TRACKING WILDLIFE BY SATELLITE 



27 



1 0.5 



Lying 

 Walking 



Feeding 

 Running 



Mean number of messages for 

 overpasses yielding locations (7.8) 



2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 



M essages / Over pass 



Fig. 23. Relation between the probability of misclassifying a 

 caribou's activity based on short-term activity counts and the 

 number of messages (i.e., activity counts) received during an 

 overpass. Each data point was estimated from 50,000 

 computer iterations. 



The experimental collar seemed to be too heavy (1.6 kg) 

 and too large (7 cm wide) for female mule deer. The deer 

 frequently attempted to remove the collar and seemed 

 uncomfortable with it, even after wearing it for 3 days. A 

 lighter collar with a shape more contoured to the slender 

 neck of female mule deer would provide activity data with 

 less disturbance to the animal. 



For elk, considerable overlap between counts from bed- 

 ding and feeding activity occurred, even at the best 

 setting of the mercury tip-switch (+2). Elk frequently 

 graze with their necks pointed downward; they also trot 

 with their heads held high, pointed upward. An orientation 

 of +2 succeeded in generating intermediate counts for 

 feeding and high counts for walking-running, but also 

 resulted in occasional intermediate and high counts while 



bedded. Orienting the switch to be less sensitive might 

 well result in low counts during prolonged grazing, when 

 the neck is held down, or perhaps even during trotting, 

 when the neck is held high. 



The captive female moose seemed to not be bothered by 

 the size and weight of the collar. Feeding by the moose 

 (browsing on tall shrubs) produced slightly better separa- 

 tion from inactive behaviors than did feeding by the elk 

 (ground-level grazing of grasses and forbs). 



General activity patterns for free-ranging caribou, as 

 predicted by the short-term activity index, correlated well 

 with information about the movement patterns of these 

 animals. However, the proportion of time estimated in 

 feeding activity was considerably lower and the time 

 spent inactive higher than documented by observational 

 studies of Alaskan caribou. It remains to be seen which 

 factors contribute to a bias in favor of the inactive category 

 at the expense of the three active categories. Such biases 

 should be identified, quantified, and corrected by calibra- 

 tion under wild conditions. We conclude that although the 

 general patterns we found by using PTT's are reliable for 

 behavior types, they should not be used quantitatively for 

 energetic studies at this time but rather as an index for 

 assessing gross seasonal trends and herd-specific differ- 

 ences in activity budgets. 



Temperature Sensor 



All PTT's we deployed since 1985 have included tem- 

 perature sensors. We have not tested all PTT's for accu- 

 racy, but a spot check of 4 third-generation PTT's that had 

 been deployed on caribou during 1987 showed that 

 temperatures were reliably reported with little variation 

 among the four. The overall mean slope relating reported 

 temperature to ambient temperature was 0.901; differ- 

 ences among the four PTT's were nonsignificant 

 (P > 0.05). 



However, we are uncertain what the temperature data 

 gathered under wild conditions represent. The tempera- 

 ture sensor reflects ambient temperatures under controlled 



Table 14. Correspondence of activity predicted by 50, 000 simulation trials with the known (observed) caribou (Rangifer 

 tarandusj activity that produced the distribution (sample sizes given for each count distribution). In all cases, the 

 length of the count series (overpass length) followed the distribution for northern Alaskan caribou. 



Predicted activity 



