Raphael and others 



Chapter 18 



Landscape-level Analysis of Habitat in Washington 



Table 3 Washington Department of Natural Resources serai stage classification (Green and others 1993) 



Class 



Description 



Late serai 

 Mid-serai 

 Early serai 



Cleared/Other 



Water 



Non-forested 



Coniferous forest stands, > 10 pet tree crown closure in trees > 21" d.b.h., with > 70 

 pet total crown closure, and < 75 percent of the crown in hardwoods or shrubs 



Coniferous forest stands, < 10 pet tree crown closure in trees > 21" d.b.h., with > 70 

 pet total crown closure and < 75 percent of the crown in hardwoods or shrubs 



Coniferous forest stands, 10-70 pet total crown closure and < 75 pet of the crown in 

 hardwoods or shrubs 



< 10 pet crown closure conifers and/or > 75 pet of the crown in hardwoods or shrubs 



Open water bodies 



Non-forest land (agriculture, urban, rock, etc.) 



occurrence against measures of landscape pattern and 

 composition. A map of major river basins depicting WDNR' s 

 Water Resource Inventory Areas (WRIA's) was obtained 

 from WDNR and used to subdivide the vegetation maps into 

 analysis units based on river drainages (fig. 1) for the basin- 

 level analysis (Green and others 1993). 



Accuracy of Forest-Cover Maps 



Forest-cover maps used for the Marbled Murrelet 

 landscape analysis were developed by WDFW and WDNR. 

 The WDFW data set was developed from 1984 and 1986 

 Landsat Multi-Spectral Scanner (MSS) imagery. This imagery 

 has a minimum spatial resolution of approximately 80 m 2 

 and collects information in four spectral bands. Digital 

 elevation models were used by WDFW to compensate for 

 shadowing on north-facing slopes (Eby and Snyder 1990). 

 The stated accuracy of this data source for mapping old- 

 growth cover is 80 percent for the Cascades (20 percent 

 error of commission and 7 percent error of omission) and 85 

 percent for the Olympic Peninsula (15 percent error of 

 commission and 7 percent error of omission) (Eby and Snyder 

 1990). Errors of commission are areas that are mapped as 

 old-growth forest, for example, but are found to be some 

 other type upon field inspection. Errors of omission are 

 areas of old-growth forest that are missed in the mapping but 

 are found to exist on the ground. Accuracy was assessed by 

 WDFW by checking mapped interpretations against field 

 observations (Eby and Snyder 1990). Other potential errors 

 in this data set are large sawtimber stands mapped as old- 

 growth forest, wind-throw or fire regenerated stands mapped 

 as old growth, and the omission of small, narrow features 

 and stand edges (Eby and Snyder 1990). In addition, areas of 

 mature deciduous forest and sapling conifer are lumped into 

 the "other forest" class, which causes difficultly in determining 

 actual stand boundaries in areas with little older forest, such as 

 in southwest Washington (Snyder, pers. comm.). In addition, 



because the focus of the mapping effort was to determine 

 areas of old-growth forest, errors associated with other types 

 of land cover were not distinguished. 



The WDNR data set was developed from 1991 TM 

 imagery. This imagery has a minimum spatial resolution of 

 30 m 2 and collects information in seven spectral bands. 

 High altitude aerial photography, field reconnaissance, and 

 WDNR maps were used to guide the classification (Green 

 and others 1993). The stated overall accuracy of this data 

 set within the range of the Marbled Murrelet is 92 percent, 

 with the lowest accuracy in the Puget lowland (87 percent) 

 and the highest in the North Cascades (97 percent) (Green 

 and others 1993). No information is given on errors of 

 commission or omission. Potential confusion in this dataset 

 may be caused by the grouping of stands with >75 percent 

 crown closure in hardwoods and young conifer in the "other 

 forest" category. 



G/5 Processing 



We subdivided both habitat maps (WDFW and WDNR) 

 into WRIA river basins by using ARC/INFO GRID commands 

 for the basin level analysis. Attributes from each basin were 

 then input to the DISPLAY landscape pattern program. 

 DISPLAY is a package of statistical routines that calculates 

 indices of landscape pattern from CIS maps (Plainer and 

 MacNeal 1993). Landscape pattern indices calculated by 

 DISPLAY (table 4) are based on pattern indices discussed in 

 O'Neill and others (1988), Milne (1991, 1992), and Krummel 

 and others (1987). 



For the site-level analysis, we subsetted the WDFW 

 forest condition map into 0.5-mile radius circles around 

 survey locations (fig. 2). We calculated indices of pattern on 

 each resulting circular landscape using the FRAGSTATS 

 program (Marks and McGarigal 1993). FRAGSTATS is a 

 set of routines that calculates indices of pattern on landscapes. 

 FRAGSTATS calculates many of the same indices as 



180 



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



