Table l-continued Common name Rodentia Forest deer mouseb Unidentified deer mouse Southern red-backed vole Long-tailed vole Creeping vole Water vole Unidentified vole Pacific jumping mouse Carnivora Ermine All species totals 1984 1985 Total 539 96 737 3; 375 60 775 0 30 6 7 51 914 156 1512 4 59 7 8 90 5 13 18 3339 3412 6751 2; 1 Common name Montane shrew Trowbridge’s shrew Vagrant shrew Coast mole Deer mouse Forest deer mouse Totals l Captured more frequently in 1985 than in 1984. Captured more frequently in 1984 than in 1985. Table 2Small mammals captured in snap traps during 1984 on 8 clearcut areas in the southern Washington Cascade Range Common name Total fnsectivora: Vagrant shrew Montane shrew Marsh shrew Trowbridge’s shrew Unidentified shrew Shrew-mole 2 34 1 7 3 2 Rodentia: Yellow-pine chipmunk Townsend’s chipmunk Cascade golden-mantled ground squirrel Northern pocket gopher Deer mouse Forest deer mouse Unidentified deer mouse Southern red-backed vole Long-tailed vole Heather vole Creeping vole Unidentified vole Pacific jumping mouse 19 15 2 4 61 36 64 6 3 6 12 1 10 All species total Table 3Small mammals captured with snap and pitfall traps on an old-growth moderate site in 1985 in the southern Washington Cascade Range (this site was not sampled in 1984) 288 Results A total of 7084 individuals of 23 species were caught over 2 years on the 54 sites. Of this total, 333 individuals of 17 species were caught on eight clearcut sites that were sampled only in 1984 (table 2), and on one old-growth moderate site Snap traps Pitfall traps 6 0 1 0 5 6 10 1 2 1 6 7 18 27 sampled in 1985 (table 3). The remaining 6751 captures of 20 species, made on 45 sites sampled in both years, constitute the focus of this paper (table 1). A consistent feature of these trapping returns is that they are numerically dominated by four species (table 1). Trowbridge’s shrew was the most common species (1946 captures), followed in abundance by the southern red-backed vole (1512 captures), the montane shrew (13 11 captures), and the forest deer mouse (9 14 captures). Insectivorous mammals (genera Sorex, Neurotrichus, and Scupanus) and rodents were roughly equal in the sample, with 3670 insectivores to 3061 rodents (table 1). Identification of all individuals to species is incomplete; 38 Sorex, one chipmunk, and eight microtine rodents, could not be reliably identified (table 1). The largest group of unidentified animals are deer mice (156 young individuals, table 1). Resolution of this problem awaits discovery of a reliable morphological means of distinguishing juvenile and subadult forest deer mice from deer mice (Bangs 1898). Recent work on this problem (Allard and others 1987, Gunn and Greenbaum 1986) was conducted on adult animals, and as the discriminating variables all depend on growth allometries, they do not help with young animals. Adult deer mice were assigned to species based on the tail-length criterion of Allard and others (1987). If captures for 1984 and 1985 are combined, at least 40 individuals were caught for 11 of the 20 species (fig, 1 and table 1). The remaining nine species, although discussed below with respect to their general habitat preferences, are not considered further here. When compared over all 45 sites, 4 of the 11 species were caught with different frequency between years (t-test, P < 0.05). The vagrant shrew and the montane shrew were caught more frequently in 1985 than in 1984, but Trowbridge’s shrew and the forest deer mouse were captured more frequently in 1984 than in 1985 (table 1). 0 Young q Matlm . Old 7 MM” TRSH S”MD TOW OEM0 Species FDMO SRBY CR”0 PlMO Species Response to Gradients of Forest Age and Moisture Forest-age gradient--Only deer mice showed statistically significant differences between age-classes. The deer mouse was more abundant in old-growth than in young forest, and the forest deer mouse was more abundant in both mature and old-growth forest than in young forest (table 4). TWO species, the Townsend’s chipmunk and the Pacific jumping mouse, were caught in higher numbers in old forest than in young forest, although the differences were not statistically significant (table 4). The Gequency of capture differed between years for four speties in the 36 sites comprising the chronosequence. The vagrant shrew, the montane shrew, and the deer mouse wcrc caught more frequently in 1985 than in 1984. Trowbridge’s shrew was captured more frequently in 1984 than in 1985. Change in capture frequency between years was apportioned similarly across age-classes for the montane shrew, Trowbridge’s shrew, and the deer mouse, but the vagrant shrew was captured much more often in mature and old-growth sites in 1985 than in 1984 (table 4). Table 4-Means, standard deviations, and frequency of occurrence (percent) oismall mammal captures (number per 100 tiap-nights) in snap (July and August) and pitfall traps (October and early November) on 36 forested sites representing an age gradient in the southern Washington CascadeRange. (where ANOVA is significant, ditferences between means are indicated with different letters (A, B); c1= 0.05) h i 213 Table &-continued Common name MWll S.D. % Rodents Townsend’s chipmunk MCXI S.D. % % Forest deer mouse MMem S.D. % Southern red-backed V& MeUl S.D. % Rodents creeping vole MWl S.D. % Pacific jumping mouse MeWI S.D. % Total mammals MWI S.D. Species richness MWI S.D. More total mammals were caught in mature.and old-growth forests than in young forests (table 4). Although not statistitally significant, the trend was toward more speciesin older forest age-classes(table 4, P = 0.07). Forest moisture gradient-Within the 27 old-growth forest sites, only the marsh shrew had an unambiguous responseto the moisture gradient (table 5, fig. 2). The marsh shrew was caught in greater numbers on wet old-growth than on moderate sites, and showed a trend of higher captures on wet than dry sites (P = 0.07). The southern red-backed vole was caught more frequently in dry than moderateor wet old-growth forests (table 5). although this pattern is complicated by an interaction with the high elevation of dry sites as discussed below. Table S-Means, standard deviations,and frequencyof occurrence(percent)of small mammal captures(number per 100trap-nights) in snap(July and August) and pitfall traps (Octoberand early November)on 21 forestedsitesrepresenting8 moishrregradient within old-growth forest in the southernWashingtonCascadeRange(whereANOVA is significant, differencesbetween meansarc indicated with different letters; a = 0.05) 1984 1985 Insectivores vagrantshrew Mean S.D. % MWtl SD. % Marshshrew MeSXl S.D. % Trowbridge’sshrewb MCXJI S.D. % Insectivores Shrew-mole MC.Wl SD. % Rodents Townsend’schipmunk MeWI 275 Tablc S-continued 1984 1985 Gammon name Forest deer ,,,oux? MWI S.D. % Southern red-backed Vole MMl S.D. % Rodents creepingvole Me%8 S.D. % Pacific jumping mouse MCUl S.D. 90 Totalmammals MUII S.D. Speciesrichness Mea S.D. of capture across moisture-classeswas different between years for both deer mice species.Becauseof the significant interaction of moisture-classby year, each yea was analyzed separately.Capturesof the deer mouse in 1984 were not significantly different acrossmoisture-classes,but deer mice were infrequently captured in wet (relative to moderate), old-growth forest in 1985 (table 5, P = 0.02), and tended to be captured in dry rather than wet old-growth forest as well (P = 0.08). In parallel fashion, capturesof the forest deer mousewere not significantly different in 1984, but forest deer mice tended to be caught in wet relative to moderate old-growth forest (table 5, P = 0.06). The pattern 1 The frequency of capture differed between years for three speciesin the 27 sites representing the moisture gradient (table 5). The montane shrew was caught more frequently in 1985 than in 1984, but Trowbridge’s shrew and the forest deer mouse were captured more often in 1984 than in 198.5. As was true acrossforest age-classes,changesin capture Table 6-Multiple correlations between individual habitat variables and small mammal abuvdance and speciesrichness for the southern Washington Cascade Raye (statistically significant regresston coefficients (I’< 0.05) are given for each variable; adjusted multiple R is listed at bottom of table) SOUth~ Deer mouse Variable 1984 BSNGSMT MCCTREE TREPIT MSHRUB CSNGL LCONIF LOGC ROCK FERN LOGAB MCDTREE WATER SlOPE ASTUMP BCSTUMP LICH SOIL BSNGL R* 0.05 -.06 i Forest deer mouse red-backed Vole 198 0.0 .l -.C .C .30 .I - frequency between years was apportioned similarly across moisture-classes for montane and Tmwbridge’s shrews (table 5). Tbe forest deer mouse, however, tended to be captured less frequently only in wet old growth. Environmental Elevation-Site Correlates With Mammalian Abundance elevation was not strongly correlated with the abundancesof most species.Only three speciesshowed a significant association with elevation, and of these,only the southern red-backedvole showed a consistent relationship between years [Trowbridge’s shrew in 1984 only, I = a.417 (P < 0.01); the shrew-mole in 1984 only, r = 0.417 (P < 0.01); the southern red-backedvole in 1984, r = 0.533 (P < 0.001) and in 1985, r = 0.527 (P < O.OOl)]. Vegetative and phyaiographic variablesThree mamma- decay class (BSNGSMT) and the percentagecover of midcanopy by coniferous trees (MCCTREE) were negatively correlated with deer mouseabundance,but deer mouseabundancein 1985 was positively correlated with the number of treepits QREPIT, or holes and root tangles associatedwith fallen trees), the percentagecover of mid-canopy shrubs (MSHRUB), and the number of large coniferous trees. Deer mooseabundancein 1985 was negatively correlated with the number of large, well-decayed snags(CSNGL). Correlations between the forest deer mouse and vegetative and physiographic variables were consistent between years, with positive associationsfor the number of large coniferous trees (LCONIF) and the percentagecover of well-decayed logs @XC). In general, the weak nature of the correlations underscoresthe fact that although more abundant in older forests,both speciesalso were found in younger forests. lian speciesand two community variables were significantly correlated with setsof vegetative and physiographic variables (table 6). Of the 18 vegetative and physiographic variables correlated with mammalian abundance,only six were correlated in more than one instance. The pattern of correlation for the southernred-backed vole was different in that some associationsheld between years but others did not. Vole abundancein 1984 was positively correlated with the percentage.cover of rock (ROCK) and the presenceof water (WATER), but it was negatively correlated with the percentagecover of ferns (FERN), logs of Although the abundanceof the deer mousewas significantly correlated with vegetative and physiographic variables in each early decay-classes(LGGAB), and mid-canopy deciduous tress (MCDTREE). In 1985, the correlations for WATER year, the variables were different between years (table 6). and MCDTREE remained, but new positive correlations were In 1984, the number of medium tall snagsof intermediate observedfor the percentagecover of mid-canopy shrubs 271 (MSHFWEQ,the number of large coniferous trees (LCONIF), and slope of the site (SLOPE), and a new negative correlation was seenfor the percentagecover of well-decayed logs (LOGC). The total number of mammals captured (TMAM) and species richness (SPRICH) were not consistently correlated between years with individual vegetative and physiographic variables, although the multiple. correlations were significant in each year (table 6). Community Composition on Gradients of Forest Age and Moisture The composition of mammalian communities basedon presence or absenceof specieswas not clearly related to the forest-agegradient. The hierarchical clustering of sites using Jaccard’scoefficient of similarity on the age gradient (36 sites) produced clusters of sites with similar mammalian composition along the left side of the dendrogmm in figure 3. Increasingly dissimilar sites (in terms of mammalian composition) were joined to the right side of the dendrogram.If forest age was a primary influence on mammalian composition, major clusters would be expected to reflect the age gradient In other words, different clusters should be composed of sites in different age-classes,which was not observed. Each of the five major clusters (cluster 1 = sites I-6; cluster 2 = sites 7-14; cluster 3 = sites 15-20; cluster 4 = sites 21. 26; cluster 5 = sites 27-36) contain sites from all three ageclasses.Inspection of the dendrogram at higher levels of similarity yields some smaller clusters of similar age-class (for example, sites 9-13 and sites 27-31). but few clusters of similar age were found. K-means clustering (specifying three clusters) by the abundanceof the most frequently captured species(11 species)showed the samelack of association between clusters and site age. The first cluster consisted of eight young, four mature, and nine old-growth sites; the secondcluster of four mature and six old-growth sites; and the thiid cluster of one young, one mature, and three oldgrowth sites. Hierarchical clustering on the moisture gradient (27 sites) showed little correspondencebetween old-growth forest moisture-classes(dry, moderate,and wet) and the clusters of sites with similar mammalian speciescomposition (fig. 4). No clusters consisted of just one moisture-class,although some clusters were composedof two classes,either moderate and wet (sites 15-21) or moderate and dry (sites 22-27). The K-means clustering procedure suggestedthat moisture status exerts somewhatmore influence on community composition. Specifying three clusters, the fmt cluster was composedof two dry, eight moderate,and eight wet sites; the second cluster of five dry, two moderate,and one wet site; and the third cluster of just one moderate site. The grouping of five 278 of the seven dry sites in the secondcluster might indicate a tendency for similar mammalian communities on drier sitea. Sites in the secondcluster supported rather high populations of red-backedvoles. This relationship is clouded by the fact that several dry sites were at high elevation, and as discussed below, whether the primary influence is due to moisture or conditions related to high elevation is not clear. Species I II forest deer mouse may expand, a consequenceof synonymizing some of the coastal forms with the forest deer mouse. Relative to the deer mouse,which has one of the widest geegraphic distributions of all speciesin North America and inhabits many different habitats, the forest deer mouse is a forest specialist. It is found in pre-canopy stagesof forest succession(table 2, fig. 5), but is generally outnumberedby the deer mouse in such habitat. A positive relationship between abundanceand forest age was an expectedpattern for the forest deer mouse. As a forest specialist, the forest deer mousewould be expectedto reach high abundancein the well-developed forests that, historically, were the prevailing forest condition in the Pacific Northwest. Observing the samepattern for the deer mouse, Discussion however, was a little surprising. Deer mice were also more abundant in old-growth than in matore or young forest on the Species Response to Gradients of Age and Moisture western slopes of the central Oregon Cascades(Anthony and Forest age gradient-The relationships of deer mice to patothers 1987). and more abundant in riparian zones of oldterns of forest successionin Washington are being reevalugrowth forest than in mature- and sawtimber-classes(sites ated in light of the recent elevation of the forest deer mouse 50-150 years) in northwestern California (Raphael 1988c). to speciesstates (Goon and Greenbaum 1986, Allard and The deer mouse may be more abundant in old rather than others 1987). The forest deer mousehas mostly been conyoung forest becauseof the diversification of ground vegetasidered a subspeciesof the deer mouse (Osgood 1909, but tion with forest succession(Spies, this volume). Older forests see Liu 1954, Ingles 1965, and Sheppe 1961), and that view may more closely resemble the diversity and abundanceof has enhancedthe conception of the deer mouse as an extreme herbaceousplants characteristic of pre-canopy successional habitat generalist. A substantial part of the habitat thought to stagesthan those of many closed-canopy young and mature be occupied by the deer mouse was actually occupied by the forests. This resemblanceis partially related to the uneven forest deer mouse. As demonstratedin this study for the nature of old-forest canopies,which allow the growth of southern Washington CascadeRange, the forest deer mouse ground-layer plant speciesthat depend on increasedlight. is the more abundant of the two speciesin forested habitat Deer mousepatterns of peak abundancein old forests,how(tables 1,4; figs. 1 and 5). Researchhasjust begun to deever, are only relevant to the stagesof forest succession scribe the distributions and patterns of abundanceof these that follow canopy closure. As in other regions (Anthony species,and work is underway on the.taxonomic affinities of and others 1987, Raphael 1988c), the deer mouse is much the forest deer mouse and other Peromyscusspeciesin coastal more abundant in habitats that occur before canopy closure British Colombia. The currently recognized range of the (table 2, fig. 5). 279 Like the deer mouse, Townsend’s chipmunk and the Pacific jumping mouse are rather common in habitats without closed forest canopies (table 2, fig. 5), and are known to inhabit meadows and edge environments (Dalquest 1948). Their tendency toward higher abundance in old-growth forest might well be a response to the complex and patchy forest floor environment of old forests, which may be more similar to their preferred habitats than ground conditions in young forests. Anthony and others (1987) found these species in greater numbers in riparian zones of young, rather than mature or old-growth forests. This difference may have been due to the younger ages (25-50 years) of sites in their “young-forest” age-class compared to those in this study (55-75 years). Given younger sites, the same pattern may hold in the Washington Cascades. Forest moisture gradient-The marsh shrew frequents wet areas (Cowan and Guiguet 1965, Pattie 1973), but it is not as tied to standing water as the water shrew. I have taken marsh shrews at a considerable distance from water in quite dry, second-growth, Douglas-fir forests. Nonetheless, the species appears to be most abundant in wet forest, regardless of forest age. Anthony and others (1987) caught six shrews in riparian zones of the western Oregon Cascade Range, three individuals in old-growth, one in mature, and two in young forests. At first inspection, the southern red-backed vole appears strongly associated with dry old-growth forest (table 5). Although these data indicate that the voles can do well in dry old-growth forests, dry conditions per se have not clearly been shown to be critical. Southern red-backed voles were caught more often at trap sites near water (table 6), and at high elevation sites, patterns that held in both years. Dry sites as a group tended to be at higher elevations (mean = 838 m) than wet (762 m) or moderate sites (754 m). More important, no dry sites were lower than 689 m in contrast to moderate and wet sites with two sites less than 500 m, four less than 600 m, and five less than 660 m. Captures of red-backed voles in these low elevation sites were few, averaging just 2.5 animals per site in 1984 and 0.5 animals in 1985. As a further demonstration of this interaction, the correlation of vole abundance for both years combined with the elevation of dry sites was quite high (r = 0.83; P = 0.02). The primary relationship may be with the environmental conditions related to high elevation, rather than to moisture. Six of the seven dry sites were located south and southeast of Mount Rainier, a region that yielded the highest capture totals for red-backed voles. In fact, nine of the ten sites with the highest capture totals for red-backed voles were from this region; these included four dry and three moderate old-growth sites, and two mature forest sites. Even though the relationship of red-backed vole abundance to dry forest appeared to be positive, it is clearly not simple; elevation and biogeographic factors must also be considered. 280 Both deer mice species shared the interaction between forest moisture-class and years, and followed the same patterns of capture in each moisture-class and year (table 5). For each species, the major difference in capture frequency across moisture-classes was the small number of captures in wet sites during 1985. With only 2 years of data, determining which, if either, of the 2 years represents the “usual” response to moisture condition is impossible. Both species may have responded to a rather widespread influence. Meteorological data from the southern (Wind River), central (Packwood), and northern (Longmire) portions of the study area indicate that 1984 was generally cooler and wetter than 1985 (Manuwal, pers. comm.). Given consistent meteorological differences between years, these may conceivably have influenced population growth differentially across moistureclasses; however, speculations on mechanisms must await clarification of the general response to moisture condition. Trends in Other Species Nine species poorly represented in the capture totals were not tested statistically for their numeric responses to forest age and moisture gradients. Because species that occur in low abundance are a special concern, noting that the small number of captures for these species (table 1: water shrew, coast mole, yellow-pine chipmunk, pika, northern flying squirrel, northern pocket gopher, water vole, long-tailed vole, and ermine) does not necessarily indicate regional scarcity is important. Most species either reach maximum abundance in adjacent habitats or were not effectively sampled by the techniques targeted for forest-floor mammals. The yellow-pine chipmunk, the northern pocket gopher, the water vole, and the pika are all found more commonly at higher elevations than were sampled in this study (Dalquest 1948). The coast mole is present in many habitats (Dalquest 1948; Hartman and Yates 1985, Maser and others 1981) and is probably most common in habitats without closed canopies at lower elevations. It was not readily caught by pitfall and snap traps and is no doubt more abundant in these forests than the capture returns would indicate. The coast mole can be sampled more effectively with traps designed to capture moles, or by indexing activity via mole-run counts (West, in press). The ermine, although typically not an abundant species, is found in a great variety of habitats (Cowan and Guiguet 1965, Maser and others 1981). Other species that were poorly represented in the capture totals are of interest for different reasons. The northern flying squirrel has received attention because of its importance as prey for the spotted owl. As expected, pitfall or snap traps did not adequately sample this species. Track plates were used to index flying squirrel activity, and the resulting frequency-counts were not positively associated with forest age or moisture condition. Whether track frequencies are highly correlated with squirrel abundance is not clear, however (Carey, pers. comm.). Recent work on the flying squirrel by the Forestry Sciences Laboratory (USDA Forest Service, Olympia) in both Oregon and the Olympic Peninsula of Washington should help clarify patterns of squirrel abundance and their association with gradients of forest age or moisture. The habitat affinities of the long-tailed vole are poorly understood in Washington. This species is captured sporadically, although it has been recorded from many different habitats (Cowan and Guiguet 1965, Dalquest 1948, Maser and others 1981, Randall and Johnson 1979). It has been found most commonly in forest-edge environments and brushy riparian zones where grass cover is usually present. I collected three individuals in clearcut areas (table 2), one in a young, one in an old-growth wet, and two in old-growth dry forests (table 1). Although it is not a species that is closely associated with old forests, it needs further ecological study. The water shrew is a riparian specialist. The techniques used in this study probably would be effective in sampling the water shrew if they were concentrated along permanent watercourses, but because the sampling did not focus on riparian strips, captures of this species were few. The water shrew is also more common at elevations higher than most sites in this study (Dalquest 1948). Its requirements for water are understood, but whether it is more abundant in riparian zones in old rather than young forests is unclear. In this study, two shrews were caught in young forests, four in oldgrowth moderate, and three in old-growth wet forests. In a study focused on riparian habitat, Anthony and others (1987) caught two shrews in old-growth, one in mature, and none in young forest. All three individuals were caught at streamside (about 1 m from water), rather than in adjacent riparian habitat (15-25 m from water). Community Patterns in the Southern Washington Cascade Range The mid-elevation forests of the southern Washington Cascades are inhabited by a small mammal fauna that is broadly adapted to naturally regenerated forests. Unique small mammal communities were not seen in forests of different age or in old-growth forests of different moisture condition. With the exception of the deer mice, no species was significantly correlated with the forest-age gradient. Four common species (Trowbridge’s shrew, southern red-backed vole, montane shrew, and forest deer mouse) and a few less common species typically were found at a given site (table 4 and fig. 1). This pattern resulted in a very similar ranking of species by abundance in each year. Clustering of similar sites by small mammal presence or absence and by abundance of common species yielded clusters that related poorly to the age gradient (fig. 3). Similarly, only the marsh shrew was unambiguously related to the moisture gradient,, and clustering of sites by abundance and by small mammal presence or absence produced clusters not clearly related to the moisture gradient (fig. 4). In concordance with these findings, correlations between small mammals and variable means describing the forest environment were generally weak. Statistically significant correlations, especially those similar in both years, were restricted to a few species which generally occurred widely across both gradients (table 6). These patterns are essentially ones found in all three provinces (Oregon Cascades, Oregon Coast Range, and southern Washington Cascades) in this study (Aubry and others, this volume; Corn and others 1988). Aside from differences in species composition related to geographic distribution, small mammal communities of Douglas-fir forests in this region are structured similarly. Community composition did not vary appreciably with forest age, but total mammalian abundance was higher in oldgrowth than in young forests (table 4). The higher abundance partially resulted from the presence of more species in old forest (table 4). Higher abundance may be linked to agerelated differences in the structure and productivity of the ground environment, which is more diverse in old forests. Whether the average number of species is typically greater in old-growth forests is uncertain, as is the associated question of whether variation in species number is related to the age gradient. In this study, coefficients of variation (C.V.) for species richness were similar between years on the age gradient (from table 4: C.V. 1984 = 21.9, 19.1, and 20.4 for young, mature, and old-growth forests; C.V. 1985 = 26.5,21.4, and 22.1). Variation in species number might be expected to be higher in young than in old-growth forest, as a reflection of the higher proportion of dispersal sinks in young forest, and perhaps as a function of the less-diverse and productive resource-base at ground level and the less physically buffered environment. Such a trend may exist, although slight, in the above coefficients, but a thorough answer to these questions requires long-term study. Of particular interest would be the variation in species number with season in young versus old forest. The initial design of this study, which was not fully implemented, included a spring and fall sampling-period for each site as a way of addressing this question. This study, therefore, cannot characterize the magnitude of seasonal changes. That snap-trapped animals were breeding is important to note; however, site occupancy was thus not simply due to the capture of individuals dispersing from favored or survival habitats (Anderson 1980). This spilling-over effect that resulted from the movement of individuals from highdensity to low-density habitats (Fretwell 1972), or from saturation dispersal (Lidicker 1975), would be more evident later in the fall and early winter. What does the apparent lack of strong community patterns on gradients of forest-age and moisture mean? Is it really true that the small mammals of this region are insensitive to variation in forest structure? Not at all; noting that this survey was done exclusively on naturally regenerated stands, 281 clearly-from this study, from work in Oregon (Corn and Bury, this volume a; Gilbert and Allwine, this volume a), and from work on the successional changes in the compositional and structural features of these forests (Spies and Franklin, this volume; Spies and others 1988~thresholds of critical habitat-variables are met for most forest-dwelling species soon after canopy closure. In a naturally regenerating landscape, the major change in the species composition of small mammals occurs before and during canopy closure. At the transition from grass- and forbdominated plant communities to later stages dominated by shrubs and young trees, small mammal communities characterized by the deer mouse, several species of meadow voles, Townsend’s and yellow-pine chipmunks, the Pacific jumping mouse, and shrews of open habitats (the vagrant shrew and the montane shrew) begin to give way to communities of the closed-canopy forest described in this study. Rather minor changes in species composition and relative abundance among species takes place in subsequent successional stages. As a consequence of this pattern of mammalian species replacement, we would expect poor correlations between small mammal abundance and habitat elements in studies that use a data set based on naturally regenerated forests, as was observed. If thresholds for limiting habitat-elements for small mammals are usually met or exceeded in naturally regenerated forests, this data set presents an interesting problem. In terms of information to guide management on forests regenerated after timber harvest, this data set is not usable for identifying critical habitat-elements, or to assess their threshold values-a straightforward consequence of the fact that the range of variation in these variables was insufficient to provide strong correlations or to indicate threshold values. For such a purpose, knowing under what conditions a species is absent is just as important as knowing the conditions under which it is 282 present. On the one hand, this study shows that most species have a fair amount of adaptability in their use of forest habitats, but on the other, it tells us little about sufficient values for habitat elements. If forests managed for timber production are to retain most of their native small mammals, we need to begin the process of identifying critical habitat-elements and assessing their minimum threshold values. That second- and third-rotation forests would resemble naturally regenerated forests in a number of respects seems unlikely, and they certainly would exceed the range of variation in habitat variables sampled in this study. The required information, consequently, can only come from a data set based on stands that are intensively managed for timber production, or an array of experimental sites designed to mimic such conditions. To this end, manipulative field experiments conducted on spatial scales sufficiently large to avoid the problem of animal movement from adjacent edges probably hold the most promise. Because of the need for large scale, such studies might best be undertaken cooperatively between the scientific community and government agencies or private companies. Acknowledgments In the course of this study, over 60 people contributed to the field work and data analyses. Field work required long hours under hard conditions, and I am indebted to these people for completing it. Diane Converse deserves special thanks for her leadership and perseverence over the course of the study. Funding was provided by the USDA Forest Service, Pacific Northwest Research Station. I thank personnel of the Wind River, Mount Adams, Packwood, and Randle Ranger Districts of the Gifford Pinchot National Forest, the Wliite River Ranger District of the Mount Baker-Snoqualmie National Forest, and Mount Rainier National Park for their advice and logistical help. This paper is contribution 125 of the Wildlife Habitat Relationships in Western Washington and Oregon Research Project, Pacific Northwest Research Station, USDA Forest Service. 0 Appendix Table 7-Physiographic and vegetative variables that describe habitats in the southern Washington Cascade Range Variable SLOPE WATER ROCK GRASS HERB SOIL MOSS LICH LITT LITDEP FERN SHRUB MSHRUB MCCTREE MCDTREE CONIFSM DECIDSM LCONIF LDECID SCAN TREPIT ASTUMP BCSTUMP LGGAB LOGC ASNGSMT BSNGSMT CSNGSM ASNGL BSNGL CSNGL Description Percentage slope Presence of permanent water Percentage cover of rock Percentage cover of grasses Percentage cover of herbs Percentage cover of mineral soil Percentage cover of moss Percentage cover of lichens Percentage cover of fine litter (<10-cm diameter) Litter depth (cm) Percentage cover of ferns Percentage cover of ground-layer shrubs Percentage cover of mid-canopy shrubs Percentage cover of canopy and mid-canopy coniferous trees Percentage cover of canopy and mid-canopy deciduous trees Number of small and medium-diameter coniferous trees (l-50 cm) Number of small and medium-diameter deciduous trees (l-50 cm) Number of large-diameter coniferous trees (>50 cm) Number of large-diameter deciduous trees (>50 cm) Presence of supercanopy trees Number of tree-fall pits Number of slightly decayed stumps Number of moderately to well-decayed stumps Number of slightly to moderately decayed logs, d.b.h. > 10 cm Number of well-decayed logs, d.b.h > 10 cm Number of slightly decayed small and mediumdiameter (l-50 cm) short (<l.5 m), medium (5-15 m), and tall (>I5 m) snags Number of moderately small and medium-diameter (l-50 cm) short (<l.5 m), medium (5-15 m), and tall (>15 m) snags Number of well-decayed small and medium-diameter (l-50 cm) short (<l.5 m) and medium (5-15 m) snags Number of slightly decayed large-diameter (>50 cm), medium and tall snags Number of moderately decayed large-diameter (>50 cm). medium and tall snags Number of well-decayed large-diameter (>50 cm), medium and tall snags Continue 283