Multiscale 1 o 1 



Habitat has been described as a western mixed mesophytic forest 

 (Braun 1950, Miller and Neiswender 1987). Dominant canopy plants are sweet 

 gum {Liquidambar stryaciflua), tulip poplar (Liriodendron tulipifera), elms 

 (Ulmus spp.), oaks (Quercus spp.), and hickories (Carya spp.). There is an 

 extensive network of grape (Vitis spp.) and poison ivy (Toxicodendron radicans) 

 vines throughout the canopy. The understory is dominated by spicebush (Lin- 

 dera benzoin). Dominate ground cover species are Osmorhiza sp., Smilacina 

 racemosa, Toxicodendron radicans, Urtica sp., various woodland grass species, 

 and seedlings of the dominant canopy and understory species. A detailed analy- 

 sis of the habitat on the station can be found in Ladine (1995). 



A 14 X 14 trapping grid was established using folding Sherman live 

 traps (H. B. Sherman Traps, Inc.; Tallahassee, Florida) spaced ca. 10-m apart. 

 Trapping was conducted from 28 January 1995 through 9 February 1995. Traps 

 were baited with oatmeal, left open during the day, and checked at sunrise. Esti- 

 mation of population size was made using the Schnabel method (Krebs 1989). 



Location of the trapping grid was entirely within a mature stand of oak, 

 sweet gum, and tulip poplar trees. The selected location has been shown to be 

 homogenous on the macrohabitat scale (Ladine 1995). Placing the grid in this 

 location avoided potential confoundment during statistical analyses posed by 

 placing traps in differing macrohabitats. 



Trap sites were classified according to the occurrence of captures of P. 

 leucopus. Trap sites with at least one capture were classified as capture sites. 

 Other sites were classified as no-capture sites. To strengthen the multivariate 

 analyses and remove the possibility of nonorthoganal functions and components 

 (Tabachnick and Fidell 1989), thirty randomly selected no-capture sites were 

 designated for habitat association analyses. 



Twelve selected habitat variables (Table 1) were measured at each cap- 

 ture site and at each no-capture sites. All selected habitat variables were mea- 

 sured at each of three spatial scales (1 m 2 , 5 m 2 , and 10 m 2 ) in circular plots cen- 

 tered on each trap. These scales were selected following Noon (1981) who sug- 

 gested that a more homogeneous habitat be sampled more finely than a hetero- 

 geneous habitat in order to detect the inherent heterogeneity. Thus, because of 

 the apparent homogeneity of the habitat within the trapping grid (Ladine 1995), 

 these selected scales were used. 



All statistical analysis were conducted using Statistical Analysis Sys- 

 tems (SAS Institute 1989). Habitat variables for capture and no-capture sites 

 were compared at each scale with a Kruskal-Wallis test of Chi-square approxi- 

 mation. Selected habitat variables between scales were compared with a 

 Kruskal-Wallis test of Chi-square approximation to test for differences among 

 selected scales. To control for group-wide Type I error, all multiple pairwise 

 comparisons were made using a sequential Bonferroni adjustment (Rice 1989) 

 with initial « = .05. 



