Maps of vegetative cover for the entire base were generated using a back-propagation
neural network (Wu & Westervelt, 1994). The neural network determined the best
correlation between the ground truth data and existing maps (satellite images
from thematic mapper bands 1 through 7, elevation, slope, watershed, and road
buffer). These correlations were then utilized to extrapolate vegetation cover
estimates from the ground truth data to the entire base. The amount of vegetation
on Fort Irwin was quantified in units of percent aerial cover rather than as
numbers of plants or amount of biomass.
Carrying capacity maps were also generated with a back-propagation neural network
that determined the best correlation between the "permanent" physical
characteristics of the cell (e.g., slope, aspect, soil type, elevation) and
data from five hundred points were randomly selected from the northern section
of Ft. Irwin. This northern area is assumed to be at or very near "original"
carrying capacity given that the region hasn't been exposed to severe human
impacts.
The vegetative community within each cell was described in terms of major plant
categories (shrubs and annuals) and phases of growth (green and brown). This
approach allowed the model to describe community level shifts in composition
due to disturbance and secondary successional changes, yet it also eliminated
the need for a highly detailed model that is required if individual plant species
are modeled.
The vegetative community was predicted to return to the climax state following
disturbance with a constant return rate, however, decades may be required for
desert vegetation communities to recover their original composition prior to
disturbance (Prose, et al., 1987; Wallace, Romney, & Hunter, 1980), and there
is debate as to whether communities ever return to the pre-disturbance state
(Knapp, 1992). The model assumes plant communities would return to the pre-disturbance
state. As a means to calculate the rate of return it was further assumed that
following a major shift in community type (from 30% to 1% shrub cover) it would
take 70 yrs. for a cell to return to climax state.
The green vegetation in both community types was estimated to closely approach
the carrying capacity of the cell during the growing season (period of vegetative
growth and reproduction). The growing season for both community types was determined
to be during the six month period from December through May (Beatley, 1974).
In order for the vegetation to near the carrying capacity during the growing
season, it was determined that an intrinsic rate of natural increase of 0.85
was required. Following the growing season, most green vegetation became senescent
and is classified as brown vegetation within the model. All communities (i.e.,
any combination of shrubs and annuals) were assumed to have green cover at the
beginning of the growing season equal to 25% of the maximum green cover of the
previous growing season.
Brown vegetation (i.e., litter and standing dead), plays important roles in
desert ecosystems, especially in nutrient cycling, energy flow, seedling establishment,
and invertebrate activity (West, 1979); it is included in the present model
because it can be consumed by tortoises, although it is not preferred forage.
Brown vegetation is lost from the system via decomposition. Decomposition rates
of the present model were estimated as a function of soil moisture and surface
temperature so that results were consistent with the observation that litter
does not accumulate to substantial levels.
|
- GRASS map of percent shrub cover at Fort Irwin, California - |
 |