Spatially explicit ecological exposure models: A rationale for and path toward their increased acceptance and use



Spatially explicit wildlife exposure models have been developed to integrate chemical concentrations dispersed in space and time, heterogeneous habitats of varying qualities, and foraging behaviors of wildlife to give more realistic wildlife exposure estimates for ecological risk assessments. These models not only improve the realism of wildlife exposure estimates, but also increase the efficiency of remedial planning. However, despite being widely available, these models are rarely used in baseline (definitive) ecological risk assessments. A lack of precedent for their use, misperceptions about models in general and spatial models in particular, non-specific or no enabling regulations, poor communication, and uncertainties regarding inputs are all impediments to greater use of such models. An expert workshop was convened as part of an Environmental Security Technology Certification Program Project to evaluate current applications for spatially explicit models and consider ways such models could bring increased realism to ecological exposure assessments. Specific actions (e.g., greater accessibility and innovation in model design, increased communication with and training opportunities for decision makers and regulators, explicit consideration during assessment planning and problem formulation) were discussed as mechanisms to increase the use of these valuable and innovative modeling tools. The intent of this workshop synopsis is to highlight for the ecological risk assessment community both the value and availability of a wide range of spatial models and to recommend specific actions that may help to increase their acceptance and use by ecological risk assessment practitioners. Integr Environ Assess Manag 2011;7:158–168. © 2011 SETAC


Exposure is critical in understanding the likelihood for adverse health effects to wildlife populations from environmental contamination. Wildlife experience the environment relative to species-specific life history constraints (e.g., habitat) which may overlap with variation in environmental contamination. Consideration of space in an environmental context is crucial, where significant resources are devoted to defining the nature and extent of contamination. Models contribute to environmental assessments from early site analyses through remedial planning, implementation, and monitoring. Accounting for spatially explicit relationships is an integral component in many of the most frequently applied fate and transport models, e.g., WASP (Gonenc et al. 2007), WASTOX (Connolly and Winfield 1994), and BASINS (Chigbu et al. 2007). Despite broad acceptance of fate and transport models that include spatial components, ecological risk assessments rarely consider the influences of habitat and contamination in a quantitative spatial context meaningful to an assemblage of individuals (populations).

Models provide scientists with the opportunity to increase the value of data collected at a particular contaminated site by facilitating additional research into alternative scenarios, leveraging data collected for predictions in areas lacking data and, when combined with non site-specific data, increasing the efficiency of future direct sampling by highlighting data gaps and clarifying data needs. Ecologists are challenged to incorporate increasingly realistic wildlife exposure scenarios in their work. Models that incorporate species- and site-specific data, and, where spatial interactions between environmental contamination and habitat preferences are made transparent and accessible, provide increased realism and enhance predictive capabilities. Recognition of the importance of spatial relationships in environmental assessments is not new. Early ecological risk assessment guidance documents discuss the importance of considering spatial characteristics in an assessment (USEPA 1997, 1998). In this context, these models are of value to risk assessors, environmental managers, and decision makers because of their ability to incorporate important spatial considerations related to exposure into risk characterization, and also to identify uncertainties associated with risk estimates. The challenge is to reach consensus among model developers, risk assessors, and regulators (and within each group, as well as among individuals who work in all three areas) regarding appropriate applications for new models. In addition, if consensus can be reached regarding the use of a spatially explicit model, then a project team must identify the appropriate model, model inputs and assumptions, and have a clear understanding about how the results may be interpreted before results are generated.

As part of an Environmental Security Technology Certification Program (ESTCP) project focusing on spatially explicit wildlife exposure model demonstration and testing, a workshop of U.S. Army, U.S. Environmental Protection Agency (USEPA), state regulators, and private sector researchers was convened to evaluate current applications of available spatially explicit wildlife exposure models and approaches for increasing future use of such models. The workshop focused on collecting insights with respect to 4 key questions: 1) What are spatially explicit wildlife exposure models and why are they valuable? 2) How have such models been applied? 3) Are there regulatory impediments to their use? and 4) What are the limitations of these models and how could they be improved? On the basis of detailed discussions during the 2-day workshop, a set of recommendations was developed using these important tools to estimate wildlife exposures. Although there are numerous applications (e.g., natural resource damage assessment, land-use planning) for these models, here, we concentrated on their applications within the ecological risk assessment (ERA) process—from initial screening assessment through remediation—with respect to contaminated sites. The following summary of our discussions during the workshop is intended to encourage ecological risk assessors to make greater use of spatially explicit exposure models and to provide recommendations for increasing the acceptance and use of such models in ecological risk assessments.


Spatially explicit relationships have been considered in site assessments in many forms for many years (Freshman and Menzie 1996; Wickwire and Menzie 2003; Hope 2004; Gaines et al. 2005). Historically, ecological risk assessors have accounted for species-specific spatial requirements using the home range (area over which a species' activities occur, excluding migration) or foraging area (area over which food is sought (USEPA 1993). The traditional approach is to either assume that the entire site represents a species' “home range” or apply an area use factor (AUF) (USEPA 1997). The AUF adjusts exposure on the basis of how much time an individual spends on-site versus off-site and is thus a measure of relative exposure. Putting a small site in combination with a large home range wildlife receptor typically results in a very small AUF. Although this approach considers use of noncontaminated areas by wildlife, it tends to be nonspecific with respect to habitat suitability and has the appearance of being a somewhat arbitrary adjustment in the absence of a consideration of habitat suitability and/or availability. Additionally, erroneously applied differences between “lifetime” home range and daily home range can significantly affect the calculation of a daily exposure estimate. Despite its limitations and uncertainties, the AUF does represent an attempt to integrate spatial considerations into an assessment and is generally an improvement on simply assuming a site is, regardless of the availability of any quality habitat and of a receptor's foraging range, the only area within which a receptor can and does forage.

Approaches for incorporating habitat heterogeneity and preferences into wildlife exposure models vary from no specific accounting to a detailed characterization of the habitat at a user-defined spatial resolution. The designation of habitat versus no habitat is important in and of itself, but there are also gradations in quality among the different habitats in an area that, if properly represented, could improve both exposure estimates and the subsequent hazard assessment. There are habitat models that define habitat suitability for a given species in terms of a habitat suitability index (HIS) (USGS 2010). Including habitat variability and the preference-guided behaviors of receptors in an exposure model increases its realism and ensures that wildlife exposure across a site is influenced by habitat quality, and that no exposure occurs in areas devoid of suitable habitat.

In contrast to traditional approaches that evaluate wildlife exposures with no, or only a limited (e.g., AUF), consideration of space and time, spatially explicit wildlife exposure models integrate the spatial determinants of biological activity, physical habitat suitability, and/or chemical variability in various media. As a result, these exposure models capture aspects of spatial (and behavioral) variability typically absent from more traditional wildlife exposure assessments. Chemical concentrations, for example, can vary widely across a site, and not necessarily in any relation to habitat type or suitability. Habitat suitability varies as well, and in many cases, habitat type and suitability vary over much larger spatial scales than does chemical contamination in specific media (e.g., soil, sediment, surface water). Capturing the intersection of habitat, media-specific chemical contamination, and wildlife receptor activity is the goal of exposure modeling. Spatially explicit exposure models increase the realism of that modeling by increasing the resolution of the calculation. By defining the spatial grid over which exposure occurs to reflect variability in both chemical contamination and habitat suitability, the variability in the exposure of each individual receptor can be captured. Some models employ algorithms that emulate movements of wildlife across the landscape guided by habitat suitability. Each day, additional exposures occur and are recorded for each individual. Ultimately, an average or high-end exposure can be estimated, reflecting all of the daily exposures. Additionally, many individuals or a virtual (statistical) population may be included in the modeled exposure estimate for the additional benefit of capturing the variability of exposure and, subsequently, of risk. These models offer assessors the opportunity to explore how known behaviors overlay with available habitat and media-specific chemical concentrations to better understand variability in the system. Ultimately, exposure estimates from these models are not estimated based on a single areal average, but rather as a statistic based on numerous model runs representing many individuals over time and space.

Value of these models

Although the goal of this communication is not to review or recommend any spatially explicit exposure model in particular, there is value in understanding how spatial considerations have been incorporated into wildlife exposure assessments historically and the variety of such models that are currently available (Loos et al. 2010). These models can be used to better inform risk assessors and risk managers at all stages of a site remediation project, including:

Problem formulation

Often boundaries and operational units are established based on non-ecological factors, such as hydrogeology, property lines, or contaminant sources or distribution. Although these are important considerations, habitat suitability should be considered as an important scaling factor overlaid on these other site demarcations. Spatially explicit models, when used early in the assessment planning stages, can help focus the assessment on areas where contamination is most likely to intersect with wildlife habitat and foraging ranges, thus minimizing exposure estimates that are unrealistically improbable. By considering the importance of habitat early in the assessment process, its evaluation can be included in the field sampling plan, evaluated during the larger field program, and, ultimately, the remedial plan.

Risk analysis and estimation

Workshop participants generally agreed that these models are more consistent with baseline (definitive) rather than screening assessments given the relative high investment in model parameterization and operation. However, during both screening (hazard) assessments and baseline (definitive) risk assessments, spatially explicit models can be employed in some capacity to determine what, if any, species might be subject to site-related exposures, and what areas of a site are likely to be the most problematic. For example, the Spatially Explicit Exposure Model (SEEM) was used effectively to evaluate several bird species with different life histories, and all of management concern, at the Eureka Mills Superfund Site in Utah (USFWS 2009). Logical recommendations regarding risk management at the site were made based on variability in the results in combination with considerations regarding land use and the spatial extent and nature of habitat and chemical contamination. Spatially explicit models allow risk assessors to: 1) Generate more realistic exposure estimates accounting for variability in habitat; 2) Account for species-specific exposure across a heterogeneous landscape by integrating actual wildlife behaviors, thereby capturing a perspective lost when relying only on site-wide, average- or maximum-based risk estimates; 3) Avoid misleading results from use of a site-wide, average-based risk estimate which might propagate errors such as remediation focused in areas where habitat is insufficient to even support wildlife; and 4) Efficiently suggest a protective risk management solution by both extending the context to consider exposures at ecologically relevant scales rather than operational or legal scales and by narrowing the focus of an analysis to the species that are most sensitive and susceptible based on the presence and orientation of favorable habitats.

Feasibility study and/or remedial planning

When a remediation project reaches the feasibility study and/or remedial planning stage, spatially explicit exposure models can be used to assess how different patterns of remedial activities could influence exposure and risk. Through iterative modeling, risk managers can arrive at a remedial solution that balances habitat loss (if any) against needed reductions in chemical concentrations. As there is no necessary correlation between static chemical concentrations in environmental media and habitat, it may be possible to craft remedies that do not significantly impact wildlife habitat. Remedial activities may also have direct costs to wildlife and can result in adverse impacts to specific species. Although there is an historical tendency to choose the most protective solution, the environmental benefits of appropriate actions that are based on more accurate prediction of exposures and risk should achieve the desired level of protection more efficiently. The role of a spatially explicit model is not to provide an absolute answer, but rather to capture a range of possible answers to serve as the basis for informed decision making.

Risk communication

Maps are traditionally an excellent medium for communicating complex assessment results, and spatially explicit models typically include map-based features. As a risk communication tool, spatially explicit exposure models can provide diverse groups of interested parties (i.e., stakeholders) with important data in a form that can be readily grasped and acted upon. Decisions regarding which areas to clean up, how much, and why can be facilitated and expedited when the options are presented visually. Ten acres of prime habitat may have no relevance (or too much relevance) until considered within the larger context of the surrounding landscape and the spatial location of possible remedial actions. Mapping tools within the models can also be used to forecast and display risk based on changing site conditions.

Examples of available models

A number of currently available models provide risk assessors with varying capabilities for assessing wildlife exposure to site-related chemicals given a consideration of habitat quality and availability. Table 1 provides an overview and comparison of a subset of available wildlife exposure models. These models have different development histories, and although many were constructed for a specific purpose or project, many are flexible enough to potentially be useful (as is or modified) for other applications. Most of the currently available wildlife models include some sort of movement of the individuals across the landscape, guided by a set of user-defined, and, where possible, species-specific movement rules (Hope 2005; Loos et al. 2010). Otherwise these models vary in terms of the exposure media considered (water, soil, sediment, etc.), the type of output offered (deterministic [point estimates] or probabilistic [distributions]), and how space is represented for landscape characterization (as a fixed grid of cells or pixel-based, which provides greater flexibility in terms of scale characterization). Some of these models are currently being used for regulatory purposes. For example, the Animal Landscape and Man Simulation System (ALMaSS), a probabilistic individual-based population model in which the individuals move around the defined landscape based on user assumptions and species-specific behaviors (Topping et al. 2003; Sibly et al. 2005; Dalkvist et al. 2009), has been used in Europe to review the registration of agrochemicals and the influence of new government policies. Similarly, HexSim, which is an individual-based, spatially explicit model for evaluating terrestrial wildlife population dynamics and interactions, is currently under development by USEPA (Lawler and Schumaker 2004).

Table 1. Comparison of spatially explicit exposure models
Model name and descriptionKey features*Exposure estimatesEffect endpoint(s)Spatial representation
  • *

    Key Features:

    OM =  organism movement; PE =  population endpoint; P/D = probabilistic and/or deterministic; n/a = not applicable; NCS = nonchemical stressors.

ALMaSS (Animal Landscape and Man Simulation System)OM - Yes - based on assumptions and observed behaviorDaily Dose (mg/kg/d)Population levelRectangular grid
A landscape scale, spatially explicit, agent based simulation system for investigating the effect of changes in landscape structure and management on the population size and distribution of animalsPE - YesInternal concentration (mg/kg)(abundance, growth rate, persistence, spatial distribution)1 cell = 1 m2
Developer - C Topping et al. 2003P/D - Probabilistic  Area = 100 km2
 NCS - Food availability, starvation, human disturbance(e.g. plowing, mowing)  104 X 104 cells
    Mapping: Spatial characterization within the model on user-supplied base map
RSEM (Resource Selection Exposure Model)OM - NoDaily Dose (mg/kg/d)Individual levelHexagonal grid
A GIS-based model for predicting exposure of midsized wildlife species to soil contaminationPE - Multiple individuals of a statistical population (compare to LOAEL) with compilation of results for population1 hexagon = 7.8 ha

Developer - Chow et al. 2005

; Gaines et al. 2005

P/D - Probabilistic  Area = 778 km2
 NCS - No  100 × 100 cells
    Mapping: Interacts with independent GIS layers
SE4M (Spatially and bioEnergetically Explicit terrestrial Ecological Exposure Model)OM - Yes, based on behavior described in the literatureInternal concentration (mg/kg)Individual levelRectangular grid
A spatially explicit, random walk model for exploring the influence of spatial and bioenergetic factors on a receptor's acquisition of energy and contaminant tissue residuesPE - No, individual based - simulation (tissue residues and energy balances)1 cell = 0.1 ha
Developer - Hope 2001, 2005P/D - Probabilistic  Area = 1.69 ha
 NCS - Food & habitat availability  13 × 13 cells
    Mapping: Spatial characterization within the model on user-supplied base map
SpaCE (Spatially Explicit Cumulative Exposure Model)OM - Yes, based on assumptions and behavior described in the literatureAverage concentration in food (mg/kg)Individual levelRectangular grid
A spatially explicit, random walk model for assessing dietary exposure of terrestrial vertebrates to cumulative chemical stressorsPE - No, individual based - simulationInternal concentration (mg/kg)(comparison with NOECs)1 cell = 25 m2

Developer - Loos et al. 2006; Schipper et al. 2008

(see also Eco-Space: Loos et al. 2009)

P/D - n/a  Area = 5.6 km2
 NCS - No  913 × 247 cells
    Mapping: Spatial characterization within the model on user-supplied base map
WBM (Wading Bird Model)OM - Yes, based on assumptions and on observed behaviorInternal concentration (mg/kg)Individual levelRectangular grid
A model to assess dietary contaminant exposure of interacting individuals of a wading bird colonyPE - Yes (foraging efficiency, reproduction success)1 cell = 6.25 ha

Developer - Wolff 1994

; Matsinos and Wolff 2003

P/D - n/a  Area = 1600 km2
 NCS - Food availability Population level160 × 160 cells
   (colony survival) 
    Mapping: Spatial characterization within the model
SEEM (Spatially Explicit Exposure Model)OM - Yes - based on 2 general foraging strategies described in the literature and implemented as movement rulesDaily dose (mg/kg/d)Statistical population levelFixed 25 × 25 cell grid
A spatially explicit, rule-based foraging model for assessing dietary exposure of terrestrial vertebrates to cumulative chemical stressorsPE - Yes, but depends on assumptions and approach; can be a statistical populationModel period dose statistics(LOAEL/NOAEL based comparisonsUser defined polygons are drawn for site chemistry and habitat suitability
Developer - US Army and ExponentP/D - Probabilistic, using 1-D Monte Carlo  Polygon data are translated into a fixed grid for calculations using area-weighted averaging to arrive at a value for each cell
 NCS - No   
    Mapping: Spatial characterization within the model on user-supplied base map
FR-M (FishRand-Migration)OM - Yes; probabilistic framework that “reseats” fish with each simulationInternal concentration (mg/kg)Individual levelUser defined
A spatially explicit dynamic aquatic bioaccumulation model (using 3-D Monte Carlo)PE - Yes (tissue concentration only) 
Developer - US Army Corps of EngineersP/D - Probabilistic, using 3-D Monte Carlo  Mapping: Spatial characterization within the model on user-supplied base map
 NCS - No   
QEAFDCHNOM - Yes, but not explicitly spatialInternal concentration (mg/kg)Individual level (body burdens)User defined
A spatially explicit, time-varying model for fish migrationPE - No  Mapping: Internal calculations only
Developer - QEA (now AnchorQEA)P/D - Deterministic   
 NCS - No   
HexSim (formerly PATCH)OM - Yes - based on assumptionsNonePopulation levelUser-defined hexangular grid
A spatially explicit, individual-based, multispecies model designed for simulating terrestrial wildlife population dynamics and interactionsPE - Yes (abundance & distribution)Mapping: Interacts with independent GIS layers
 P/D - Probabilistic   
Developer - US EPA ( - Habitat quality   
RAMAS GISOM - Yes - based on assumptionsNonePopulation level (abundance & distribution)User-defined rectangular grid
A metapopulation modeling platform for exposure analysis, population viability analysis and extinction risk assessmentPE - Yes   
Developer - Applied Biomathematics ( - Probabilistic  Mapping: Links directly to user-supplied GIS software
 NCS - Habitat quality & location   
3MRA (Multimedia, Multipathway, Multireceptor Exposure and Risk Assessment)OM - No (?)Daily dose (mg/kg-day)Statistical population levelRectangular grid
A system of models for conducting screening-level risk-based assessments of potential human and ecological health risks resulting from chronic exposure to chemicals released from land-based waste management units    
 PE - Yes (HQs based on mortality, growth, survival, reproductive success)100 × 100 m cells
Developer - US EPA Center for Exposure Assessment Modeling (CEAM) ( - Deterministic (probabilistic in FRAMES)   
 NCS - Habitat influences exposure  Mapping: Uses external GIS layers within an internal GIS
AQUATOXOM - Yes, pre-specified by userExternal exposureA large variety of both individual and population level endpointsLinked Segments
An ecosystem simulation model for aquatic systems which predicts the fate of various pollutants, such as nutrients and organic chemicals, and their effects on the ecosystem, including fish, invertebrates, and aquatic plantsPE - Yes (as biomass)Internal dose Thermal stratification
Developer - US EPA ( - Both options available  Flexible scale
 NCS - Yes (DO, suspended & bedded sediment)   
    Mapping: No mapping component, however the program can pull data from BASINS, EPA's GIS and water quality modeling system
MEERC modelsOM - n/aWater circulationEvaluate how spatially homogeneous vs heterogeneous ecosystems respond differently to perturbations (simulating mesocosms), and how ecological effects differ from pulse versus press predation by schooling fish 
Individual-based methods to simulate spatial movements of fish schools through the water columnPE - n/aNutrients  
Developer - Multiscale Experimental Ecosystem Research Center ( - n/a   
 NCS - n/a   
SADA (Spatial Analysis and Decision Assistance)OM - NoDaily Dose (mg/kg/d)Individual level (benchmark screening)User-defined
integrated modules for visualization, geospatial analysis, statistical analysis, human health risk assessment, ecological risk assessment, cost/benefit analysis, sampling design, and decision analysisPE - No (based on benchmark screening of individuals)  Includes a geospatial estimator (e.g. kriging)
Developer - University of Tennessee (∼sada/index.shtml)P/D - n/a   
 NCS - n/a  Mapping: Includes internal mapping tool and also accepts externally generated GIS layers

Following are 2 examples of how spatial model outputs might benefit risk assessors and risk managers.

Example 1: Deterministic versus probabilistic approaches

This example compares exposure estimates for songbird species exposed to Pb at 2 small arms ranges made with a deterministic model and site-wide statistics to those generated with SEEM, a spatially explicit wildlife exposure model with probabilistic outputs (Johnson et al. 2007). Avian dietary exposure was estimated with SEEM, which considers the spatial relationships of habitat and receptors, and a deterministic point estimate method with no spatial or habitat considerations. Exposure criteria used for each species were identical for each model. Exposure estimates from both models were compared to a Pb dietary dose-based toxicity reference value (TRV) using a hazard quotient methodology. These results were then compared to a site-specific risk estimate developed from blood Pb data collected on site. The investigators concluded that SEEM modeling results were more closely aligned with the risk estimate generated from the directly measured blood Pb concentrations (Figure 1). SEEM also made a daily maximum calculation for each individual to ensure that acute thresholds are not exceeded. In contrast, the conventional deterministic risk estimates were significantly higher than both the spatially explicit and directly measured risk estimates (Johnson et al. 2007). An additional benefit is that the SEEM can be summarized as a probability of exceeding a TRV, based on Monte Carlo–generated means for each individual over time (Figure 2). This probability distribution is more than a simple hazard classification and is the type of information consistent with a risk assessment (Tannenbaum et al. 2003).

Figure 1.

Comparison of hazard quotients determined using measured values, a deterministic reasonable maximum exposure (RME) method, and a spatially explicit exposure model (SEEM). Note that SEEM produces estimates more aligned with measured values than does the deterministic RME method (adapted from Johnson et al. 2007).

Figure 2.

Output from the SEEM model showing the percentage of a statistical population experiencing a given mean hazard quotient, based on a lowest-observed-adverse-effect-level (LOAEL) toxicity reference value (TRV). An increasingly smaller percentage of the population experiences increasingly larger hazard quotients (adapted from Johnson et al. 2007).

Example 2: Risk management decision making

Because spatially explicit models require users to collect spatially specific chemical and habitat suitability data, these models can be redeployed during risk management to evaluate how different remedial options may influence risk to wildlife, due to changes in both chemical concentrations and habitat due to remedial actions, thus allowing risk management plans to be fine tuned to meet site-specific goals, including the protection of wildlife. For example, remedial managers may be considering different cleanup options. A brief description of possible options and the benefits and costs of each is provided in Table 2. Habitat use affects exposure estimates. Areas with higher contamination may not be seen as presenting a risk if not used by valued receptors. Conversely, areas of high habitat use combined with relatively high contamination may result in very high exposure and risk estimates. Virtual risk management of various remedial alternatives can be conducted and compared efficiently using these models and optimal solutions for remediation can be determined through iterative model runs.

Table 2. Summary of costs and benefits relative to various remedial options
Remedial optionBenefitsCosts
Site-Wide Removal: Remove all soils with a contaminant concentration greater than a specific preliminary remediation goal (PRG).Public perception – clean the entire siteNo consideration of habitat quality or replacement; potential unnecessary loss of habitat and indirect wildlife harm; high costs associated with remediation.
Hotspot Removal: Remediate a few hotspot areas. A reduced acreage of habitat removed, but some elevated locations of a chemical may remain on the site.More focused clean-up areas – potentially lower costHabitat quality not considered; potential unnecessary loss of habitat and indirect wildlife harm; high costs
Risk-Based Balanced Remediation: targeted soil remediation to ensure population protections while balancing conservation of habitat; arrive at the best protective plan through iterative risk management planning.Protect the wildlife population – save key habitat when possible; Arrive at a defensible approach using transparent cost-benefit analysis; visualize landscape changes and their influence before remediation begins.Public perception may be that pollution is not being fully addressed. Managers have the information to illustrate important choices and trade-offs between removal, habitat preservation, and habitat restoration.


Although workshop participants could not identify any specific regulatory restrictions on the application of spatially explicit wildlife exposure models, a number of impediments to the application of such models were identified. These include few precedents for their use, misguided perceptions as to their purpose, traditional regulatory practices, when such models are considered during the site assessment process, and specific technical concerns, including the quality of input data. Each impediment is discussed briefly below.

Few precedents for use

Though the models have been available, discussed, and updated for years, there are few examples of their application in risk assessments for regulatory purposes. Although the current use of AUFs illustrates that spatial factors are being given consideration in some risk assessments, attempts to use exposure estimates based on more detailed, and possibly more realistic, characterizations of habitat suitability have largely been met with concern. Recently, however, SEEM was used with success at the Eureka Mills Superfund site in Utah (USFWS 2009). This former mining site was divided into 8 exposure areas, varying from 15 to 69 ha in size, and 23 unique exposure and/or habitat areas were defined. To support the spatially explicit calculations employed by SEEM, Thiesson polygons were used to assign surface soil concentration values to every sample collection point on the site. Exposure profiles for 5 songbird species were developed and assessed. The results corroborated life history attributes and were used in a weight-of-evidence approach to characterize risk.

Misperceptions as to purpose

Spatially explicit exposure models can be perceived, incorrectly, as simply a means to “dilute” exposure estimates so that subsequent risk estimates are lower and less “protective” remedial options are favored. However, simply basing remedial decisions on a protective, rather than a predictive, model may not result in actual protection of wildlife populations. Active remediation (e.g., soil excavation with attendant removal of overlying vegetation [i.e., habitat], dredging of sediment) comes at a cost, specifically in the reduction in habitat use during restoration, which can be especially significant to threatened and endangered species where habitat is the often the primary regulating factor (Wilcove et al. 1998). Other misperceptions include believing that modeling cannot possibly provide useful outputs because it requires so little investment of time, data, or resources or that using a model will lead to protracted and burdensome disputes about parameter values, disputes that will slow progress toward actual remediation.

Regulatory practices

Few precedents for use, in combination with these misperceptions, has made regulators hesitant to specifically support use of spatially explicit models. In addition, there is an expectation on the part of many regulators that risk assessments will, for simplicity and consistency with common practice, incorporate default, nonspatial, protective (reasonable maximum), and nonprobabilistic approaches. Although they predate recent advancements in spatially explicit models, current guidance documents do not specifically preclude or endorse the use of spatially explicit models but instead tend to focus on spatially neutral approaches. Larger framework statements about risk assessment do recognize the need for more flexible and realistic exposure tools, and some risk assessments have given limited consideration to spatial factors through the use of AUFs. However, as mentioned previously, approaches that realistically capture the variability in both habitat and chemical distribution are rarely used or discussed.

Timing of inclusion in the assessment process

Although spatially explicit methods are not suitable for every risk assessment, they can add value to baseline (or definitive) assessments at larger, more complex sites, particularly those involving highly valued habitats and ecological receptors (e.g., critical habitat, endangered species). They can be of assistance only if their role and data needs are given consideration early in the assessment process, and ideally during planning or problem formulation phases. It is extremely hard to retrofit the data needs of a spatially explicit model into an assessment once the project plan has been finalized and field work has commenced in a nonspatially explicit manner. Therefore, during the initial stages of an assessment, preliminary data should be collected that may allow spatially explicit exposure models to be of value to the baseline or definitive assessment. Following identification of assessment and measurement endpoints, specific and pertinent data can be collected that will help delineate habitat suitability as well as assist in better characterizing the extent of chemical contamination. Often, these data are not extensive nor is obtaining them particularly burdensome—habitat suitability indices typically require only 2 to 4 data for their calculation. Often these data can be gleaned from existing data: Natural Resource Management Plans, aerial photography, or limited field surveys.

Technical concerns

Although use of spatially explicit models can raise a legitimate number of technical and data quality concerns, these concerns should not be used as excuses for not employing such models where they might otherwise add value to an assessment. Some typical technical concerns and means for their possible resolution are outlined below.

Determination of suitability

Habitat suitability is an important component of many spatial models, and its determination can be contentious, particularly when a species-specific HSI model does not already exist. Although suitability is most effectively determined by an ecologist or by natural resource personnel who have worked with the receptor and know its habitat requirements well, a consensus may be also reached through course habitat suitability assignments (e.g., high, medium, low) based on species presence data.

Assessment population size

The size (as the number of model iterations) of the assessment (statistical) population can affect the shape of the distribution of output values, with the curve becoming “smoother” as more and more individuals are modeled. An inclination to use the most accurate population size often results in a “saw-toothed” curve, which is hard to interpret. One solution is to use a high estimate of population size combined with a series of iterations to achieve the best exposure estimate for the individuals of the population.

Assessment area

Reducing the assessment area to include only “hot spots” (small areas with very high contaminant concentrations) may reduce the population size to a few (perhaps unrealistically few) individuals. Conversely, expanding the assessment area to include substantial areas with habitat but without contamination is often criticized as an attempt to “dilute” exposure estimates. Workshop participants agreed that species-specific life history criteria combined with relevant population information could be used to select an assessment area. That is, the area of exposure should be determined after careful consideration of the habitat available for utilization by a population during an exposure period (e.g., a season). Assessors are advised to take receptor's requirements into consideration and consult with stakeholders to determine an optimal assessment area (stakeholder involvement may be needed to allow for data collection on adjoining private lands).


Balance complexity with accessibility

As many model developers know, finding the proper balance between a feature-rich model, user accessibility, and transparency for decision makers is one of the more challenging steps in the model design process. Users may include risk assessors, environmental engineers, and biologists. Decision makers may include risk managers and policy makers and, by extension, the constituencies to which they respond. Users must be able to easily describe to decision makers and stakeholders the features of the model and how it works—that it is not simply a black box. The optimal choice generally is a model that is no more complicated than necessary to inform the regulatory decision. The challenge for designers is how to achieve the stated goals in a transparent, accessible model. Selection of a model and its features requires, for example, a well-defined objective, an understanding of technical possibilities, close communication with the user community, an understanding of monetary constraints, and a distribution method. Close and frequent communication with the user community and decision makers is necessary to ensure that an appropriate balance of features and complexity is achieved as these improvements are included.

Enhanced guidance with examples and case studies

Even if the design of the model is intuitive, a user-friendly, clear, and comprehensive guidance manual provides a foundation for new users. Step-by-step instructions are important, but examples of applications and model outputs, along with well-documented case studies, will likely be the most encouraging and useful to new users. Case studies offer the opportunity to state a specific question, illustrate how the model is setup to answer that question, demonstrate how data are entered and results generated, and how the results were interpreted for and used by risk managers. The tendency to create a guidance document early on in the process with little attention to the application and with no plan for guidance updates leads to models that are applied by the developers but few others. Guidance documents can be considered living documents, and developers should consider updating them on a routine basis, including the incorporation of user input. One of the most frequently mentioned concerns about spatially explicit models is that they are “fixed” to produce lower estimates of exposure and risk than are deterministic (point estimate) models. There is no better way to dispel this mistaken belief and convey the value of a model than to provide examples of real-world applications. Case study examples might include comparing an example of an analysis made with data collected while considering habitat to one made with data collected only from chemical hot spots. The value of comparative case studies is their being able to demonstrate why results differ between different approaches to risk estimation. By presenting detailed examples, potential new users can see what data and assumptions underlie the differing estimates. Case studies can also be used to illustrate how iterative model runs can be used to focus on the most effective and efficient remedial solution.

Emphasize earlier consideration in the process

One of the historic limitations on the use of spatially explicit models has been the timing of their introduction into the assessment process. If they are introduced to a project at all, it is too often after study areas within the site (e.g., operable units) have been delineated without consideration of habitat. It is often unclear at a site a priori how habitat availability and chemical concentrations interacts might influence wildlife exposures—the highest chemical concentrations may not be colocated with habitat, rendering wildlife exposures negligible or, conversely, highly attractive habitat could harbor sufficient contamination to pose a problem with prolonged exposure. Earlier consideration of a spatially explicit model's data needs may provide insights into the need for specific sampling in order to understand the potential exposures. Identifying these needs early in the assessment process allows for coordination with other sampling that is planned for the site. It can be prohibitively resource intensive to remobilize a project's workforce simply to gather habitat data. If, however, these models were introduced at the beginning of the assessment process, during the planning and problem formulation phases, then study areas could be demarcated so as to ensure collection, congruent with collection of all other site-specific data, of habitat data supportive of a spatially explicit assessment, thereby avoiding the impediment of remobilization costs.

Increase interactions with the regulatory community

Increased appropriate use of spatially explicit models would be greatly encouraged if the results from such models were shown to be useful to and accepted by regulatory decision makers. Even if such models are not specifically recognized in existing regulatory guidance, their use by regulators to solve specific problems would likely lead to their use within the wider environmental community. Working collaboratively with the regulatory community to develop new guidance would also help align model products with the expectations of the various agencies who may ultimately provide oversight at sites seeking to use spatially explicit models. Increasing opportunities for hands-on training with direct access to the models, as well as including regulators on peer-review panels, may also model visibility and assist in understanding needed improvements in modeling features, approaches, and assumptions. External, independent reviews are another way to increase awareness within the regulatory community of models and their uses. The Center for Exposure Assessment Modeling (USEPA 2010) offers an existing option for formalized review and distribution through appropriate regulatory agencies.

Expanded communication with risk community

Developers of spatially explicit models need to identify and employ consistent communication channels to alert risk assessors to the availability of these models and any updates to them, as well as to receive feedback from users as to model performance and applications. Some avenues for communication between developers and users are briefly discussed below. More than one of these channels may need to be used on a consistent basis if effective bilateral communications are to occur. New media such as Web site blogs and Listservs also provide the opportunity to more directly communicate with users and encourage new applications.


Publications are the traditional platform for presenting technical information about new models, the results of research using them, as well as case studies using model results. Publications can, depending on the nature and extent of their reader base, be important tools for reaching a large and varied groups, encompassing both users and decision makers. A publication series focused on different spatially explicit models and their applications would help increase the visibility of these tools and provide a focal point for comparisons of functionality and features.


Conferences offer a venue where it is possible to reach a large number of people at one time, present the latest findings and developments, obtain real-time feedback, and offer hands-on demonstrations. Conferences should, ideally, include a diversity of environmental practitioners from regulators to consultants, researchers and industrial representatives. The opportunity for real-time interactions among a number of different practitioners and decision makers allows for valuable feedback while expanding the number of risk assessors who are comfortable with the models.

Other recommendations

Workshop participants discussed and generally agreed upon a number of other ideas for improving the utility and utilization of spatially explicit models, including (in no specific order): 1) Providing linkages to conventional, preferably widely available GIS applications, because more sophisticated geospatial tools and geostatistical approaches may be better at delineating the extent of both contamination and habitats; 2) Ability to easily adjust exposure calculation algorithms; 3) Polygon-specific adjustments to site-specific soil bioavailability values; 4) Use bioaccumulation regression equations to estimate food uptake; 5) Adding interactions between individuals and predator–prey relationships for greater realism; 6) Including life history parameters for various life stages; 7) Harmonizing output from these models with those from human health assessments, because some spatially explicit models provide results that are directly (e.g., FishRand-Migration estimates fish body burdens for human consumption) and indirectly (early indications of stressed sentinel populations) applicable to human health; and 8) Further research is needed to corroborate model estimates with field measures.


Workshop participants reached a consensus on spatially explicit wildlife exposure models being an important tool for increasing the predictive power of ecological risk assessments, and for improving the effectiveness and efficiency of risk management decisions. The participants also agreed that developers have not yet succeeded in fostering widespread acceptance and application of these models. By increasing model visibility in both the regulatory and risk assessment communities, opportunities will be identified, and the existing models can be more directly tuned to meet users' needs and expectations. Ultimately, as more sites use these models, they will gain greater acceptance. Additional research aimed at corroborating model results with field observations will be key to user acceptance, allowing for these important tools to gain greater application in baseline ecological risk assessments. The models are valuable throughout the different stages in the assessment process, but increasing the acceptance and use of these models will require a concerted effort by developers and regulators. Ultimately, the goal of increasing the realism of ecological assessments can be attained with a balanced and thoughtful integration of spatially explicit wildlife models.


The authors wish to recognize the support of the Department of Defense (DoD) Environmental Security Technology Certification Program (ESTCP) for supporting this work. The constructive comments of 3 anonymous reviewers are also appreciated. The following experts participated in the workshop that served as the basis for the summary and recommendations presented in this publication: Johnathan Clough, Anne Fairbrother, Anita Meyer, Larry Kapustka, John Kern, Igor Linkov, Charles Menzie, Wayne Munns, Drew Rak, Anne Rea, Randy Ryti, Brad Sample, and Katherine von Stackelberg. All views or opinions expressed herein are solely those of the authors and do not necessarily represent the policy or guidance of any other public or private entity. No official endorsement is implied or to be inferred.