TRACKING WILDLIFE BY SATELLITE 



23 



10 



o 



< 



CM 



Fig. 19. Mean 24-h activity indices during 

 3-day intervals for female caribou rela- 

 tive to their calving date. A "calving" 

 date of 2 June was assigned to barren 

 females. 



-15 -10 -5 5 10 15 20 25 30 



Days Before or After Calving 



35 



bouts consisted of combinations of browsing, grazing, 

 walking, standing alert, grooming, and interactions with 

 other animals, most of which lasted <60 s. 



Sensor counts were invariably low when the animal was 

 bedded, although good separation of counts was also 

 noted among the three active categories (Fig. 20A). Feed- 

 ing produced counts from to 48 but had little overlap 

 with walking (mostly 30-48) and running (always >36). 



We obtained 229 sampling periods of 60 s each for the 

 captive mule deer, and activities were categorized as for 

 caribou. The running category included only trotting the 

 bounding gait typical of mule deer escape behavior was 

 not observed. Results were similar using the mercury tip- 

 switch at angles of +6 and +10; therefore, data were 

 combined (Fig. 20B). Sensor counts during bedding ac- 

 tivity were clearly distinguishable from all other behav- 

 iors; of 53 counts during bedding activity, only one regis- 

 tered >2. Sensor counts during feeding activity were 

 variable, ranging from to 48, but with the exception of 

 the class, they rarely overlapped counts from the other 

 three activities. Walking activity produced sensor counts 

 of 46-60; running produced sensor counts of 54-60. 



We recorded 46 1 sampling periods of 60 s each for the 

 captive elk. Activities were categorized as for caribou and 

 mule deer, except that continuous walking and running 

 were combined into a single category. (The animal gener- 

 ally responded to being chased by trotting quickly away 

 for a few seconds, followed immediately by walking.) 

 With the activity sensor inclined at +6, grazing activity 

 frequently failed to trigger the mercury switch, resulting in 

 many zero counts during feeding. Zero readings for feed- 

 ing activity overlapped unacceptably with bedding ac- 



tivity, greatly reducing the sensor's usefulness. With the 

 mercury switch inclined at +2, discrimination among the 

 three defined activity categories improved, although over- 

 lap between inactive and feeding remained higher than in 

 the caribou and mule deer experiments (Fig. 20C). At +2, 

 177 of 207 (85%) zero counts occurred while the animal 

 was bedded, although zero was also the most common 

 count for feeding activity. Counts >20 were sometimes 

 obtained while the elk was bedded and displayed no de- 

 tectable movement. Breathing or ruminating may have 

 triggered the mercury switch. Walking-running produced 

 uniformly high counts. 



For the wild elk in Yellowstone National Park, ranges of 

 sensor counts overlapped for all activities, although cate- 

 gories were significantly different (P < 0.001; Fig. 21). 

 The distribution of sensor counts for moving, feeding, and 

 sparring were not significantly different from normal (P > 

 0.05), although the other categories were significantly 

 non-normal (P < 0.05, Shapiro- Wilk statistic). For the 

 bedded category, >97% of observations had a sensor count 

 of zero and 2% had a count of 1 . The elk was involved in 

 grooming behavior during the two remaining bedded ob- 

 servations with higher sensor counts. Other categories 

 generally yielded higher sensor counts, especially spar- 

 ring activity, which had the highest counts. 



Because of the extent of overlap of counts from differ- 

 ent activities of the elk, prediction on the basis of short- 

 term activity was concluded to be limited to active versus 

 nonactive behaviors. Additionally, inferences of time 

 spent active versus nonactive were restricted to those 

 times when data were received, and 24-h patterns could 

 not be determined. 



