Kuletz and others 



Chapter 15 



Inland Habitat Suitability in Southcentral Alaska 



Table 1 The number of detections for categorical variables considered for inclusion in multiple regression analyses 

 relating activity of Marbled Murrelets to survey period, weather, topographic, and vegetation variables. A Kruskal-WaJUs 

 nonparametric analysis of variance tested the null hypotheses that murrelet activity did not differ between (or among) classes 

 of each variable 



of bays than elsewhere in bays or on exposed shorelines 

 (table 1). Windspeed did not significantly affect murrelet 

 activity and activity did not vary significantly among survey 

 methods (by boat, from shore or upland; Table 1). 



Correlation coefficients between Marbled Murrelet 

 activity and continuous weather, topographic, and vegetation 

 variables measured in all four areas varied from -0.16 for 

 alder cover to 0.39 for d.b.h. (table 2). The largest correlation 

 coefficients were between murrelet activity and variables 

 directly related to nest site selection (epiphyte cover, platforms 

 per tree; table 2). 



Our reduced model explained 52 percent of the total 

 variation in murrelet activity (table 3). Parameters for survey 

 period, location relative to the head of a bay, and epiphyte 

 cover were highly significant. Based on ratios of parameters 

 to their standard errors (table 3), epiphyte cover, survey 

 period, and location relative to the head of a bay were the 

 most important predictors of murrelet activity. 



Across all four study areas combined, tree d.b.h. (x 2 = 

 7.58, df = 2, P = 0.02), number of potential nesting platforms 

 (X 2 = 7.08, df = 2, P = 0.03), and percent epiphyte cover (x 2 



Table 2 Pearson correlation coefficients between continuous variables 

 considered for inclusion in multiple regression model and murrelet activity 

 (Overland detections <200 mfrom observer) 



1 Not estimated at Naked Island 



USDA Forest Sen-ice Gen. Tech. Rep. PSW-152. 1995. 



145 



