Our winter severity index (WSI) gave values highly 

 correlated (r=0.92, n=19, P<0.01) with values from a modified 

 Leckenby Index (Leckenby and Adams 1986), required less 

 detailed data, and could be calculated for a longer series of 

 years . 



Statistical procedures generally followed Snedecor and 

 Cochran (1967) and Zar (1984). Analyses were conducted using 

 the Montana State University computing service and personal 

 computers. Statistical packages used included the 

 Statistical Analysis System (SAS) (Ray 1982) and MSUSTAT (Lund 

 1983) . 



Estimating Population Numbers 



Like most mule deer habitats in the northern Great 

 Plains, our study area was "open". Grassland - low shrub 

 vegetation covered about half of the area and stands of 

 coniferous vegetation were typically small or patchy, 

 distributed along slopes, and held only scattered to low 

 densities of trees that rarely exceeded 15 m in height. The 

 area also comprised a relatively discrete population-habitat 

 unit because mule deer were usually distributed on yearlong 

 individual home ranges scattered throughout the area. During 

 late autumn, winter, and especially early spring, the deer 

 tended to group locally and utilize uplands and open ridgetops 

 such that observability and counting efficiency was high. 

 Because of this, aerial census provided a highly efficient and 

 effective means of measuring population characteristics and 

 trend. 



Our use of complete-coverage surveys eliminated possible 

 bias resulting from sampling design. Quadrat and transect 

 sampling require precise sampling systems that ensure random 

 and representative effort over complex mosaics of topography 

 and vegetation. They are also subject to the same visibility 

 bias of any census that results in fewer animals seen than 

 actually occur (Caughley and Goddard 1972) . Our counts always 

 represented an absolute minimum estimate of the number of deer 

 on the area. To develop a reasonable total population 

 estimate, we had only to account and adjust for visibility 

 bias. This we accomplished by developing observability 

 indices (estimates of proportions of total deer observed) 

 relative to season, survey conditions, aircraft, and 

 observers. We also collected data on a variety of population 

 characteristics for comparison and reconciliation with 

 population estimates through arithmetic modeling (Mackie et 

 al . 1981). A major thrust of our intensive studies during 

 1975-1987 was to further develop and test this approach to 

 population estimation. 



30 



