. The model evaluated nesting likelihood and foraging areas relative to the distribution of contaminated site soils within a spatially heterogeneous landscape. The model can be run probabilistically to look at exposure distributions and ask questions such as “How likely is a given exposure to occur?” And Wickwire discussed FishRAND, a data-intensive GIS-based model for evaluating risk to fish from contaminated habitats in a spatially explicit manner. Species-specific parameters including foraging strategies are combined with biological, chemical and physical parameters for the study system to produce outputs that can include population distributions as well as uncertainty and variability around these distributions and provide a point of comparison for more spatially static outputs (Figure 1). The probabilistic output includes distributions of endpoints and the data needed to answer questions such as the likelihood that a reduction in fecundity in a population subjected to a chemical stressor will occur.
Figure 1. Output from FishRAND showing comparison of a static output using spatially weighted average concentrations (SWACs) of PCBs in sediment with the outputs from the model using spatially explicit information about fish exposure. Figure courtesy of Katherine von Stackelberg, E Risk Sciences, LLP.
Deb Rudnick of Integral Consulting Inc. discussed temporal, spatial and biogeochemical issues at play in forecasting risk for a pit mine expansion. An integrative conceptual model of lake ecological succession integrating robust geochemical modeling for surface water quality, spatial distribution of chemicals and temporal infill trajectories for the future pit lake were presented as critical components of understanding and accurately estimating risk. Examples of the benefits of this integrated approach were described, including the identification of exposure areas that might provide insight into possible mitigation areas (Figure 2).
Figure 2. A spatially explicit model of future potential mule deer exposure at a Nevada pit mine identifies the areas most likely to be accessed by ungulates. This model provided the basis for evaluating risk management scenarios such as placing material with lower concentrations of metals in the identified exposure area and consequent risk reduction. Figure courtesy Sean Kosinski, Integral Consulting Inc.
Chris Salice of Texas Tech University provided some important conceptual ideas and case studies describing the importance of spatially explicit exposure modeling to improve a broader range of risk assessment scenarios, recognizing the key issues that contaminants vary across sites, and so do organisms, and discussed how individual-based models can provide insights into population-level risks. Case studies included evaluating a mink population exposed to PCBs and avian risks from pesticides using a modified version of EPA’s T-REX model.
It was clear from this session that there are many concepts and tools being developed for addressing ecological risk assessment in spatially and temporally realistic ways. The speakers hope that ecological risk assessors are willing to adopt, adapt and try these tools and approaches so that we can collectively bring greater understanding to and accuracy in predictions of ecological risks.
Author's contact information: email@example.com
Return to the Globe