23.5 Characterization of Dynamics and Evidence Used in Assessment
347
ing on interactions with other processes, and data
confined to one time period cannot be used to extrapolate conditions over other time periods.
23.5.3 Simulating Dynamic Patterns
Ecological modeling, or the explicit mathematical
simulation of physical and biological relationships,
has been used in numerous applications to estimate
and predict ecosystem processes. Modeling has
been used on a global scale to estimate the impacts
of climate change (Claussen and Esch, 1994; Dale
and Rauscher, 1994), to simulate biogeochemical
cycles (Parton et aI., 1988; Bonan, 1995), and to
estimate net primary production (Lieth, 1975; Running and Coughlan, 1988). It has also been used
extensively to model hydrologic processes (Beven
and Kirkby, 1979; Moore et aI., 1991), usually
through the use of raster, or cell-based, geographical information systems (GIS) and digital elevation models (Davis and Goetz, 1990; Drayton et aI.,
1992). Application to terrestrial dynamic patterns
has been limited largely to fire behavior (Kessell,
1979; Baker, 1994; Davis and Burrows, 1994; Vasconcelos et aI., 1994), insect outbreak (Gribko et
al., 1995), and the dynamics of vegetation change
(forest succession, or gap-phase, models) (Urban
and Shugart, 1992). This is likely due to the fact
that these disturbance types are largely endogenous
(a product of life-history traits in the case of forest
succession and insect outbreak) or controlled by
characteristics of the system (i.e., landform and interval since last fire in the case of fire behavior).
Exogenous disturbances, such as weather events,
flooding, earthquakes, and exotic invasions, are
nearly impossible to predict and therefore model
effectively.
Models are useful for projecting future landscape
structure and for investigating the impacts of a variety of management scenarios (Baker, 1992a). This
is usually perceived as their primary function.
However, models are also useful for testing the validity of assumptions made about a system. By using historic conditions as a starting point and comparing results to current conditions, spatial models
can reveal the extent and limits of our understanding. Models can provide estimates of historical
range of variability when historical data are incomplete or inadequate (Morgan et al., 1994).
Models can also be used to determine the circumstances under which a system might be considered to be in an equilibrium state, wherein the
proportion of the landscape in various successional
states remains fairly constant, although the location
of patches constantly shift. This basic definition of
eqUilibrium can be modified in highly dynamic
landscapes to include systems in which all successional stages are constantly present, although
highly variable in relative proportion (White and
Walker, 1997). Being able to model a patch dynamic equilibrium state for a particular landscape
is a significant route by which the natural, stable
disturbance regime can be identified. Several researchers have concluded that dynamic eqUilibrium
is a function of the relationship between the mean
size of disturbances in a stable disturbance regime
and the size of the area of interest (Shugart, 1984;
White, 1987). Thus simulation modeling can aid in
determining (1) the stable disturbance regime of a
region, (2) whether a particular study area is large
enough to capture this disturbance regime, and (3)
the appropriate temporal scale for assessing the disturbance regime. For example, in applying a spatially explicit, GIS-based fire model (REFIRES) to
a chaparral ecosystem in southern California, Davis
and Burrows (1994) found that the variation between long model runs with different initial conditions but the same parameter settings was only
slight. This suggests that, within regions with uniform climate, vegetation, and physiography and
over long periods of time, a fire-controlled chaparral system may approach an equilibrium state.
Such an assessment can aid in determining the important components in a system's natural disturbance regime, and the appropriate spatial and temporal scales for describing that disturbance regime.
Models range widely in their degree of specificity and spatial detail, from whole landscape
models, which estimate the value of a variable over
an entire landscape, to distributional models, which
estimate the distribution of values of some variables(s) over time, and finally to spatial landscape
models, in which the spatial location and configuration of a variable can be tied to any number of
landscape measures and submodels (Baker, 1989).
Models may be either continuous or discrete in their
treatment of time. Although spatial models are the
most detailed and informative, their computational
demands are prohibitive. Distributional models,
therefore, have been most widely developed and
applied, whereas whole landscape models are usually integrated as submodels. Many recent models
have been developed and implemented in a GIS environment (see Goodchild et aI., 1993, and HainesYoung et aI., 1993) and structured on a raster format, often using a hexagonal grid tessellation to
facilitate contagious diffusion among grid cells.
Such a format allows for easy integration with other
347
ing on interactions with other processes, and data
confined to one time period cannot be used to extrapolate conditions over other time periods.
23.5.3 Simulating Dynamic Patterns
Ecological modeling, or the explicit mathematical
simulation of physical and biological relationships,
has been used in numerous applications to estimate
and predict ecosystem processes. Modeling has
been used on a global scale to estimate the impacts
of climate change (Claussen and Esch, 1994; Dale
and Rauscher, 1994), to simulate biogeochemical
cycles (Parton et aI., 1988; Bonan, 1995), and to
estimate net primary production (Lieth, 1975; Running and Coughlan, 1988). It has also been used
extensively to model hydrologic processes (Beven
and Kirkby, 1979; Moore et aI., 1991), usually
through the use of raster, or cell-based, geographical information systems (GIS) and digital elevation models (Davis and Goetz, 1990; Drayton et aI.,
1992). Application to terrestrial dynamic patterns
has been limited largely to fire behavior (Kessell,
1979; Baker, 1994; Davis and Burrows, 1994; Vasconcelos et aI., 1994), insect outbreak (Gribko et
al., 1995), and the dynamics of vegetation change
(forest succession, or gap-phase, models) (Urban
and Shugart, 1992). This is likely due to the fact
that these disturbance types are largely endogenous
(a product of life-history traits in the case of forest
succession and insect outbreak) or controlled by
characteristics of the system (i.e., landform and interval since last fire in the case of fire behavior).
Exogenous disturbances, such as weather events,
flooding, earthquakes, and exotic invasions, are
nearly impossible to predict and therefore model
effectively.
Models are useful for projecting future landscape
structure and for investigating the impacts of a variety of management scenarios (Baker, 1992a). This
is usually perceived as their primary function.
However, models are also useful for testing the validity of assumptions made about a system. By using historic conditions as a starting point and comparing results to current conditions, spatial models
can reveal the extent and limits of our understanding. Models can provide estimates of historical
range of variability when historical data are incomplete or inadequate (Morgan et al., 1994).
Models can also be used to determine the circumstances under which a system might be considered to be in an equilibrium state, wherein the
proportion of the landscape in various successional
states remains fairly constant, although the location
of patches constantly shift. This basic definition of
eqUilibrium can be modified in highly dynamic
landscapes to include systems in which all successional stages are constantly present, although
highly variable in relative proportion (White and
Walker, 1997). Being able to model a patch dynamic equilibrium state for a particular landscape
is a significant route by which the natural, stable
disturbance regime can be identified. Several researchers have concluded that dynamic eqUilibrium
is a function of the relationship between the mean
size of disturbances in a stable disturbance regime
and the size of the area of interest (Shugart, 1984;
White, 1987). Thus simulation modeling can aid in
determining (1) the stable disturbance regime of a
region, (2) whether a particular study area is large
enough to capture this disturbance regime, and (3)
the appropriate temporal scale for assessing the disturbance regime. For example, in applying a spatially explicit, GIS-based fire model (REFIRES) to
a chaparral ecosystem in southern California, Davis
and Burrows (1994) found that the variation between long model runs with different initial conditions but the same parameter settings was only
slight. This suggests that, within regions with uniform climate, vegetation, and physiography and
over long periods of time, a fire-controlled chaparral system may approach an equilibrium state.
Such an assessment can aid in determining the important components in a system's natural disturbance regime, and the appropriate spatial and temporal scales for describing that disturbance regime.
Models range widely in their degree of specificity and spatial detail, from whole landscape
models, which estimate the value of a variable over
an entire landscape, to distributional models, which
estimate the distribution of values of some variables(s) over time, and finally to spatial landscape
models, in which the spatial location and configuration of a variable can be tied to any number of
landscape measures and submodels (Baker, 1989).
Models may be either continuous or discrete in their
treatment of time. Although spatial models are the
most detailed and informative, their computational
demands are prohibitive. Distributional models,
therefore, have been most widely developed and
applied, whereas whole landscape models are usually integrated as submodels. Many recent models
have been developed and implemented in a GIS environment (see Goodchild et aI., 1993, and HainesYoung et aI., 1993) and structured on a raster format, often using a hexagonal grid tessellation to
facilitate contagious diffusion among grid cells.
Such a format allows for easy integration with other