American Bison - Buffalo Field Campaign

Appropriate facilities usually include custom sized and
constructed chutes and alleyways, crash gates, and chute
crowding tubes. It is expensive to construct facilities safe for
bison (and the people working with them), and we strongly
recommend visiting facilities that have proven to be safe
and effective. Highly credible facilities include those at YNP
(Gardiner, Montana), the Baca Ranch (Colorado), Badlands
National Park (South Dakota), and EINP (Alberta).
9.7 Modelling to Assess Bison
Populations and Habitat
Computer models are routinely used to improve our
understanding of bison population and disease dynamics,
and to forecast probable genetic consequences resulting from
particular management actions. In the future, we should expect
even more widespread use of quantitative models, which can,
and likely will, be used for a broad range of purposes. A detailed
treatise on modelling is well beyond the scope of this plan. The
main goals of this section are, therefore, to provide readers
with the minimal background necessary to seriously consider
the utility of using an existing model, or of constructing a new
management-oriented model, and to provide sufficient insight
to the modelling process, that they can reasonably evaluate the
validity and usefulness of model results, or at least ask questions
that will help resolve these issues.
For conservation purposes, population viability analysis (PVA)
and population habitat viability assessment (PHVA) have
become common, and important, approaches for assessing
existing populations and for evaluating potential restoration
or reintroduction projects. We restrict PVA and PHVA to
analyses that employ quantitative modelling to assess the
risk of extinction, or which attain a quantitative population
threshold greater than extinction (“quasi-extinction”, from
Ginzburg et al. 1982; Burgman et al. 1992; Ralls et al. 2002).
Other thresholds for evaluation could include attaining a
specified level of inbreeding depression or allelic diversity, or
estimating the likelihood that a proposed introduction plan will
result in establishment. Conclusions drawn from expert panels,
committees, and other source of opinions, in the absence of a
quantitative model, do not constitute a PVA (Reed et al. 2002).
PHVA is a much broader process than PVA, and includes
evaluation of geographical, social, regulatory, and ecological
considerations that may significantly affect a species. The PHVA
process includes a broad range of stakeholders and leads to
specific recommendations for conserving a species in the area
considered (http://www.cbsg.org/cbsg/phva/index.asp). Viability
analysis is important to bison conservation because so many
bison populations are small and clearly at risk, and because we
have a rich knowledge of factors necessary to conduct credible
and insightful evaluations.
100 American Bison: Status Survey and Conservation Guidelines 2010
The small size of many bison herds has raised concerns about
retention of genetic diversity, and these concerns motivated
detailed simulations to evaluate effects of management actions
on retention of genetic variation in bison herds (Gross et al.
2006; Halbert et al. 2005; Wilson and Zittlau 2004). Other
modelling studies have focused on brucellosis dynamics and its
control in bison (Dobson and Meagher 1996; Gross et al. 1998;
2002; Peterson et al. 1991; Treanor et al. 2007) and on illustrating
population dynamics of bison (Brodie 2008). All wildlife models
are ultimately limited by data availability, and model results
can be misleading when forecasts are presented with an
apparent precision that is not justified by the underlying model
assumptions, structure, or the accuracy of model parameters
(Ralls et al. 2002; Reed et al. 2002). In general, the most
appropriate use of simulation model results is to evaluate the
merits of alternative management actions, rather than to define
an absolute threshold population size. In particular, minimum
critical population sizes may be sensitive to small errors in
parameter estimates, or to the functional structure of strong
environmental perturbations.
9.7.1 Guidelines for using computer simulations
The first critical step is to clearly define the objectives of the
modelling exercise. If the intent is to evaluate management
actions, the best objectives are quantitative, specific, time-
bound, and consist of “treatment” variables (e.g., number of
founders, number or proportion removed) that can reasonably be
simulated by a computer model. A good objective must include
the likelihood of achieving the desired results, the quantitative
value of a threshold, and a time horizon. For example, a bison
PVA used the genetic objective to achieve a 90% probability of
retaining 90% of currently observed selectively neutral genetic
heterozygosity for 200 years (Gross et al. 2006).
Below, we list steps that will be required to construct a computer
model to support bison conservation. A number of recent
treatises provide more detailed information about this process
(we especially recommend Burgman et al. 1993; Bessinger
and Westphal 1998; Bessinger and McCullough 2002; Hilborn
and Mangel 1997). Although we list steps sequentially, most
modelling exercises are iterative and involve simultaneously
working through a number of these tasks and revisiting them as
more information or insight becomes available.
1. Clearly articulate the objectives of the modelling exercise. It is
essential to clearly identify a small, discrete set of “treatments”
and “responses”.
What management must be evaluated?
What is the relevant time frame?
What model outputs are to be evaluated?