(Beissinger and Westphal 1998). Rather than taking predictions of extinction risk or population size at face value to make a decision, demographic
models of population viability are better used to compare the effects of
different management options with the goal of setting priorities.
We found it very useful to have a basic model that could be readily
adapted to alternative management questions (Rule No. 6), but only
because the scale and important factors were similar enough among our
projects that it was appropriate to use the same model structure (Rules Nos.
7 and 9).All our projects were concerned with small wolf populations where
stochasticity and social population structure influence population densities.
Each of our wolf projects asked such distinct questions, however, that
different experiments, model adaptations, and output were required.
Our ability to address different management questions was enhanced by
developing case-specific versions of our computer code, not a finished
package that could be used in multiple ways (Rules Nos. 6 and 10). Our
attempt to create a user-friendly version of our model did not work because
the model kept changing to meet case-specific needs. The user shell rapidly
became obsolete and was not worth the investment. The development of a
simpler, educational version of the model may be useful, but this should be
a separate project with its own objectives (Rule No. 2).
We contend it would not have been useful to have a “standing” model or
box to be pulled out and plugged in to answer these management questions.
For the kind of management questions we explored, it was better to keep
a modeler involved and working hand-in-hand with biologists and managers than to try to write a model that staff without programming ability
could use. We repeatedly revised elements of our modeling experiments
beyond the basic model structure. For example in the cumulative effect
experiments for Voyageurs Park (see Section 2.3.3), we tested different
algorithms for compensation between discrete mortality sources, linked
disease to different population segments, considered alternatives with and
without density responses in four demographic rates, and so on. In addition, in some of our projects we were able to quickly address questions
about model and experimental structures as they arose by producing
preliminary results from model prototypes or iterative versions of the
model (Rule No. 5). Building a single, general model retaining all these
options would have been terrifically cumbersome, more time consuming,
and error prone.
Even with our “simple” model, the experiments were at times sufficiently
complex to be overwhelming, especially if all assumptions were challenged
and tested. We recommend that when modeling exercises bog down in
details or complexity or the next step becomes unclear, the modeler should
step back and look for ways to simplify the situation and get the next phase
started somehow. In other words, cut through the details to keep focusing
on what is important (Rule No. 7). Using an iterative or top-down modeling approach (Starfield and Bleloch 1986) was helpful, starting with the
2. Modeling for Endangered-Species Recovery
41
models of population viability are better used to compare the effects of
different management options with the goal of setting priorities.
We found it very useful to have a basic model that could be readily
adapted to alternative management questions (Rule No. 6), but only
because the scale and important factors were similar enough among our
projects that it was appropriate to use the same model structure (Rules Nos.
7 and 9).All our projects were concerned with small wolf populations where
stochasticity and social population structure influence population densities.
Each of our wolf projects asked such distinct questions, however, that
different experiments, model adaptations, and output were required.
Our ability to address different management questions was enhanced by
developing case-specific versions of our computer code, not a finished
package that could be used in multiple ways (Rules Nos. 6 and 10). Our
attempt to create a user-friendly version of our model did not work because
the model kept changing to meet case-specific needs. The user shell rapidly
became obsolete and was not worth the investment. The development of a
simpler, educational version of the model may be useful, but this should be
a separate project with its own objectives (Rule No. 2).
We contend it would not have been useful to have a “standing” model or
box to be pulled out and plugged in to answer these management questions.
For the kind of management questions we explored, it was better to keep
a modeler involved and working hand-in-hand with biologists and managers than to try to write a model that staff without programming ability
could use. We repeatedly revised elements of our modeling experiments
beyond the basic model structure. For example in the cumulative effect
experiments for Voyageurs Park (see Section 2.3.3), we tested different
algorithms for compensation between discrete mortality sources, linked
disease to different population segments, considered alternatives with and
without density responses in four demographic rates, and so on. In addition, in some of our projects we were able to quickly address questions
about model and experimental structures as they arose by producing
preliminary results from model prototypes or iterative versions of the
model (Rule No. 5). Building a single, general model retaining all these
options would have been terrifically cumbersome, more time consuming,
and error prone.
Even with our “simple” model, the experiments were at times sufficiently
complex to be overwhelming, especially if all assumptions were challenged
and tested. We recommend that when modeling exercises bog down in
details or complexity or the next step becomes unclear, the modeler should
step back and look for ways to simplify the situation and get the next phase
started somehow. In other words, cut through the details to keep focusing
on what is important (Rule No. 7). Using an iterative or top-down modeling approach (Starfield and Bleloch 1986) was helpful, starting with the
2. Modeling for Endangered-Species Recovery
41