First, all the neighboring forest polygons next to the currently occupied one were scored using the weighted parameters. For example, say that a neighboring polygon had the following values: food quality = 4, cover quality = 1, slope = 5, proximity to water bodies = 1, and proximity to roadside salt pools = 3. Then its score would have been equal to the following:
score = (food weight) * (food quality) + (cover weight) * (cover quality) + (slope weight) * slope + (proxWB weight) * (proximity to water bodies) + (proxSP) * (proximity to roadside salt pools)
= 0.40 * 4 + 0.10 * 1 + 0.05 * 5 + 0.10 * 1 + 0.35 * 3 = 3.10.
Next the scores were turned into preferences by determining the two habitat quality ranks on either side of the score and applying a linear interpolation to obtain a precise preference value. For example, the score 3.10 obtained above fell between the habitat quality rankings of 3 and 4, which had values of 0.125 and 0.25 respectively. Consequently, the score 3.10 became 0.125 + (3.10 -3) * (0.25 -0.125) = 0.1375.
These preferences were normalized to 100%. The normalized preferences were laid out along a line from 0 to 1, a random number was selected between 0 and 1, and the preference that bound the selected random number was determined and its corresponding polygon chosen. If, however, this chosen polygon was the one previously visited or the one before that, the process was repeated until a new polygon was selected. This was to avoid the moose returning too soon to a previously visited forest polygon. If the current polygon had only one or two neighbors, the first neighbor was selected in both cases. If the polygon had no neighbors, the moose returned to the forest polygon it had visited before the previous one. It was anticipated that at that point the moose would pick a different neighboring forest polygon and not end up in the same dead end.
It was clear that individual moose had distinct home ranges from the moose movement GPS data. Each moose’s starting location was assumed to be the center of its circular home range, the radius of which was entered into the moose location shape field. All moose ranges were given a radius of 10 km, a value somewhat larger than the 6 km of Voigt et al. (2000) but which approximated the real study-area moose. This radius was used when the moose was traveling to check if it exceeded its home range limit. If this happened, then the moose had to choose a different polygon to move to. This rule was relaxed to just a warning in the final version of the model, because otherwise the moose tended to get stuck on the perimeter of its home range.