Ecological Modeling in Risk Assessment - Chapter 13 pdf

© 2002 by CRC Press LLC
CHAPTER 13
Profiles of Selected Models
Robert A. Pastorok
Table 13.1 summarizes the models selected for further development and use in ecological risk
assessment in the near future. These models were rated relatively high in the evaluation within
model type. Profiles of models selected for further development and use are shown in Tables 13.2
through 13.7.
Toxicity extrapolation models are intended mainly for use in screening-level risk assessments
and for developing generic environmental criteria (e.g., EPA’s ambient water quality criteria). They
are not recommended for use alone in a detailed (e.g., baseline) risk assessment. Nevertheless, they
may be useful in supporting population, ecosystem, or landscape models as part of a detailed
assessment.
Of the population models, stochastic scalar abundance models (either discrete or continuous-
time) and deterministic life-history matrix models are most appropriate for screening-level ecolog
-
ical risk assessments (Table 13.1). Some simple food-web models, founded on RAMAS Ecosystem
or Populus, for example, may be appropriate for screening-level assessments at larger, complex
sites, especially where disruption of food-web structure may be an issue. More complex ecosystem
models and landscape models are not recommended for screening-level assessments because of the
relatively high level of effort and expense involved in developing and running these models.
For detailed assessments, stochastic life-history matrix models and metapopulation models
(e.g., RAMAS GIS and VORTEX) are recommended. These models, as well as aquatic ecosystem
models like AQUATOX, CASM, and IFEM, aquatic landscape models like ATLSS, AQUATOX,
and CASM, and terrestrial landscape models like LANDIS, JABOWA, and the disturbance bioge
-
ography model, are suitable for application in detailed ecological risk assessments. Several model
categories lack specific examples of available models for detailed assessments. Further development
of such models could include integration of metapopulation models with food-web models and
other ecosystem models and with landscape models (Table 13.1).

The selection of specific models for addressing an ecological risk issue depends on the balance
between model complexity and the availability of data, the degree of site-specificity of available
models, and the issue, ecosystem, endpoints, and chemicals of interest. The models must be
appropriate for the context, whether for the evaluation of risks associated with new chemicals and
their uses, of ecological impacts and risks associated with past uses, or of clean-up and restoration
issues. Because the selection of models is specific to the issue, chemical, and site of interest, we
have provided only general guidance for selecting ecological models (see Chapter 1, Introduction,
1574CH13.fm Page 195 Tuesday, November 26, 2002 6:23 PM
© 2002 by CRC Press LLC
Table 13.1 Ecological-Effects Models Selected for Further Development and Use in Chemical Risk Assessment
Ecosystem Landscape
Toxicity-
Extrapolation
Level of
Effort
Population
General
Food-Web
Aquatic Terrestrial General Aquatic Terrestrial
Screening Stochastic scalar
abundance
Populus NA NA NA NA NA Interendpoint
interspecies
Life-history matrix
(deterministic)
RAMAS
Ecosystem
Species-
sensitivity

distribution
Detailed Life-history matrix
(stochastic)
Integration of
RAMAS
Ecosystem
with spatially
explicit
metapopu-
lation models
AQUATOX
CASM
IFEM
Integration of
spatially
explicit
metapopu-
lation and
food-web
models
Integration of
spatially
explicit
metapopu-
lation and
landscape
models
AT LS S
AQUATOX
CASM

IFEM
LANDIS
JABOWA
Island
disturbance
biogeo-
graphic
N/A
a
Metapopulation
RAMAS GIS
VORTEX
Note: N/A - not applicable.
a
Relationships for use with ecological models.
1574CH13.fm Page 196 Tuesday, November 26, 2002 6:23 PM
© 2002 by CRC Press LLC
The Process of Ecological Modeling for Chemical Risk Assessment). Moreover, the models that
we selected for further development and use are not necessarily the only models that would be
useful for ecological risk assessments in the near future.
Many of the models reviewed here are applicable for answering ecological risk questions. We
urge the reader to consider his/her needs, select a general category of models according to the endpoint
Table 13.2 Profiles of Selected Population Models — Scalar Abundance Models
Model: Stochastic Continuous-Time Models
References: Poole (1974), Goel and Richter-Dyn (1974), May (1974), Ginzburg et al. (1982),
Iwasa (1998), Tanaka (1998), Hakoyama and Iwasa (1999)
Stressor: Any stressor whose impacts can be summarized as a change in the mean or
variance of the population growth rate or density-dependence parameters
Type of ecosystem: Various (terrestrial, freshwater, estuarine, marine)
Temporal resolution: Determined by resolution of available census data and generation time of the

species; time step is usually between 1 week and 1 decade
Time scale: Determined by the modeler according to the question addressed, the quality
and amount of available population data (e.g., period during which the data
were collected), and generation length of the species; time period is usually
between several weeks and a century
Spatial resolution: Not spatially resolved
Geographic scale: Implicitly identified with the geographical scale of modeled population
Biological scale: Population
Model type: Simple scalar abundance model
Level of detail: Screening
Endpoints: Expected future population abundance; statistical distribution of abundance;
risk of a population decline of a specified size; time to a population decline
of a specified size
Comments: Has minimum data requirements for any population-level, risk-analytic model;
can include effects of density dependence or omit them in a conservative
assessment; can integrate effects of chemical and non-chemical stressors
through time; methodology can be used now but would benefit from further
development of allometric relationships to inform data parameterizations when
empirical information is sparse
Model: Stochastic Discrete-Time Models
References: Poole (1974), Capocelli and Ricciardi (1974), Goel and Richter-Dyn (1974),
May (1974), Ginzburg et al. (1982), Ferson (1999)
Stressor: Any stressor whose impacts can be summarized as a change in the mean or
variance of the population growth rate or density-dependence parameters
Type of ecosystem: Various (terrestrial, freshwater, estuarine, marine)
Temporal resolution: Determined by resolution of available census data and generation time of the
species; timestep is usually between 1 week and 1 decade
Time scale: Determined by the modeler according to the question addressed, the quality
and amount of available population data (e.g., period during which the data
were collected), and generation length of the species; time period is usually

between several weeks and a century
Spatial resolution: Not spatially resolved
Geographic scale: Implicitly identified with the geographical scale of the modeled population
Biological scale: Population
Model type: Simple scalar abundance model
Level of detail: Screening
Endpoints: Expected future population abundance; statistical distribution of abundance;
risk of a population decline of a specified size; time to a population decline
of a specified size
Comments: Has minimum data requirements for any population-level, risk-analytic model;
can include effects of density dependence or omit them in a conservative
assessment; can integrate effects of chemical and non-chemical stressors
through time; methodology can be used now but would benefit from further
development of allometric relationships to inform data parameterizations when
empirical information is sparse; easier both to implement in software and to
parameterize than continuous stochastic models
1574CH13.fm Page 197 Tuesday, November 26, 2002 6:23 PM
© 2002 by CRC Press LLC
desired, and review the ratings of individual models. For specific needs, the ratings may be differen-
tially weighted, and the best models may differ from the ones selected by us for further development.
Table 13.3 Profiles of Selected Population Models — Life-History Models
Model: General Matrix Models
Reference: Caswell (2001)
Stressor: Any (including chemical and thermal pollution, harvest, entrainment,
impingement, habitat loss, habitat fragmentation); depends on implementation
Type of ecosystem: Various (terrestrial, freshwater, estuarine, marine)
Temporal resolution: Theoretically any temporal resolution is allowed; practically, temporal resolution
of the model depends on temporal resolution of the data; generally ranges
from 1 day to 10 years
Spatial resolution: Not spatially explicit, but can be embedded within a metapopulation model

Geographic scale: Not spatially explicit
Biological scale: Population
Time scale: Theoretically any; practically, depends on temporal resolution and question
being asked (endpoint); generally ranges from several weeks to several
decades
Model type: Structured population
Level of detail: Tier 2
Endpoints: Depends on implementation (e.g., which software package is being used);
possibilities include expected future abundance, risk of decline in abundance,
risk of extinction, time to extinction, time to decline, harvested biomass, as
well as measures of uncertainty about these endpoints
Comments: Highly recommended as a general class of models; assumptions are realistic
and endpoints are comprehensive; implementation generally requires the use
of a software package or programming; matrix models have been embedded
in metapopulation models in which each subpopulation is modeled with a
matrix; new theory that will add to the utility of matrix models is the
development of risk-based sensitivity analysis
Model: RAMAS Age, Stage, Metapop, or Ecotoxicology
Reference: Applied Biomathematics (2000)
Stressor: Any stressor, including chemical and thermal pollution, harvest, entrainment,
impingement, habitat loss, habitat fragmentation
Type of ecosystem: Various (terrestrial, freshwater, estuarine, marine)
Temporal resolution: Theoretically any temporal resolution is allowed; practically, temporal resolution
of the model depends on temporal resolution of the data; generally ranges
from 1 day to 10 years
Spatial resolution: Not spatially explicit (but see comments)
Geographic scale: Not spatially explicit (but see comments)
Biological scale: Population
Time scale: Theoretically any; practically, depends on temporal resolution and question
being asked (endpoint); generally ranges from several weeks to several

decades
Model type: Structured population
Level of detail: Tier 2
Endpoints: Expected future abundance, risk of decline in abundance, risk of extinction,
time to extinction, time to decline, harvested biomass, and measures of
uncertainty around each endpoint
Comments: These programs comprehensively apply matrix models; Age and Stage are
only available in DOS, whereas Metapop and Ecotoxicology are Windows
®
-
based; only Ecotoxicology explicitly models the effects of toxic chemicals; the
other programs can be manipulated to model toxicant effects by running
several simulations (e.g., with and without toxic chemicals)
1574CH13.fm Page 198 Tuesday, November 26, 2002 6:23 PM
© 2002 by CRC Press LLC
Table 13.4 Profiles of Selected Population Models — Metapopulation Models
Model: RAMAS GIS (and Metapop)
Reference: Akçakaya (1998a,b); Applied Biomathematics (2000)
Stressor: Any stressor, including chemical and thermal pollution, harvest, entrainment,
impingement, habitat loss, habitat fragmentation
Type of ecosystem: Various (terrestrial, freshwater, estuarine, marine)
Temporal resolution: Variable; determined by resolution of available demographic data (e.g., census
frequency) and generation length of the species; typically between 1 week
and 1 decade
Spatial resolution: Variable; determined by resolution of available spatial data (e.g., size of raster
map cells) and spatial behavior of the species (e.g., home range size, daily
movement distance, etc.); typically between one meter and several kilometers
Geographic scale: Variable; determined by the range of the species, the question addressed, and
the availability of geographic data; typically between one hectare and
thousands of square kilometers

Biological scale: Population or metapopulation
Time scale: Variable; determined by the question addressed, the quality and amount of
available demographic data (e.g., period during which the data were
collected), and generation length of the species; typically between several
weeks and several centuries
Model type: Metapopulation
Level of detail: Tier 2
Endpoints: Expected future abundance, risk of decline in abundance, risk of extinction,
time to extinction, time to decline, harvested biomass
Comments: Has high realism, relevance, and flexibility and can be used for risk assessment
without additional development; however, further development that will likely
make it more suitable includes linking it to a GIS-based hydrodynamic model,
adding sex structure, genetics, and different types of density-dependence;
GIS/Metapop enables one to model several structured subpopulations that
together comprise a metapopulation
Model: VORTEX
Reference: Lacy (1993, 1999)
Stressor: Any stressor, including chemical and thermal pollution, harvest, entrainment,
impingement, habitat loss
Type of ecosystem: Various, but mostly terrestrial
Temporal resolution: Variable; determined by resolution of available demographic data (e.g., census
frequency) and generation length of the species; typically between 1 week
and 1 decade
Spatial resolution: Not spatially explicit; spatial structure based on predefined populations
Geographic scale: Variable; determined by the range of the populations modeled and the question
addressed
Biological scale: Population or metapopulation
Time scale: Variable; determined by the question addressed, the quality and amount of
available demographic data (e.g., period during which the data were
collected), and generation length of the species; typically between several

weeks and several centuries
Model type: Individual-based metapopulation
Level of detail: Tier 2
Endpoints: Expected future abundance, risk of extinction, time to extinction
Comments: Has medium level of realism (in comparison with other individual-based models
reviewed), high relevance, and high resource efficiency; parameters in the
model can be estimated with relative ease; may not be suitable for plants,
highly fecund species, and very abundant species; does not incorporate
habitat relationships
1574CH13.fm Page 199 Tuesday, November 26, 2002 6:23 PM
© 2002 by CRC Press LLC
Table 13.5 Profiles of Selected Ecosystem Models — Food-Web Models
Model: RAMAS Ecosystem
Reference: Spencer and Ferson (1997a); Applied Biomathematics (2000)
Stressor: Chemical, physical
Type of ecosystem: Various (terrestrial, freshwater, estuarine, marine)
Temporal resolution: Any temporal resolution is allowed by the program, but generally ranges from
weeks to years
Spatial resolution: Not spatially explicit
Geographic scale: Not spatially explicit
Biological scale: Food web, food chain, or simple predator–prey interaction
Time scale: Any time frame is allowed, but the general range is between several months
and several decades
Model type: Food web/food chain
Level of detail: Tier 2
Endpoints: Expected future abundance, risk of decline in abundance, risk of extinction,
time to extinction, time to decline, toxicant concentration in the environment,
toxicant concentration in the organisms
Comments: This program is quite flexible and easy to use; it explicitly models the effects
of toxic chemicals, and does not assume that populations are at equilibrium

(which is in its favor); several relevant endpoints are available; a generalization
to allow an intermediate between prey-dependent and ratio-dependent
functional responses would increase its utility
Model: Populus
Reference: Alstad et al. (1994a,b); Alstad (2001)
Stressor: Chemical, physical
Type of ecosystem: Various (terrestrial, freshwater, estuarine, marine)
Temporal resolution: Any temporal resolution is allowed by the program, but generally ranges from
weeks to years
Spatial resolution: Not spatially explicit
Geographic scale: Not spatially explicit
Biological scale: Food web, food chain, or simple predator–prey interaction
Time scale: Any time frame is allowed, but the general range is between several months
and several decades
Model type: Food-web/food-chain
Level of detail: Tier 2
Endpoints: Expected future abundance
Comments: Endpoints are limited, but the program is flexible in that any kind of
predator–prey model can be implemented, as equations can be defined by
the user; also offers several predefined predator–prey models; however,
Populus does not explicitly model the effects of toxic chemicals and has no
treatment of uncertainty
1574CH13.fm Page 200 Tuesday, November 26, 2002 6:23 PM
© 2002 by CRC Press LLC
Table 13.6 Profiles of Selected Ecosystem Models — Aquatic Ecosystem Models
Model: AQUATOX
Reference: Park et al. (1974); Park (1998); U.S. EPA (2000a,b,c)
Stressor: User-defined toxic chemicals
Type of ecosystem: Freshwater, estuarine
Temporal resolution: Days

Spatial resolution: Not applicable
a
Geographic scale: Not applicable
a
Biological scale: Guild
Time scale: User-defined
Model type: Multicompartment; linear additive integral
Level of detail: Tier 2
Endpoints: Guild biomass, body burdens of contaminants; specific fish type abundance
Comments: AQUATOX was specifically designed to model the ecological impact of chemical
contaminants on aquatic ecosystems; it not only models changes in
productivity and standing stock abundance but also the transfer of
contaminants between media and biota as well as through various trophic
levels; AQUATOX is supported by EPA and is available on a software platform
that is still undergoing revision and development
Model: CASM (Modified SWACOM)
Reference: DeAngelis et al. (1989); Bartell et al. (1992, 1999)
Stressor: Nutrients and user-defined toxic chemicals
Type of ecosystem: Freshwater, estuarine
Temporal resolution: User-defined periodicity
Spatial resolution: Not applicable
a
Geographic scale: User-defined; area treated as a homogeneous unit
a
Biological scale: Guild
Time scale: User-defined
Model type: Multicompartment; linear additive integral
Level of detail: Tier 2
Endpoints: Nutrient concentrations, productivity, abundance, and diversity of fish
Comments: This model is very similar to AQUATOX; unlike AQUATOX, SWACOM was

initially developed to model nutrient impacts
Model: IFEM
Reference: Bartell et al. (1988)
Stressor: User-defined; established for naphthalene
Type of ecosystem: Freshwater, estuarine
Temporal resolution: Days
Spatial resolution: Not spatially explicit
a
Geographic scale: Not applicable
a
Biological scale: Guild
Time scale: User-defined
Model type: Compartment dynamic; integrated linear additive
Level of detail: Tier 2
Endpoints: Guild productivity, toxic impact, guild-specific body burdens of contaminants
Comments: This model is an iterative simulation that describes food-web transfer for both
carbon and contaminants and estimates both primary and secondary impacts
of toxicity
a
Spatially aggregated model. May also be applied as a landscape model.
1574CH13.fm Page 201 Tuesday, November 26, 2002 6:23 PM
© 2002 by CRC Press LLC
Table 13.7 Profiles of Selected Landscape Models
Model: ATLSS
Reference: DeAngelis (1996)
Stressor: Developed for mercury, but may be modified for other contaminants
Type of ecosystem: Large-scale wetlands (Florida Everglades)
Temporal resolution: Variable; executed on user-defined cycles
Spatial resolution: Currently based on 100-m
2 gr

ids
Geographic scale: Variable; based on multiple grids
Biological scale: Variable; guild at lower trophic levels and species at upper trophic levels
Time scale: Open
Model type: Spatially explicit; low trophic levels are process models, middle trophic levels
are structured cohort models, and high trophic levels are individual-based
models
Level of detail: Tier 2
Endpoints: Abundance of guild or species; intertrophic carbon flow and populations
(abundance and diversity) of keystone species; mercury bioaccumulation
Comments: ATLSS is one of the most comprehensive landscape models available; its
ambition is balanced by its flexibility in that the creators recognized that
different trophic levels require different types of model formats; this allows
the simulations to be tailored not only based on the life histories of a particular
species but also on the available data and understanding of that receptor’s
behavior; although this model was originally developed for application to the
Florida Everglades, it could be used in other biomes where there is a terrestrial
food web reliant on the productivity of a large-scale aquatic habitat
Model: LANDIS
Reference: Mladenoff et al. (1996); Mladenoff and He (1999)
Stressor: Wind and fire disturbances; harvesting; any physical disturbance
Type of ecosystem: Northern temperate zone forests
Temporal resolution: Hundreds of years
Spatial resolution: 10 × 10 m
Geographic scale: Thousands of hectares
Biological scale: Tree species as the presence or absence of 10-year age cohorts (not as
individual stems)
Time scale: Hundreds of years
Model type: Stochastic, spatially explicit model of forest landscape processes; raster-based
Level of detail: Tier 2

Endpoints: Tree species presence/absence, stand age structure and species richness,
potential impact of toxic chemicals on species-specific life-history traits (e.g.,
mortality, seed dispersal, etc.)
Comments: LANDIS is an elaboration of the VAFS/LANDSIM model; however, LANDIS
operates in raster mode, and the spatial interactions are based on distances
(instead of polygon neighborhoods); operation can be at different scales of
resolution and is programmed in C++ by using hierarchical, object-oriented
data structures; LANDIS uses a free-standing spatial analysis package, reads
and writes ERDAS raster files, and incorporates wind-throw and fire
disturbance, which are absent from other models
1574CH13.fm Page 202 Tuesday, November 26, 2002 6:23 PM
© 2002 by CRC Press LLC
Table 13.7 (cont.)
Model: JABOWA
Reference: Botkin et al. (1972); West et al. (1981); Bodkin (1993a,b)
Stressor: Physical disturbance, primarily harvesting
Type of ecosystem: Multispecies forests throughout the world
Temporal resolution: Annual time step
Spatial resolution: 10-m × 10-m grid sections (default value, which is user-adjustable in the early
versions of the model)
Geographic scale: User-defined landscape
Biological scale: Individual trees of various species
Time scale: User-defined
Model type: Spatially explicit model of forest landscape with stochastic functions for
mortality and reproduction
Level of detail: Tier 2
Endpoints: Tree biomass, stand biomass, stand age structure, forest productivity, and
species richness based on growth, reproduction, and mortality of individual
trees; species and biomass distribution in space
Comments: JABOWA was developed as a general model of forest dynamics; it was among

the first successful multispecies computer simulations of terrestrial
ecosystems and has gone through extensive modification and use during the
30 years since it originated; one of the best-documented and most flexible
models of its type
Model: Island Disturbance Biogeographic Model
Reference: Villa et al. (1992)
Stressor: Variable; manifest as impacts to populations
Type of ecosystem: Various; model must be tailored by the user
Temporal resolution: Periodicity is user defined
Spatial resolution: Grid size and resolution are user defined
Geographic scale: Grid size and resolution are user defined
Biological scale: Communities across a landscape
Time scale: Periodicity is user defined
Model type: Spatially explicit; quantitative rule-based model based on iterative Lotka-
Volterra kinetics
Level of detail: Tier 2
Endpoints: Biodiversity and temporal species sensitivity
Comments: The Villa model uses the same principles as most island biogeographic models;
however, this implementation (1) is spatially explicit, (2) has implicit integration
of the disturbance function, (3) has multiple population interactions, and
(4) calculates immigration relative to island carrying capacities; this model,
as presented, is highly theoretical; this increases its flexibility but would require
explicit modifications to tailor it to a specific situation; it does not currently
have functions to explicitly model toxicity
1574CH13.fm Page 203 Tuesday, November 26, 2002 6:23 PM