128
Danny C. Lee
The key to developing suitable management tools is to find a way to link
conventional well-understood viability models with the more nebulous reality of
habitat impacts on population parameters. Management concerns for Bull Trout
(Salvelinus confluentus) motivate the following demonstration of an approach to
risk assessment that combines a stochastic logistic model of population abundance with conditional probability matrices that tie habitat conditions and proposed activities to model parameters. The model parameters, model outputs, and
the conditional probability matrices are all combined within a Bayesian belief
network (BBN), which provides the analytical structure for examining possible
implications of proposed actions.
The example network presented here was constructed in 1993 as a prototype to
explore the potential for BBN technology and to help identify research needs. It
was not developed for real-world decision making, nor has it been used as such.
Because BBNs are a relatively recent development in the field of artificial intelligence, a brief introduction to the structure and purpose of BBNs is provided.
Bull Trout: A Species in Jeopardy
The plight of the Bull Trout is illustrative of the problems facing many native trout
and char of western North America and of the challenges facing land-use managers that consider the needs of sensitive species in their management activities.
Twenty-four endemic species and subspecies of the genera Oncorhynchus, Salvelinus, and Thymallus join the Bull Trout on the American Fisheries Society’s list
of endangered, threatened, or fishes of special concern (Williams et al. 1989).
Heightened concern for the viability of the Bull Trout prompted the U.S. Fish and
Wildlife Service recently to review the status of the Bull Trout and list it as a
threatened species.
A principal threat to Bull Trout and other native salmonids is loss or disruption
of habitat. Such losses result from a combination of changing land-use patterns
and instream alterations over the past 150 years (Meehan and Bjornn 1991;
Nehlsen et al. 1991; Behnke 1992; Rieman and McIntyre 1993). Logging, mining,
agricultural practices, and urbanization have degraded watersheds, whereas irrigation and hydroelectric dams or diversions have reduced the quantity and quality
of available habitat. In addition, overharvest and hybridization with introduced
species and stocks have compounded the problems. The decline in suitable habitat
has fragmented and isolated native trout populations. Many of the strongest populations supported by natural reproduction are found in undisturbed or wilderness
areas (Rieman and Apperson 1989; Rieman and McIntyre 1993; Lee et al. 1997).
Bull Trout pose a special challenge to land managers because they are widely
distributed throughout their range (Fig. 9.1), but they are seldom locally abundant
due to demanding habitat requirements (see Howell and Buchanan 1992; Rieman
and McIntyre 1993). This combination increases both the likelihood of the presence of a Bull Trout population within a given watershed and the sensitivity of that
population to watershed disruption. In the United States, much of the best remain-
Danny C. Lee
The key to developing suitable management tools is to find a way to link
conventional well-understood viability models with the more nebulous reality of
habitat impacts on population parameters. Management concerns for Bull Trout
(Salvelinus confluentus) motivate the following demonstration of an approach to
risk assessment that combines a stochastic logistic model of population abundance with conditional probability matrices that tie habitat conditions and proposed activities to model parameters. The model parameters, model outputs, and
the conditional probability matrices are all combined within a Bayesian belief
network (BBN), which provides the analytical structure for examining possible
implications of proposed actions.
The example network presented here was constructed in 1993 as a prototype to
explore the potential for BBN technology and to help identify research needs. It
was not developed for real-world decision making, nor has it been used as such.
Because BBNs are a relatively recent development in the field of artificial intelligence, a brief introduction to the structure and purpose of BBNs is provided.
Bull Trout: A Species in Jeopardy
The plight of the Bull Trout is illustrative of the problems facing many native trout
and char of western North America and of the challenges facing land-use managers that consider the needs of sensitive species in their management activities.
Twenty-four endemic species and subspecies of the genera Oncorhynchus, Salvelinus, and Thymallus join the Bull Trout on the American Fisheries Society’s list
of endangered, threatened, or fishes of special concern (Williams et al. 1989).
Heightened concern for the viability of the Bull Trout prompted the U.S. Fish and
Wildlife Service recently to review the status of the Bull Trout and list it as a
threatened species.
A principal threat to Bull Trout and other native salmonids is loss or disruption
of habitat. Such losses result from a combination of changing land-use patterns
and instream alterations over the past 150 years (Meehan and Bjornn 1991;
Nehlsen et al. 1991; Behnke 1992; Rieman and McIntyre 1993). Logging, mining,
agricultural practices, and urbanization have degraded watersheds, whereas irrigation and hydroelectric dams or diversions have reduced the quantity and quality
of available habitat. In addition, overharvest and hybridization with introduced
species and stocks have compounded the problems. The decline in suitable habitat
has fragmented and isolated native trout populations. Many of the strongest populations supported by natural reproduction are found in undisturbed or wilderness
areas (Rieman and Apperson 1989; Rieman and McIntyre 1993; Lee et al. 1997).
Bull Trout pose a special challenge to land managers because they are widely
distributed throughout their range (Fig. 9.1), but they are seldom locally abundant
due to demanding habitat requirements (see Howell and Buchanan 1992; Rieman
and McIntyre 1993). This combination increases both the likelihood of the presence of a Bull Trout population within a given watershed and the sensitivity of that
population to watershed disruption. In the United States, much of the best remain-