Raphael and others 



Chapter 18 



Landscape-level Analysis of Habitat in Washington 



Table S Frequency of Marbled Murrelet survey sites among forest-cover 

 classes by detection class, western Washington 



'Cover classes from Washington Department of Rsh and Wildlife (Eby and 

 Snyder 1990. Collins 1993), updated by Washington Department of Natural 

 Resources (Collins, pers. comm.). See table 2 for cover class descriptions. 



Table 9 Analysis of variance of patch size in relation to survey status 

 (occupied, detected, unoccupied) and cover class (old-growth, large 

 sawtimber, small sawtimber) of Marbled Murrelet survey sites, western 

 Washington 1 



'Patch area was estimated for contiguous cover surrounding each survey 

 site as classified from cover maps of Eby and Snyder ( 1 990), Collins ( 1 993). 



detected sites (31.2, 30.3, 30.6 km, respectively). Elevation 

 of survey sites averaged 482 meters and mean elevation did 

 not significantly differ among occupied, detected, and 

 undetected sites (520, 467, and 473 meters, respectively). 

 Maximum elevation for all surveys was 1,455 meters, 

 minimum elevation was sea-level (0 meters). 



Site Characteristics 



We investigated two general characteristics in describing 

 the 203-ha area surrounding each site amount and pattern 

 of forest-cover classes (WDFW forest-cover map). The 

 relative amounts of each of four general forest cover classes 

 varied significantly among each of the detection classes 

 (table 10). Over the entire sample of 261 survey sites, old- 

 growth forest averaged 18 percent of the 203-ha landscape 

 surrounding each site. Percentages of large sawtimber, small 

 sawtimber and other averaged 9 percent, 1 1 percent, and 61 

 percent, respectively. Percentage of old-growth forest was 

 significantly greater on occupied sites compared to undetected 

 sites (table 10). Similarly, the proportion of large sawtimber 

 was greater on occupied sites than on undetected sites. 

 Proportion of other forest land was greater on undetected 

 sites than occupied sites. 



Many of the landscape pattern indices are correlated. 

 Rather than report estimates for each of the 21 different 

 indices we computed, we used principal components analysis 

 to produce composite landscape shape index variables. This 

 analysis resulted in four factors that contained about 88 

 percent of the variation inherent in the original set of variables. 

 The first factor contained about 61 percent of the variation 

 in the original variables and was used in subsequent analyses. 

 This composite factor was highly correlated (r > 0.80) with 

 10 of the original variables. Values of this composite index 

 increased with increasing number of patches, landscape 

 shape index, Shannon's diversity index, Simpson's diversity 

 index, modified Simpson's density index, Shannon's and 

 Simpson's evenness indices, modified Simpson's evenness 

 index, contagion index, and total edge. Mean values of this 

 composite landscape pattern index (table 11) varied 

 significantly among detection classes (F= 14.88, P = 0.000), 

 and was significantly greater among occupied sites than in 

 either of the other detection classes (planned contrast, t = 

 5.17,P = 0.000). 



We also investigated the influence of shape and size of 

 old-growth and large sawtimber patches (table 11). These 

 attributes are correlated with the amount of each cover 



Table 10 Forest cover (mean percentage) within 203-ha circles centered on Marbled Murrelet survey sites, western Washington 1 



1 Letters indicate results of pairwise comparisons among means; experiment-wise P <0.05, using Tukey 's test Means with same 

 letter (within columns) did not differ significantly. 



: Forest cover map from Washington Department of Fish and Wildlife (Eby and Snyder 1990, Collins 1993). 

 3 Includes status codes 1. 2, and 3 from table 1. 



USDA Forest Service Gen. Tech. Rep. PS W- 152. 1995. 



187 



