possible fit of model predictions to observed data.
Calibration is sometimes referred to as “fitting” or
“tuning” a model.
Compartment model–A model that simulates pools of
biomass, concentration, numbers, or energy content
aggregated across populations or communities
rather than of individuals.
Conceptual (or descriptive) model–A diagrammatic
representation of the feedbacks and relationships
operating within a system; these models
synthesize existing understanding and enable
one to make qualitative predictions as well
as decisions regarding numerical model
formulation.
Deterministic model–A model that uses exact values
for all parameters (no uncertainty) and produces
a single prediction for each state variable at each
point in time.
Domain–The spatial extent of the model, i.e., the area
or volume being modeled.
Dynamic model–A model that produces predictions
that vary over time.
Empirical model–A model based on statistical
relationships between variables of interest
rather than theoretical or mechanistic relationships.
Forcing functions–Exogenous variables that impact
the system but are not explicitly modeled; instead,
they are input or “forced” into the model.
Formulation–(Verb) The process by which model
equations are developed (as in Figure 3) or (noun)
a model equation.
Individual-based (or agent-based) model (IBM)–A
model that simulates the unique behavior of
individuals (or agents) in a population; IBMs apply
the perspective that ecosystem patterns emerge as
a product of the interactions of individuals with each
other and their environment.
Initial conditions–The starting values used for each
state variable at the beginning of a model run.
Mechanistic model–A model that formulates processes
as the combined result of physiological,
biogeochemical, physical, and/or behavioral
mechanisms which are typically functions of
environmental or model state variables.
Model–A conceptual or mathematical simplification
(or abstraction) of a real system.
Model currency–The units used to simulate state
variables and rate processes in the model (e.g.,
g C m
À3
, g C m
À3 d
À1 ).
Optimization–An automated process in which
parameters are adjusted to achieve the best possible
fit to observations.
Parameterization–The process in which values or
distributions are chosen for model parameters,
typically from the literature, previous models, and
field and/or experimental data.
Parameters–Components (e.g., intercepts, slopes) of
both empirical and mechanistic model formulations
which can be fixed (deterministic models) or
allowed to vary (stochastic models).
Sensitivity analysis–The process by which individual
parameters of a model are changed by some amount
(e.g., Æ10 % or within known ranges) and the effect
on model predictions (e.g., % change in predicted
state variables) is quantified. Sensitivity analysis
can also be performed on initial conditions,
boundary conditions, and forcing functions
(although modification of forcing functions is
typically referred to as simulation analysis).
Simulation (or numerical) model–A model with
equation(s) that do not have an exact solution and
therefore must be solved over successive time steps,
a process called numerical iteration.
Simulation (or scenario) analysis–The use of a model
to understand system structure and function; this
often includes “what if?” scenarios in which the
model is used to predict system response to changes
in an important parameter or forcing function.
Skill assessment–Statistically quantifying the degree
to which the model reproduces the data (i.e., the
model-data misfit).
Spatial resolution–The degree to which space is
resolved by a model; models are typically 0D (single
point or box), 1D (e.g., boxes along the axis of
a system or layers through the water column), 2D
(e.g., boxes along the axis of a system with >1 depth
layer), or 3D (e.g., gridded models that resolve the x,
y, and z dimensions).
Standard run–The final, best-fitting predictions of
a model given the data available for calibration
which forms the basis for skill assessment,
sensitivity analysis, and simulation analysis.
State variables–Quantities within the system to be
modeled through time (e.g., biomass, concentration,
numbers, energy content).
Static model–A model that produces predictions
that are constant over time; however, these
models can be solved iteratively making them
quasi-dynamic.
Stochastic model–A model that incorporates
uncertainty in parameter values and produces
predictions with error distributions or a set of
possible model trajectories.
Temporal resolution–The unit of time for which rate
processes are resolved by a model.
Time step–The increment of time over which
successive iterations of a model are solved.
Verification (also called validation or confirmation)–
The process by which a model is tested against an
independent dataset (e.g., different year or system)
without changing the parameter values used in
calibration.
220
ECOLOGICAL MODELING
Calibration is sometimes referred to as “fitting” or
“tuning” a model.
Compartment model–A model that simulates pools of
biomass, concentration, numbers, or energy content
aggregated across populations or communities
rather than of individuals.
Conceptual (or descriptive) model–A diagrammatic
representation of the feedbacks and relationships
operating within a system; these models
synthesize existing understanding and enable
one to make qualitative predictions as well
as decisions regarding numerical model
formulation.
Deterministic model–A model that uses exact values
for all parameters (no uncertainty) and produces
a single prediction for each state variable at each
point in time.
Domain–The spatial extent of the model, i.e., the area
or volume being modeled.
Dynamic model–A model that produces predictions
that vary over time.
Empirical model–A model based on statistical
relationships between variables of interest
rather than theoretical or mechanistic relationships.
Forcing functions–Exogenous variables that impact
the system but are not explicitly modeled; instead,
they are input or “forced” into the model.
Formulation–(Verb) The process by which model
equations are developed (as in Figure 3) or (noun)
a model equation.
Individual-based (or agent-based) model (IBM)–A
model that simulates the unique behavior of
individuals (or agents) in a population; IBMs apply
the perspective that ecosystem patterns emerge as
a product of the interactions of individuals with each
other and their environment.
Initial conditions–The starting values used for each
state variable at the beginning of a model run.
Mechanistic model–A model that formulates processes
as the combined result of physiological,
biogeochemical, physical, and/or behavioral
mechanisms which are typically functions of
environmental or model state variables.
Model–A conceptual or mathematical simplification
(or abstraction) of a real system.
Model currency–The units used to simulate state
variables and rate processes in the model (e.g.,
g C m
À3
, g C m
À3 d
À1 ).
Optimization–An automated process in which
parameters are adjusted to achieve the best possible
fit to observations.
Parameterization–The process in which values or
distributions are chosen for model parameters,
typically from the literature, previous models, and
field and/or experimental data.
Parameters–Components (e.g., intercepts, slopes) of
both empirical and mechanistic model formulations
which can be fixed (deterministic models) or
allowed to vary (stochastic models).
Sensitivity analysis–The process by which individual
parameters of a model are changed by some amount
(e.g., Æ10 % or within known ranges) and the effect
on model predictions (e.g., % change in predicted
state variables) is quantified. Sensitivity analysis
can also be performed on initial conditions,
boundary conditions, and forcing functions
(although modification of forcing functions is
typically referred to as simulation analysis).
Simulation (or numerical) model–A model with
equation(s) that do not have an exact solution and
therefore must be solved over successive time steps,
a process called numerical iteration.
Simulation (or scenario) analysis–The use of a model
to understand system structure and function; this
often includes “what if?” scenarios in which the
model is used to predict system response to changes
in an important parameter or forcing function.
Skill assessment–Statistically quantifying the degree
to which the model reproduces the data (i.e., the
model-data misfit).
Spatial resolution–The degree to which space is
resolved by a model; models are typically 0D (single
point or box), 1D (e.g., boxes along the axis of
a system or layers through the water column), 2D
(e.g., boxes along the axis of a system with >1 depth
layer), or 3D (e.g., gridded models that resolve the x,
y, and z dimensions).
Standard run–The final, best-fitting predictions of
a model given the data available for calibration
which forms the basis for skill assessment,
sensitivity analysis, and simulation analysis.
State variables–Quantities within the system to be
modeled through time (e.g., biomass, concentration,
numbers, energy content).
Static model–A model that produces predictions
that are constant over time; however, these
models can be solved iteratively making them
quasi-dynamic.
Stochastic model–A model that incorporates
uncertainty in parameter values and produces
predictions with error distributions or a set of
possible model trajectories.
Temporal resolution–The unit of time for which rate
processes are resolved by a model.
Time step–The increment of time over which
successive iterations of a model are solved.
Verification (also called validation or confirmation)–
The process by which a model is tested against an
independent dataset (e.g., different year or system)
without changing the parameter values used in
calibration.
220
ECOLOGICAL MODELING