262
models. Daily spread of a fire front is predicted
based on vegetation, fuels, a weather submodel, and
a fire behavior submodei. Simulated fire behavior
includes rate of fire spread and height of flames.
The heat flux from fire in a cell raises the temperature of fuel in an adjacent cell, allowing for the
prediction of fire spread rate (Sklar and Costanza,
1991). A submodel also has been developed to estimate the response of small and large mammals to
fire disturbance. The effect of thinning or clearcutting on fire spread is included in the gradient
model FORPLAN (Kessell, 1976; Potter et aI.,
1979).
Models have been constructed for various coniferous forest locations in Montana, the Blue Mountains of Oregon, southern California chaparral, and
Australian parks and reserves. Operation of the
models requires a comprehensive resource inventory to populate grid cells with the necessary environmental attributes. For a model implemented in
Glacier National Park, Montana, the necessary data
were derived from aerial photographs, topographic
maps, and fire history maps (Kessell, 1976).
The simplicity of the submodels allows large
numbers of grid cells to be processed (Baker,
1989a). The models have been designed to be easily operated by resource managers in the management of small to medium fires. Limitations include
simulation of succession as a deterministic process
and the lack of submodels for plant dispersal and
disturbances other than fire. In addition, natural or
artificial fire breaks cannot be accurately simulated
if the cell size (e.g., 1 ha) is too coarse to represent
them (Sklar and Costanza, 1991).
18.2.5 LANDIS
LANDIS is a stochastic, spatially explicit gridbased model of forest succession that incorporates
the effects of fire and windthrow disturbances as
well as harvesting (He and Mladenoff, 1999;
Mladenoff et aI., 1996). A stochastic approach is
taken to simulate landscape dynamics over broad
spatial scales and long time periods, rather than predicting individual events, such as are simulated using the gradient models described previously. As
in gap models, interactions among species life history characteristics, site conditions, and disturbance
regimes determine the course of succession
(Mladenoff et aI., 1996). However, LANDIS can
simulate larger areas than gap models with the use
of a ten-year time step and aggregation of tree
species into ten-year age classes. An objectoriented approach is taken in which each cell in the
simulated grid is a spatial object (He and MladeEcosystem Structure and Function Modeling
noff, 1999). The cell size can be modified for operation of the model at different scales of resolution.
Species life history parameters, including dispersal and establishment characteristics, longevity,
age at sexual maturity, and response to disturbance,
are obtained from the literature (Mladenoff et aI.,
1996). Seed dispersal is a spatially explicit process
based on dispersal distance curves. Establishment
of trees is a function of site characteristics, which
are provided as input to the model in the form of
a land-type data layer. Land types are intended
to correspond to soil moisture and nutrient conditions. Site characteristics as expressed by land types
also affect fuel accumulation and decomposition
(Mladenoff et al., 1996). Disturbances are generated stochastically by selecting from disturbancesized distributions. The actual size of a disturbance
also depends on local susceptibility conditions. Fire
disturbance is assigned to a fire severity class based
on time since last fire. Fire severity and species fire
tolerance determine which trees are killed, beginning with younger age classes (He and Mladenoff,
1999). Windthrow susceptibility increases with tree
age, leading to removal of older age classes first.
Great flexibility in model output is provided, including maps of cover types or age classes and an
interface with a spatial analysis package for production of summaries and landscape indices such
as fractal dimension, connectivity, and patch statistics.
The model has been parameterized for a northern Wisconsin landscape and verification and sensitivity analysis have been performed (He and
Mladenoff, 1999; Mladenoff and He, 1999). A predecessor model, LANDSIM (Roberts, 1 996a,
1996b), was parameterized for the southwest
United States. A potential limitation is the representation of environmental characteristics by a single data layer.
18.2.6 CRBSUM
CRBSUM is a vegetation succession simulation
model that was developed for predicting landscape
characteristics as part of the interior Columbia
River basin (ICRB) ecological assessment (Keane
et aI., 1996a). Successional development is simulated as a change in structural stage and cover type
keyed to successional age, using an annual time
step over a set of grid cells. The ICRB assessment
included two spatial scales: the coarse scale, covering the entire basin at a grid cell size of 1 km 2 ,
and the mid-scale, with a cell size of 0.01 km 2 , covering a set of representative watersheds. CRBSUM
models. Daily spread of a fire front is predicted
based on vegetation, fuels, a weather submodel, and
a fire behavior submodei. Simulated fire behavior
includes rate of fire spread and height of flames.
The heat flux from fire in a cell raises the temperature of fuel in an adjacent cell, allowing for the
prediction of fire spread rate (Sklar and Costanza,
1991). A submodel also has been developed to estimate the response of small and large mammals to
fire disturbance. The effect of thinning or clearcutting on fire spread is included in the gradient
model FORPLAN (Kessell, 1976; Potter et aI.,
1979).
Models have been constructed for various coniferous forest locations in Montana, the Blue Mountains of Oregon, southern California chaparral, and
Australian parks and reserves. Operation of the
models requires a comprehensive resource inventory to populate grid cells with the necessary environmental attributes. For a model implemented in
Glacier National Park, Montana, the necessary data
were derived from aerial photographs, topographic
maps, and fire history maps (Kessell, 1976).
The simplicity of the submodels allows large
numbers of grid cells to be processed (Baker,
1989a). The models have been designed to be easily operated by resource managers in the management of small to medium fires. Limitations include
simulation of succession as a deterministic process
and the lack of submodels for plant dispersal and
disturbances other than fire. In addition, natural or
artificial fire breaks cannot be accurately simulated
if the cell size (e.g., 1 ha) is too coarse to represent
them (Sklar and Costanza, 1991).
18.2.5 LANDIS
LANDIS is a stochastic, spatially explicit gridbased model of forest succession that incorporates
the effects of fire and windthrow disturbances as
well as harvesting (He and Mladenoff, 1999;
Mladenoff et aI., 1996). A stochastic approach is
taken to simulate landscape dynamics over broad
spatial scales and long time periods, rather than predicting individual events, such as are simulated using the gradient models described previously. As
in gap models, interactions among species life history characteristics, site conditions, and disturbance
regimes determine the course of succession
(Mladenoff et aI., 1996). However, LANDIS can
simulate larger areas than gap models with the use
of a ten-year time step and aggregation of tree
species into ten-year age classes. An objectoriented approach is taken in which each cell in the
simulated grid is a spatial object (He and MladeEcosystem Structure and Function Modeling
noff, 1999). The cell size can be modified for operation of the model at different scales of resolution.
Species life history parameters, including dispersal and establishment characteristics, longevity,
age at sexual maturity, and response to disturbance,
are obtained from the literature (Mladenoff et aI.,
1996). Seed dispersal is a spatially explicit process
based on dispersal distance curves. Establishment
of trees is a function of site characteristics, which
are provided as input to the model in the form of
a land-type data layer. Land types are intended
to correspond to soil moisture and nutrient conditions. Site characteristics as expressed by land types
also affect fuel accumulation and decomposition
(Mladenoff et al., 1996). Disturbances are generated stochastically by selecting from disturbancesized distributions. The actual size of a disturbance
also depends on local susceptibility conditions. Fire
disturbance is assigned to a fire severity class based
on time since last fire. Fire severity and species fire
tolerance determine which trees are killed, beginning with younger age classes (He and Mladenoff,
1999). Windthrow susceptibility increases with tree
age, leading to removal of older age classes first.
Great flexibility in model output is provided, including maps of cover types or age classes and an
interface with a spatial analysis package for production of summaries and landscape indices such
as fractal dimension, connectivity, and patch statistics.
The model has been parameterized for a northern Wisconsin landscape and verification and sensitivity analysis have been performed (He and
Mladenoff, 1999; Mladenoff and He, 1999). A predecessor model, LANDSIM (Roberts, 1 996a,
1996b), was parameterized for the southwest
United States. A potential limitation is the representation of environmental characteristics by a single data layer.
18.2.6 CRBSUM
CRBSUM is a vegetation succession simulation
model that was developed for predicting landscape
characteristics as part of the interior Columbia
River basin (ICRB) ecological assessment (Keane
et aI., 1996a). Successional development is simulated as a change in structural stage and cover type
keyed to successional age, using an annual time
step over a set of grid cells. The ICRB assessment
included two spatial scales: the coarse scale, covering the entire basin at a grid cell size of 1 km 2 ,
and the mid-scale, with a cell size of 0.01 km 2 , covering a set of representative watersheds. CRBSUM