9. Assessing Land-Use Impacts on Bull Trout Using Bayesian Belief Networks
139
No information is available on recent population trends, but a survey provides a
rough estimate of the current number of spawning redds within the basin. Other
analyses of populations in similar circumstances suggest that density dependence
is most likely to be moderate, but weak or strong levels are not unreasonable.
Based on the topography and climate of the basin, natural variation could fall
anywhere but is most likely to be in the moderate range.
This information is entered into the Bull Trout BBN by entering findings for the
respective root nodes. By sequentially entering findings for the proposed-activity
node, a series of belief vectors for a minimum run are generated that can be used
as relative indices of the risk to the population associated with each option. To
facilitate comparison, the forest supervisor plots the biological risk, defined as the
probability of a run of five or fewer (five redds might represent a lower threshold
with unacceptable consequences) versus the net economic return (revenues minus
costs) of each timber sale option (Fig. 9.4). From this graph, the supervisor
decides that the increased economic return of the no-effect option is worth the
additional risk over options with low or high positive benefits, but it would not be
worth the additional risk to proceed with options that degrade the habitat.
Example 2: Search for Guidelines
In the second example, a fisheries biologist is asked to participate in developing a
conservation strategy for Bull Trout within the federal lands of his region. Part of
the strategy calls for setting aside selected watersheds that will act as population
reserves. Potentially disruptive activities will be severely limited within watersheds so designated. There are numerous candidate watersheds available; the
biologist wants to be able to identify those most likely to support long-term
populations. Because there are too many candidate watersheds to examine them
all individually, the biologist looks to screen candidates by using the BBN.
The list of possible candidates ranges from relatively pristine streams in parks
and wilderness areas to those heavily impacted by logging, mining, agriculture,
and other activities. As a first step, the biologist hypothesizes that the areas with
lowest risk will be undisturbed watersheds with high initial numbers and low
natural population variation. Findings are entered to reflect these conditions,
given that there is no other information about the streams. The biologist also
minimizes the effect of management by setting the proposed activity node to “no
effect.” The resultant belief vectors (Fig. 9.5) provide a benchmark that can be
used when searching for possible combinations of conditions that would be exceptions to this rule (i.e., streams not fitting this description that have equal or lower
risks). The results suggest that even relatively pristine areas are not risk free,
because there are factors other than habitat that threaten some populations (e.g.,
the spread of introduced competitors such as Brook Trout [Salvelinus fontinalis]).
To guide the search, the biologist uses the diagnostic features of the BBN and
enters findings of “>60” for minimum run, “>100 yr” for years to extinction, and
“no effect” for proposed activity. Propagation of the network reveals the most
139
No information is available on recent population trends, but a survey provides a
rough estimate of the current number of spawning redds within the basin. Other
analyses of populations in similar circumstances suggest that density dependence
is most likely to be moderate, but weak or strong levels are not unreasonable.
Based on the topography and climate of the basin, natural variation could fall
anywhere but is most likely to be in the moderate range.
This information is entered into the Bull Trout BBN by entering findings for the
respective root nodes. By sequentially entering findings for the proposed-activity
node, a series of belief vectors for a minimum run are generated that can be used
as relative indices of the risk to the population associated with each option. To
facilitate comparison, the forest supervisor plots the biological risk, defined as the
probability of a run of five or fewer (five redds might represent a lower threshold
with unacceptable consequences) versus the net economic return (revenues minus
costs) of each timber sale option (Fig. 9.4). From this graph, the supervisor
decides that the increased economic return of the no-effect option is worth the
additional risk over options with low or high positive benefits, but it would not be
worth the additional risk to proceed with options that degrade the habitat.
Example 2: Search for Guidelines
In the second example, a fisheries biologist is asked to participate in developing a
conservation strategy for Bull Trout within the federal lands of his region. Part of
the strategy calls for setting aside selected watersheds that will act as population
reserves. Potentially disruptive activities will be severely limited within watersheds so designated. There are numerous candidate watersheds available; the
biologist wants to be able to identify those most likely to support long-term
populations. Because there are too many candidate watersheds to examine them
all individually, the biologist looks to screen candidates by using the BBN.
The list of possible candidates ranges from relatively pristine streams in parks
and wilderness areas to those heavily impacted by logging, mining, agriculture,
and other activities. As a first step, the biologist hypothesizes that the areas with
lowest risk will be undisturbed watersheds with high initial numbers and low
natural population variation. Findings are entered to reflect these conditions,
given that there is no other information about the streams. The biologist also
minimizes the effect of management by setting the proposed activity node to “no
effect.” The resultant belief vectors (Fig. 9.5) provide a benchmark that can be
used when searching for possible combinations of conditions that would be exceptions to this rule (i.e., streams not fitting this description that have equal or lower
risks). The results suggest that even relatively pristine areas are not risk free,
because there are factors other than habitat that threaten some populations (e.g.,
the spread of introduced competitors such as Brook Trout [Salvelinus fontinalis]).
To guide the search, the biologist uses the diagnostic features of the BBN and
enters findings of “>60” for minimum run, “>100 yr” for years to extinction, and
“no effect” for proposed activity. Propagation of the network reveals the most