Overview

The cheetah viability influence diagram (Figure 2) consists of five sub-diagrams: decisions, habitat, direct effects on population dynamics, population dynamics, and loss. The decision nodes represent time (t), region (q), and management option (m) values at which influence diagram outputs are desired. Cheetah habitat is characterized by chance nodes for the region's climate (C), unprotected land use (U), and the proportion of a region's area that is protected (reserves) (R_t). A single direct effect chance node follows: within-region hunting pressure due to poaching or pest control (H_t). As more data becomes available, other direct effects such as disease could be added.

Next, cheetah population dynamics is represented as as a system of SDE's. This system consists of the within-region nodes: birth rate (f_t), death rate (r_t), number of cheetah-prey herbivores (hereafter, herbivores) (B_t), cheetah carrying capacity (K_t), cheetah count (N_t), and fraction of the region's area on which cheetah are detected (D_t). The final section contains the utility node - expressed here as loss (L).

***** [Figure 2 about here] *****

There are three management options: do nothing, increase anti-poaching enforcement, and expand ranching. The first option would be used as a benchmark to evaluate cheetah viability if current conditions are unchanged. Other options such as increase protected areas or livestock loss compensation could be added. Increasing anti-poaching enforcement lowers the cheetah death rate over all regions. Extending ranch lands has two effects: first, each region's unprotected land use distribution is modified so that the event U= ranching is more likely, and second, hunting pressure on both cheetah prey and cheetahs is increased.

The influence diagram's marginal distribution of N_t within each region at a given time and given management option can be used to assess cheetah viability. For example, a typical risk-based definition of viability is: ``with probability .99, a species' count is positive 1000 years into the future'' (see Burgman et al. 1993, p. 37).

Table 1 lists all nodes and their governing distributions.

***** Table 1 about here *****