TRACKING WILDLIFE BY SATELLITE 



17 



c 



o 

 o 







D 



^ 8 



1 7 



- 6 







o 5 



.2 4 



CO 



Fencepost 



A. 



_- 



Caribou 



B. 



H 



JJI 



I 



Satellite Overpass 



Fig. 11. Location errors for a FIT placed on a fencepost (A), and 

 another simultaneously placed on a captive caribou (B) in an 

 enclosure. Each bar from (A) corresponds with the one below 



location; nor was any correlation found between the longi- 

 tudinal or latitudinal components of error and slope (Table 

 12). Similarly, no correlations were found between the 

 azimuth of location error and the aspect of the elk's loca- 

 tion (rank circular correlation, Batschelet 198 1 : 1 87), both 

 when all positions were considered and when only loca- 

 tions with > slope were considered. Keating (personal 

 communication, 1988) also found no correlation between 

 error and aspect of known locations in Glacier National 

 Park, Montana. 



We did not test the effects of vegetative cover on accu- 

 racy or precision. However, Squires and Anderson (Wyo- 

 ming Department of Fish and Game, personal communi- 

 cation), in a test of fixed PTT's in Yellowstone and Glacier 

 national parks, found no association between location er- 

 ror and vegetative covers classed as open meadow, 

 medium-density conifers, and high-density conifers. 



Finally, we were unable to demonstrate a correlation 

 between the elevations of the 44 known elk locations and 

 their magnitude of error (Table 1 2). Fancy et al. ( 1 988) and 



Keating (unpublished report) were similarly unable to find 

 significant correlations between PTT elevation and error 

 magnitude within a single data set. However, comparison 

 among data sets confirmed the findings of French ( 1 986) 

 that locations were biased when calculated using an incor- 

 rect elevation (Fig. 13). The practical effect of Argos 

 estimating locations using sea level when PTT's were at 

 higher elevations was an increase in the spread of location 

 estimates in the longitudinal directions. Two independent 

 studies of location error at high elevations supported this 

 conclusion. Squires and Anderson (unpublished data) cal- 

 culated a mean error of 2,722 m from fixed locations in 

 Yellowstone and Glacier national parks, a much higher 

 error for PTT's at lower elevations than reported by Fancy 

 et al. (1988) or this report. Although the national park 

 transmitter elevations were not reported, they were con- 

 siderably above sea level. Keating (unpublished data) cal- 

 culated a median error of 2,325 m from 23 test locations 

 (n = 691) at elevations ranging from 1,463 to 3,052 m 

 above sea level. 



Elevation-related errors were primarily longitudinal be- 

 cause the satellites travel in nearly north-south orbits. 

 When signals come from PTT's that are higher than the 

 assumed elevation, Argos interprets them as coming from 

 locations that are closer than they actually are to the satel- 

 lite along its across-track direction (Fig. 14). Thus, when 

 PTT's were above sea level, errors tended to be along a 

 direct line from the PTT to the satellite as it passed over. 

 We found highly significant (P < 0.01) circular correla- 

 tions between the azimuth of error and that of the satellite 

 at its nearest approach to the PTT in both the Brooks 

 Range (radio-collared Dall sheep) and Yellowstone Na- 

 tional Park (elk; Fig. 15). Keating (unpublished data) and 

 Squires and Anderson (unpublished data) also found sig- 

 nificant relations between azimuths. All these studies 

 were at relatively high elevations. We found that the cor- 

 relation of error azimuth with satellite azimuth was itself 

 correlated with the elevation of the PTT, strengthening 

 with increasing elevation (Fig. 16). Keating (personal 

 communication) observed a similar strengthening of the 

 azimuth-azimuth relation as PTT elevation increased. 



The magnitude of error arising from a discrepancy be- 

 tween actual and assumed PTT elevation was related to the 

 maximum elevation of the satellite overpass. As suggested 

 by the sea-level data (Fig. 8), errors were greatest at very 

 low and very high satellite elevations, and least at inter- 

 mediate satellite elevations. These errors intensified con- 

 siderably when data were from an elevation considerably 

 above sea level. Keating fit a second-order polynomial 

 regression to data collected from a PTT he had placed at 

 different elevations; he found that almost 53% of the vari- 



