Kuletz and others 



Chapter 15 



Inland Habitat Suitability in Southcentral Alaska 



Table 3 Multiple regression model relating activity of Marbled Murrelets 1 to survey period, weather, topographic, 

 and vegetation variables at three study areas: western Prince William Sound, Kenai Fjords National Park, and 

 Afognak Island. Categorical variables were entered into the regression as dummy variables 



1 Variable was natural log transformed 



2 Tested null hypothesis that coefficient estimate = 



3 Variable was square root transformed 



= 6.73, df = 2, P = 0.03) were greater at sites located at heads 

 of bays, than at more exposed sites. 



Identification of Occupied Sites 



The probability of observing occupied behavior was 

 greater: (1) at Afognak Island than at other areas; (2) during 

 upland surveys than during boat or shore surveys; (3) during 

 days with a high percentage of clouds than during clear 

 days; and (4) at bays (especially at heads of bays) than at 

 exposed sites (table 4). The probability of observing occupied 

 behaviors did not vary with survey period or windspeed. 

 Occupied sites had greater levels of cloud cover, forest 

 cover, canopy cover, canopy height, d.b.h., epiphyte cover, 

 and platforms per tree, than other sites (table 5). Alder cover 

 was greater at other sites than at occupied sites. 



Tree size (d.b.h.) and location relative to the head of a 

 bay entered the model at the P < 0.10 level; survey method 

 and platforms per tree were also included. Standardized 

 parameter estimates (table 6) indicated that d.b.h. and 

 platforms per tree were the most important predictors of 

 occupied sites. The logistic function correctly classified 78.9 

 percent of observations in a jackknife procedure; 82.7 percent 

 of occupied sites, and 74.6 percent of sites of unknown 

 status were correctly classified. 



Discussion 



Habitat Predictors Of Murrelet Use 



Murrelet Activity Levels 



Several variables were consistent predictors of high 

 murrelet activity. Allowing for survey period, activity was 

 highest at the heads of bays, at low elevations, and in areas 

 with a high percentage of forest cover and large diameter 

 trees. The most important habitat variables across all study 

 areas were location relative to heads of bays, tree size (d.b.h.), 

 and epiphyte cover on trees (excluding the Naked Island 

 group for which there was no data on epiphyte cover). The 

 number of platforms per tree was also important because it is 

 highly correlated with epiphyte cover. 



The importance of tree size and the number of platforms 

 per tree was consistent with results from other studies and 

 with attributes of nest trees found in southcentral Alaska 

 (Hamer and Cummins 1991; Hamer, this volume; Naslund 

 and others, in press). The importance of location relative to 

 heads of bays was noted in earlier analyses of Prince William 

 Sound data (Kuletz and others, in press; Marks and others, in 

 press) but has not been reported elsewhere. Further, the 

 trend for a bay effect in Kenai Fjords National Park was not 

 significant in prior analyses (Kuletz and others 1994b). It is 



146 



USDA Forest Service Gen. Tech. Rep. PSW-152. 1995. 



