Windstorm Disturbance Effects on Forest Structure and Black Bear Dens in Southeast Alaska

Ecological Applications, 15(4), 2005, pp. 1306±1316 ᭧ 2005 by the Ecological Society of America

WINDSTORM DISTURBANCE EFFECTS ON FOREST STRUCTURE AND BLACK BEAR DENS IN SOUTHEAST ALASKA

EUGENE J. DEGAYNER,1,4 MARC G. KRAMER,2 JOSEPH G. DOERR,1,5 AND MARGARET J. ROBERTSEN3,6 1USDA Forest Service, Alaska Region, P.O. Box 304, Petersburg, Alaska 99833 USA 2NASA AMES, Ecosystem Science and Technology Branch, Mail Stop 242-4, Moffett Field, California 94035 USA 3USDA Forest Service, Wrangell Ranger District, P.O. Box 51, Wrangell, Alaska 99929 USA

Abstract. We examined the relationships among forest susceptibility to windstorm damage, forest structure, and black bear winter den site selection in the coastal temperate rain forests of southeastern Alaska, USA. Forest susceptibility to maritime windstorms was determined by applying a wind-disturbance model (WINDSTORM) to two study areas where bear dens had been identi®ed by radio telemetry. We evaluated model predictions for forest structure against ®eld data for one study area. As predicted, forests in wind-protected landscapes contained features indicative of later stages of forest development where larger and older trees are more abundant. By contrast, forests located in storm-susceptible locations contained higher tree densities, less variation in tree diameters, smaller trees, and less evidence of large, old trees with heart rot. Analysis of habitat use relative to its availability on the landscape indicates that bears selected wind-protected landscapes over storm-prone landscapes in both study areas for winter dens. The majority (58%) of den sites were located in forests most protected from catastrophic storm effects. By contrast, only 6% were located in forests most exposed to storm damage. Results suggest that forests most protected from catastrophic windstorm disturbance contain more suitable overwinter habitat for black bears as evidenced by den site selection. Key words: black bear denning ecology; disturbance; forest dynamics; forest succession; spatially explicit models; telemetry; windthrow; winter den site selection.

INTRODUCTION enough protection from the elements and are susceptible to ¯ooding (Alt 1984, Smith 1985, Weaver and Most black bears (Ursus americanus) and brown Pelton 1994). A recent study of black bear dens in the bears (Ursus arctos) hibernate in dens as part of an coastal rain forest of British Columbia, Canada, re- energy-conserving strategy and are expected to bene®t ported that all dens (n ϭ 67) were in trees or in wooden from choosing dens that are thermally ef®cient (John- structures derived from trees (logs, root boles, stumps; son et al. 1978, Hayes and Pelton 1994, Oli et al. 1997). Davis 1996). Bears give birth in their dens and can spend over half While our knowledge of den structure is good, what their lives in these structures. The reproductive success is less understood is how black bears choose a patch, and survivability of bears has been linked to avail- stand, or landscape as their denning area. Landscape- ability of adequate den sites although this has never scale habitat selection may be in¯uenced by a variety been experimentally demonstrated (Hamilton and Mar- of factors, including the density of large decadent trees, chinton 1980, Alt 1984, Oli et al. 1997, McDonald and topography, climate, and elevation. Bears in coastal Fuller 1998). British Columbia select for middle-elevation land- Although black bears are reported to choose den sites scapes, late-successional stands, and patches with high opportunistically, bears living in wet environments structural complexity contributed by coarse woody de- have fewer options when it comes to ®nding dry, ther- bris (Davis 1996). In a review, Linnell et al. (2000) mally ef®cient den sites and may depend on the pres- reported that black bears select for 20Њ±50Њ slopes in ence of suitable trees (Alt 1984, Davis 1996). In these middle altitudes where high topographic variation ex- areas, caves and excavated dens often do not provide ists. Schwartz et al. (1987) found no signi®cant dif- ference among aspect groups but found dens to be more Manuscript received 2 December 2003; revised 15 July 2004; common in steep areas of south-central Alaska and less accepted 30 August 2004; ®nal version received 6 December 2004. Corresponding Editor: D. J. Mladenoff. common in low, wet areas subject to spring ¯ooding. 4 Present address: USDA Forest Service, Eastern Region, At the landscape scale, the abundance of large, often 626 East Wisconsin Avenue, Milwaukee, Wisconsin 53201 hollow, trees may be in¯uenced by natural disturbance USA. E-mail: [email protected] (Pickett and White 1985). The mapping of prevailing 5 Present address: USDA Forest Service, White River Na- tional Forest, P.O. Box 720, Eagle, Colorado 81631 USA. disturbance regimes may play an important role in pre- 6 Present address: USDA Forest Service, Superior Nation- dicting bear denning habitat, especially if the regimes al Forest, P.O. Box 2159, Tofte, Minnesota 55615 USA. in¯uence the quality and quantity of den structures. In 1306 August 2005 WIND DISTURBANCE AND BLACK BEAR DENS 1307 concert with other agents of forest disturbance, such as landslides, avalanches, insects, fungi, and snow breakage, wind plays a fundamental role in shaping forest structures in the temperate rain forests of southeastern Alaska (Harris 1989, Veblen and Alaback 1996, Nowacki and Kramer 1998). Forests most susceptible to maritime cyclonic storm damage are characterized by recurrent small-scale disturbances punctuated by oc- casional large and severe wind storms with blow-down patches averaging 7.4 ha and ranging up to 70.8 ha (Harris 1989). Kramer et al. (2001) found that in southeastern Alaska, forests most prone to storm winds may never reach an old-growth stage of forest development before the next catastrophic storm occurs. By contrast, forests protected from maritime windstorm disturbances tend to harbor trees with more old-growth characteristics, regulated by small-scale, low-intensity gap- phase mortality (Kramer et al. 2001). FIG. 1. Sequence of the analyses used to examine the Few studies have investigated how large mammals relationships among windstorm exposure, forest structure, respond to natural disturbance regimes (e.g., Gaillard and bear den selection. et al. 2003). An improved understanding of the factors that in¯uence den site selection at the landscape scale, Study areas such as disturbance regimes, can lead to the development of predictive habitat models for land use planning. Southeastern Alaska provides an excellent environ- This information is needed for understanding the ef- ment for investigating the in¯uence of a single distur- fects of forest management on bear habitat and de- bance agent, windthrow, on black bear winter den site signing prescriptions that conserve important bear win- selection at the landscape scale. The primarily pristine tering habitat. coastal rain forests of the region are largely regulated This study examined the in¯uence of maritime wind- by natural disturbance regimes. The shallow soils that storm activity on winter den site selection at the land- typify this area make these forests particularly suscep- scape scale in the coastal temperate rain forests of tible to windthrow from maritime windstorms. Wind- southeastern Alaska, USA. The objectives were to de- storms in southeastern Alaska occur as a result of ex- termine: (1) whether forests most protected from mar- tratropical cyclones that pass as frequently as every itime windstorm disturbances tend to harbor more trees four or ®ve days during late fall and winter (Shumacher suitable for bear dens; and (2) whether bears select for and Wilson 1986). These storms produce winds up to these areas in locating den sites. To accomplish these and occasionally exceeding 40 m/s, persistent cloud objectives, we evaluated a wind disturbance model cover, and up to4mofannual precipitation in the (WINDSTORM) against actual measures of forest coastal mountains. Forest ®res and large-scale insect structure patterns. We then used the model in con- outbreaks are uncommon in the region and do not pre- junction with black bear den site locations to investi- sent a confounding factor for the study of wind dis- gate the relationship between forest structure patterns turbance on forest structure and habitat selection. and bear den locations. We used black bear den locations from two study areas in southeastern Alaska (Fig. 2), Mitkof Island (56Њ38Ј N, 132Њ50Ј W) (Erickson et al. 1982) and Anan METHODS Creek watershed (56Њ10Ј N, 131Њ52Ј W) (Robertsen This project is a multidisciplinary synthesis of sev- 1998). Both areas are centrally located in the Tongass eral information sets; Fig. 1 provides an overview of National Forest and are representative of the coastal the analysis process. This analysis used two sets of temperate rain forest found in the archipelago and bear den locations, gathered as parts of other studies, coastal mainland, respectively. Mitkof Island is slightly to de®ne study areas. Subsequently the model WIND- more than 518 km2 in area and occurs within the Wran- STORM was run on these study areas. To identify wind gell Narrows Metasediments Ecological Subsection disturbance areas the model classi®cations were vali- characterized by clusters of glacially rounded moun- dated with an empirical independent ®eld data set col- tains separated by broad U-shaped valleys (Nowacki lected in one of the study areas. The bear den locations et al. 2001). The Anan Creek watershed is located ap- were classi®ed by wind disturbance categories. Com- proximately 45 km south of the town of Wrangell on positional analysis was used to determine whether the northern end of the Cleveland Peninsula and 64 km bears from the two study areas preferentially select southeast of Mitkof Island. Two ecological subsections landscapes based on disturbance regime. bisect this watershed: Eastern Passage Complex and Ecological Applications 1308 EUGENE J. DEGAYNER ET AL. Vol. 15, No. 4

Bell Island Granitics (Nowacki et al. 2001). Charac- storm susceptibility model (WINDSTORM; Kramer et terized by steep rugged terrain, elevation in the study al. 2001) across the forested landscape for each study areas extends from sea level to 1125 m. Alpine mead- area. WINDSTORM uses four abiotic factors (slope, ows are extensive at higher elevations, whereas dense elevation, soil stability, and exposure to prevailing forests cover the coastlines and mountain slopes. Six storm winds) to spatially predict long-term patterns of conifer species dominate the region (Pawuk and Kis- windthrow. The model was developed and tested on singer 1989). On well-drained sites, productive western nearby Kuiu and Zarembo islands. On Zarembo Island, hemlock (Tsuga heterophylla (Raf.) Sarg.) and Sitka the model correctly classi®ed 72% of both windthrow- spruce forests (Picea sitchensis (Bong.) Carr.) are com- prone and non-windthrown forest. Wind direction for mon. On less well-drained sites, hemlock and spruce maritime storm activity was taken to be 160Њ±220Њ still dominate, but are mixed with Alaska yellow cedar (Kramer et al. 2001). Model output generated three (Chamaecyparis nootkatensis (D. Don) Spach) and broad categories of storm susceptibility: storm prone, western red cedar (Thuja plicata Donn ex D. Don). storm protected, and intermediate susceptibility to Above 400 m, mountain hemlock (Tsuga mertensiana wind damage. (Bong.) Carr.) typically replaces western hemlock. Low-productivity mixed conifer-scrub forests, often Model evaluation dominated by lodgepole pine (Pinus contorta Dougl. We evaluated model predictions and described forest ex Loud. var. contorta), occur extensively on the land- characteristics of storm-prone and storm-protected for- scape, along with muskeg plant communities on low- ests on Mitkof Island (Fig. 3). Twenty-®ve ®xed-radius productivity hydric soils or wetlands (Pojar and White ®eld plots (0.1 ha) were located in these two landscape 1994, Alaback 1996). Sitka spruce is more abundant types by using a strati®ed random design. We only in stands that have experienced disturbance. Sitka placed plots in stands Ͼ15 ha and within 2 km of the spruce is a long-lived seral species that will, without road network. In each plot, we recorded dbh, tree spe- disturbance, eventually succumb to more shade-toler- cies, and estimated canopy position (dominant/codom- ant western and mountain hemlock (Eyre 1980). Dis- inant, intermediate) of all overstory-standing trees turbance (windthrow and ¯ood scouring) plays a key (dead or alive with dbh Ͼ12 cm). We did not include role in the maintenance of Sitka spruce in mixed stands. suppressed trees in the standing structure analysis, al- Indeed, Sitka spruce prefers regenerating on mineral though they were counted. We cored 10±20 overstory seedbeds and open conditions that these disturbances trees of representative diameter classes in each plot to provide (Barrett 1995). Past logging activity consists determine age. Annual rings on the cores were counted primarily of small-scale selective logging along the in the laboratory using a dissection microscope. When saltwater shoreline in the Anan watershed, whereas ap- necessitated by obscured rings, we mounted and sanded proximately 20% of the original productive forest has the cores and then counted rings. been logged on Mitkof Island. We calculated standing live and dead trees per hect- Mitkof Island telemetry data set.ÐThe Mitkof den are, the standard deviation, and mean diameter for each site data originated as a 2-yr habitat use and den ecol- plot. We summarized stand structure by plotting trees ogy study in the early 1980s (Erickson et al. 1982; B. per hectare (TPHA) against quadratic mean diameter M. Hansen and J. G. Doerr, unpublished manuscript). (QMD). We generated frequency distributions of tree This study located 16 black bear winter dens by ground sizes. We estimated the number of large decadent (dead radio-tracking marked bears to winter den sites. All or hollow) trees based on visual observations recorded dens were associated with large woody structures. in the ®eld, including visual characteristics (standing Anan telemetry data set.ÐThe Anan den site data dead trees or visual evidence of rot), and cored trees came from a black bear behavior study conducted near found to be hollow. We compared the standard devi- the Anan Wildlife Viewing Area from 1993 to 1996 ation of diameters from all plots located in windstorm- (Chi and Gilbert 1996). Thirteen black bears were cap- prone with those from windstorm-protected forests. We tured and radio-collared in the Anan Creek vicinity clustered data from all the ®xed radius plots (from both from 25 to 31 July 1993. One or two ¯ights per winter windthrow prone and protected areas) using a Che- identi®ed den locations for these bears. Den locations bychev distance metric (Michalski et al. 1981). Clus- were recorded on aerial photos (1973, 1:15 000) and tering was based on structural and age characteristics digitized using high-resolution orthophotoquads (1: of the ®xed-radius plots to provide a more detailed 36 680). Only three of these den locations were veri®ed description of age and structural characteristics of for- by walk-in ground telemetry; all were found to be with- ests in various stages of recovery from catastrophic in trees. and partial canopy disturbance (Kramer et al. 2001). Four distinct clusters were selected, ranging from pre- Modeling maritime windstorm effects on Mitkof dominantly young even-aged stands to old all-aged for- Island and Anan study areas ests. Stand age and size characteristics in relation to To predict long-term effects of maritime windstorm WINDSTORM predictions were not investigated for activity on the forest landscape we ran a maritime wind- the Anan study area. August 2005 WIND DISTURBANCE AND BLACK BEAR DENS 1309

FIG. 2. (a) The State of Alaska vicinity map. (b,c) Relief maps for (b) Mitkof Island and (c) Anan Creek study areas, with bear den locations shown as small dots.

Bear den locations relative to We used the composition of these buffers to describe WINDSTORM predictions den sites in the statistical analysis. To avoid double-counting dens that were reused from We used the 16 den locations of 13 bears from Mitkof year to year, unveri®ed den locations (located from the Island and the 15 den locations of 11 bears from Anan air, but not ®eld veri®ed) for the same bear Ͻ305 m Creek. We incorporated these data into a geographic apart were spatially averaged and treated as one den. information system (GIS) and overlaid them on WIND- Using this rule, dens reused by the same bear in dif- STORM model output. We assigned a storm suscep- fering years were classed as one den. However, two tibility category to each den location. In southeastern dens from the same bear Ͻ305 m apart could be mis- Alaska, estimated median observer error is 152 m for taken for a single den using this methodology. telemetry ¯ights in similar terrain (J. Doerr, unpub- For use/availability analyses, the Anan Creek study lished data). Since spatial accuracy of den locations area is de®ned as the minimum convex polygon that and data within the GIS are imperfect, we buffered den enclosed the winter dens and expanded by 457 m to sites by 152 m (7.24 ha) to develop an error polygon. easily accommodate the 152-m den site buffer (Fig. 2). Ecological Applications 1310 EUGENE J. DEGAYNER ET AL. Vol. 15, No. 4

4) indicative of old-growth conditions (Caouette et al. 2000). Large trees (Ͼ88 cm) were twice as dense in storm- protected forests (28 Ϯ 6 trees/ha) than in unprotected forests (14 Ϯ 5 trees/ha, P Ͻ 0.01). Forested stands sampled in storm-prone areas showed greater evidence of past catastrophic disturbance. We detected evidence of either catastrophic or partial canopy disturbance from windthrow in 85% of the storm-susceptible plots vs. 50% in storm-protected. Tree core data indicate that most of the even-aged windthrow originated from a storm that occurred approximately 200 yr ago, although some forests originated from more recent (150 yr ago) and older (250 and 300 yr ago) storms. The four Chebychev forest-type clusters yielded stands ranging from predominantly young even-aged stands to old all-aged forests. All plots in cluster 1 (n ϭ 4) were located in areas of storm-prone landscapes as predicted by WINDSTORM and showed strong evidence of recent (Ͻ200 yr) catastrophic or multiple partial-canopy events. Ages were tightly grouped (Fig. 5), and most of the trees in the stand were small (0± 68 cm). The high number of small, standing dead trees in these plots (Fig. 6) suggests that self-thinning may be a dominant mechanism for mortality. These stands had a high density of trees (440 TPHA). Nearly 30% FIG. 3. Output from WINDSTORM for Mitkof Island and locations of ®xed-radius plots collected in storm-susceptible of the trees were spruce. Stands in this cluster represent and storm-protected locations. an early-to-middle stage of stand development. Plots in cluster 2 (n ϭ 8) showed evidence of an old (Ͼ150 yr) catastrophic or recent partial canopy distur- For the Mitkof Island study area, natural water bound- bance, resulting in some tight clustering of age groups aries were available for de®ning available habitat. We (Fig. 5). Few large trees (Ͼ88 cm) exist in these stands. de®ned available habitat as the portion of the island Spruce abundance was still high (30%). These stands north and east of Blind Slough omitting the residential may have experienced one or more partial-canopy dis- area of Petersburg. We compared the composition of turbance events. Forests in this cluster are in an inter- storm susceptibility (high, medium, low) for the pro- mediate stage of forest development characterized by ductive forest for each study area with the storm sus- many stands being 200 yr old and some being multi- ceptibility compositions of the 7.24-ha buffers around each bear den (Aebischer et al. 1993) using RESELECT software to complete the compositional analysis (Leban 1994).

RESULTS Maritime windstorms and forest structure on Mitkof Island Our analysis elucidated differences in stand age and structural characteristics between storm-prone and storm-protected forests. Forest age characteristics were more homogenous in areas of high storm susceptibility than in areas of low storm susceptibility (SD of dbh ϭ 19 cm [SE ϭ 1.7] vs. SD of dbh ϭ 26.5 cm [SE ϭ 2], comparison of SD of dbh, P Ͻ 0.01). Tree density was greater (means Ϯ SE, 335 Ϯ 27 trees/ha vs. 220 Ϯ 12.5 trees/ha, P Ͻ 0.01) and stem size was signi®cantly smaller in forests most susceptible to maritime windstorms (dbh, 47 Ϯ 2 cm vs. 56.5 Ϯ 2.4 cm, P Ͻ 0.01). FIG. 4. Stem size (quadratic mean diameter, QMD) vs. Scatterplots from all plots (n ϭ 25) show the higher stem density from 0.1-ha ®xed-radius plots. Key: ®lled cir- QMD and lower TPHA in storm-protected plots (Fig. cles, storm prone; open triangles, storm protected. August 2005 WIND DISTURBANCE AND BLACK BEAR DENS 1311

FIG. 5. Percentage of sampled trees in age intervals for each cluster on Mitkof Island. Clusters 1 through 4 show increasing age complexity from even-aged to all-aged stand structure. aged. Large standing dead trees are not common in these stands. Competitive mortality processes are still occurring in the stem exclusion stage (Oliver and Lar- son 1996), as evidenced by the abundance of small, dead trees. Most of the plots in this cluster (63%) were located in storm-susceptible locations. Plots in cluster 3 (n ϭ 9) showed evidence of many small-scale disturbances that contributed to present- day stand structure and age attributes. The majority of these plots (68%) occurred in WINDSTORM storm- protected locations. Ages in these plots spanned a wide range (Fig. 5), and diameter distributions were uniform (Table 1). Spruce trees comprised less than 10% of the overstory. This cluster had more large, standing dead and live trees than clusters 1 or 2. In addition, trees cored for age determination displayed more evidence of heart rot. These forests have old-growth characteristics. Plots in cluster 4 (n ϭ 3) showed evidence of some partial-canopy disturbance and multiple small-scale FIG. 6. Frequency of live stem (dbh) classes and fre- disturbance events. All of the plots were located in quency of spruce, small dead trees, and large dead trees, by storm-protected locations. Ages were variable (Fig. 5), cluster, on Mitkof Island. and most of the diameter distributions were negative exponential with a wide distribution of stem sizes. Most plots contained large, standing dead trees, but spruce as exhibiting age and structural characteristics indic- trees were uncommon (4%). ative of later stage forests. These plots were most abun- Based on the stand structural differences described, dant in WINDSTORM-modeled storm-protected land- we characterized plots in clusters 3 and 4 (12 of 25) scapes. Plots in clusters 1 and 2 were generally con-

TABLE 1. WINDSTORM predictions for plot storm exposure and measures of forest structure for each Chebychev cluster in two study areas in southeastern Alaska, USA.

No. plots Tree Density of large Percentage No. plots storm- Forest in density trees (Ͼ88 cm) of spruce Age of protected vs. development Cluster cluster (no. trees/ha) (no. trees/ha) (%) trees storm-prone² stage 1 4 440 15 28 youngest 0 vs. 4 early-mid 2 8 274 20 30 intermediate 3 vs. 5 mature 3 9 247 26 10 old 6 vs. 3 gap-phase 4 3 187 23 4 oldest 3 vs. 0 gap-phase ² According to WINDSTORM. Ecological Applications 1312 EUGENE J. DEGAYNER ET AL. Vol. 15, No. 4

TABLE 2. Use vs. availability analysis of storm exposure classes of bear dens by model WINDSTORM for the two study areas.

Storm damage Area of Relative Relative use susceptibility productive availability w/error Use/ class forest (ha) (%) No. dens polygon availability Mitkof Island Protected 8192 35 8 57.1 1.6 Moderate 9181 39 7 34.6 0.9 Exposed 6096 26 1 8.2 0.31 Anan Protected 2450 52 10 69.4 1.3 Moderate 1352 29 4 24.5 0.86 Exposed 915 19 1 6.1 0.31 Note: Availability was restricted to productive forest lands capable of producing 1.4 m3 wood ®ber´haϪ1´yrϪ1.

sistent with stands in early-to-middle succession re- DISCUSSION covering from a disturbance event. These plots were Our analysis suggests that the WINDSTORM model most abundant (9 of 12) in storm-exposed landscapes. does predict forest structure and that spatially explicit modeled wind disturbance regimes are useful in pre- Bear selection of denning sites dicting black bear winter den sites at the landscape Only two of 31 winter bear dens were found in storm- scale. exposed habitats, one in each study area (Table 2). Both dens were Ͻ50 m from a moderate/low storm exposure Windstorm effects on stand structure polygon boundary. on Mitkof Island Storm-protected habitats had 1.6 and 1.3 times the Forests in storm-susceptible landscapes showed expected number of dens, based on the relative avail- strong evidence of past windstorm damage. The ma- ability of this habitat in the Mitkof and Anan study jority of these stands originated from a catastrophic areas, respectively. Storm-prone areas in both study storm that occurred approximately 200 years ago. The areas had only 31% of the expected number of dens effects of past windstorms on forest age and structural based on the relative availability of this habitat type. characteristics were complicated by partial mortality at The use-relative-to-availability analysis demonstrat- the time of disturbance, recruitment of advance regen- ed that bear dens are not randomly located with regard eration trees from the understory, and tree mortality to windstorm susceptibility in the Anan (P Ͻ 0.001) after the disturbance event (Fox 1989, Franklin et al. and Mitkof (P Ͻ 0.013) study areas. Den selection is 2002). Overall, maritime storm damage resulted in positively related to areas most protected from wind- more homogenous stand conditions than those found storms and negatively related to areas most susceptible in storm-protected landscapes. Forests in storm-sus- to windstorm damage (P Ͻ 0.05). Landscapes with ceptible areas characteristically showed signs of early- moderate storm susceptibility were selected over to-middle developmental stages and may remain per- storm-exposed landscapes in the Anan study area (P petually in these early stages as a consequence of fre- Ͻ 0.05), but similar selection was not found in the quent and intense maritime storm winds (Nowacki and Mitkof study area (P Ͼ 0.05). The relative orders of Kramer 1998, Kramer et al. 2001). storm exposure selection rankings were similar for each Forests more protected from maritime storm winds study area (Table 3). showed limited evidence of catastrophic disturbance.

TABLE 3. Pairwise comparison and ranking of habitats using compositional analysis for the ϭ Ͻ ϭ Ͻ Anan (F2, 13 19.77, P 0.001) and Mitkof (F2, 14 5.27, P 0.0134) study areas.

Anan Mitkof Storm damage susceptibility Protected Moderate Exposed Rank Protected Moderate Exposed Rank Protected ϩ* ϩ*2 ϩ* ϩ*2 Moderate Ϫ* ϩ*1 Ϫ* ϩ NS 1 Exposed Ϫ* Ϫ*0Ϫ* Ϫ NS 0 Notes: The greatest value is the highest rank. Plus signs indicate selection for a habitat, either signi®cant (* P Ͻ 0.05) or nonsigni®cant (NS); minus signs indicate selection against a habitat, signi®cant (* P Ͻ 0.05) or nonsigni®cant (NS). August 2005 WIND DISTURBANCE AND BLACK BEAR DENS 1313

We found greater evidence of partial canopy and small- scale gap disturbances in these stands. Large (Ͼ88 cm) standing dead and live trees were more common in these landscape settings. Heart rot and other noncom- petitive mortality processes likely play a stronger role in determining stand characteristics (density, tree size, and number of standing dead trees) in protected areas. Though some trees were impossible to age due to hollow boles or heart rot decay, those larger trees that could be aged were old (Ͼ300 years old). These forests tended to be in later stages of development and main- tained a wider range of tree ages and diameters with low stocking. Maritime windstorm in¯uence on winter den site selection Large old trees with complex ``stilted'' root systems and large hollow trees are important habitat features for black bear den sites. Tree dens provide better protection from disturbances, weather, predation, and ¯ooding and are more energetically ef®cient than other types of dens (Oli et al. 1997). In areas where soil conditions are not suitable for excavating hibernacula, black bears may be dependent on old-growth timber stands for denning sites (Beecham et al. 1983). Our data suggest that these features are most abundant in landscapes with relatively infrequent wind disturbance events and that black bears prefer these structures for winter den sites. These sheltered landscapes favor persistence of very large old trees that are structurally vulnerable to wind damage (Fig. 7). By contrast, forest FIG. 7. Defect in a hemlock tree that allows bears access landscapes most susceptible to recurrent maritime to the hollow core. This tree was used for denning by an adult female and two yearling cubs. Note the susceptibility of the windstorms offered fewer potential den sites, such as tree to stem breakage due to the structural weakness of the large and hollow trees. bole. (Photo by A. Erickson.) Large fallen trees are important for the creation of denning sites. These trees are frequently associated with stem breakage that creates access to otherwise otic factors such as cooler climates, deeper snow pack, sealed heart rot hollows (Fig. 8) and provide large nurse and less windy microclimates commonly found on logs of suf®cient size to create trees with large stilted north-facing (and wind-protected) slopes could also in- root systems. Dens in stilted root systems are formed ¯uence den site selection. as the nurse log or wad decays. This study did not present data on whether abun- Heart rot fungi are essential for the creation of hol- dance of suitable bear dens actually in¯uence bear den- low trees suitable as denning sites for black bear. Ad- sity or demography in southeastern Alaska. However, vanced encroachment of heart rot fungi is likely in trees we demonstrated that bear dens are not randomly lo- more than 200 years old in both storm-protected and cated across the landscape, and this implies that bears storm-prone forests (Hennon 1995). However, we sus- are attempting to meet some metabolic or security re- pect that large hollow trees are less abundant in forests quirement. Bears frequently traveled outside their sum- susceptible to maritime storm winds. A larger propor- mer home range to select winter dens, another indi- tion of the trees are more than 200 years old in storm- cation that bears were selecting for certain habitats. protected landscapes; and large hollow trees with sig- Reducing the availability of high-quality dens may in- ni®cant fungal decay are particularly susceptible to ¯uence populations, but this paper cannot suggest at stem snap and breakage from maritime storms (Fig. 7). what threshold changes in demographics would man- We believe that the selection of wind-protected land- ifest. scapes for denning is likely related to the abundance of large and hollow trees. Management implications However, there are confounding factors with respect Conservation of den structures and den sites may be to bear den site selection. A limitation of our study is most important in areas where black bears rely on trees that we could not determine to what degree other abi- for denning (Beecham et al. 1983, Linnell et al. 2000), Ecological Applications 1314 EUGENE J. DEGAYNER ET AL. Vol. 15, No. 4

FIG. 8. A bear den in a hollow hemlock log with heart rot in a recent clearcut on Mitkof Island. Management practices that encourage leaving of legacy trees may be bene®cial to bears. (Photo by A. Erickson.) as is the case in southeastern Alaska. Better under- timber sale by reducing logging costs associated with standing of the relationships between windstorm dam- handling low-value logs. age gradients, disturbance regimes, and the creation Planners should take into account the natural dis- and renewal of habitat features used by bears can in- turbance regime when designing and evaluating alter- form the creation of habitat conservation areas and en- native management and conservation scenarios (Han- sure long-term habitat capability of bear habitat. Our sen et al. 1996, Nowacki and Kramer 1998). To con- study suggests that wind disturbance modeling is a use- serve denning habitat for bears in southeastern Alaska, ful tool for identifying suitable bear denning areas. the timber harvest practice of clearcutting should be Timber harvest operations, especially clearcutting, avoided in those areas protected from maritime wind- may have a deleterious effect on the long-term avail- storms, where winter den sites occur in greater abun- ability of black bear den sites (Hansen 1988). Potential dance. den sites may be eliminated directly during harvest or decline in abundance afterward as the residual fallen Future work structures decay, leading to long-term den site short- ages within cutover areas. Managers may mitigate such Although the model WINDSTORM was developed impacts by leaving clumps of standing large and small using data from Kuiu and Zarembo Islands, it appeared trees within harvest units to ensure persistence and to be effective in predicting winter bear den locations future recruitment of large wood structure. In the short- in the Mitkof and Anan study areas. The model's ability term, clearcuts can create den sites, but conditions more to predict broad forest structure patterns and, in turn, favorable to denning bears can be created by requiring bear denning habitat demonstrates the relevance of nat- that large cull logs be left within the harvest unit (Fig. ural disturbance regimes for bear habitat management 8) (Erickson et al. 1982). This approach may be con- in southeastern Alaska. More effective identi®cation of sistent with timber harvest economic objectives since forests containing high concentrations of large, old, many of the large old-growth trees that may recruit into hollow trees that may serve as bear dens should be a potential black bear dens have little timber value due high priority for future research. Calibrating the to their poor form and high degree of wood defect. WINDSTORM model to speci®c study area conditions Leaving these trees for wildlife and general biodiver- may strengthen the association between wind mapping sity objectives may improve the economic value of a and bear den site selection. Secondly, future work is August 2005 WIND DISTURBANCE AND BLACK BEAR DENS 1315 needed to determine the importance of high-quality Hansen, B. M. 1988. Habitat use and den ecology of black bear dens on bear demography. bears on Mitkof Island, southeast Alaska. Thesis. Univer- sity of Washington, Seattle, Washington, USA. ACKNOWLEDGMENTS Harris, A. S. 1989. Wind in the forests of southeast Alaska We acknowledge Doug Larson and Lavern Beier of the and guides for reducing damage. USDA Forest Service, Alaska Department of Fish and Game, Division of Wildlife General Technical Report PNW-GTR-244. Conservation for capturing and radio marking Anan study Hayes, S. G., and M. R. Pelton. 1994. Habitat characteristics bears and conducting telemetry ¯ights. Dennis Chester ad- of female black bear dens in northwestern Arkansas. In- ministered the Anan study, conducted telemetry ¯ights, and ternational Conference on Bear Research and Management prepared the Anan data set. Terry Shaw, Chris Iverson, Me- 9:411±418. lissa Cady, Greg Nowacki, and Lance Craighead provided Hennon, P. E. 1995. Are heart rot fungi major factors of encouragement and reviewed drafts of the manuscript. disturbance in gap-dynamic forests? Northwest Science 69: 284±293. LITERATURE CITED Johnson, K. G., D. O. Johnson, and M. R. Pelton. 1978. Aebischer, N. J., P. A. Robertsen, and R. E. Kenward. 1993. Simulation of winter heat loss for a black bear in a closed Compositional analysis of habitat use from animal radio- tree den. Pages 155±166 in Proceedings of Fourth Eastern tracking data. Ecology 74:1313±1325. Workshop on Black Bear Research and Management. Vol- Alaback, P. B. 1996. Biodiversity patterns in relation to cli- ume 4. Maine Department of Inland Fisheries and Wildlife, mate: the coastal temperate rainforests of North America. Greeneville, Maine, USA. Pages 105±133 in R. G. Lawford, P. B. Alaback, and E. Kramer, M. G., A. J. Hansen, M. Taper, and E. Kissinger. Fuentes, editors. High-latitude rainforests and associated ecosystems of the West Coast of the Americas. Ecological 2001. Abiotic controls on windthrow and natural forest studies. Volume 116. Springer-Verlag, New York, New dynamics in a coastal temperate rainforest. Ecology 82: York, USA. 2749±2768. Alt, G. L. 1984. Black bear cub mortality due to ¯ooding of Leban, F. 1994. Compositional analysis of habitat use: RE- natal dens. Journal Wildlife Management 48:1432±1434. SELECT. University of Idaho, Moscow, Idaho, USA. Barrett, J. W. 1995. Regional silviculture of the United States. Linnell, J. D., J. E. Swenson, R. Andersen, and B. Barnes. John Wiley and Sons, New York, New York, USA. 2000. How vulnerable are denning bears to disturbance? Beecham, J. J., D. G. Reynolds, and M. G. Hornocker. 1983. Wildlife Society Bulletin 28:400±413. Black bear denning activities and den characteristics in McDonald, J. E., and T. K. Fuller. 1998. Testing assumptions west-central Idaho. International Conference on Bear Re- in bear research: using statistical power analysis to estimate search and Management 5:79±86. effects of den type on black bear cub survival. Ursus 10: Caouette, J. P., M. G. Kramer, and G. J. Nowacki. 2000. 405±411. Deconstructing the timber volume paradigm in the man- Michalski, R. S., R. E. Strepp, and E. Diday. 1981. A recent agement of the Tongass National Forest. USDA Forest Ser- advance in data analysis: clustering objects into class char- vice, General Technical Report PNW-GTR-482. acterized by conjunctive concepts. Pages 33±56 in L. N. Chi, D. K., and B. K. Gilbert. 1996. Human±bear interactions Kanal and A. Rosen®ld, editors. Progress in pattern rec- at Anan Creek, Tongass National Forest, Alaska. Final re- ognition. Volume 1. North-Holland, New York, New York, port, USDA Forest Service Cooperative Agreement Num- USA. ber 93-265. Tongass National Forest, Stikine Area, Wran- Nowacki, G. J., and M. G. Kramer. 1998. The effects of wind gell, Alaska, USA. disturbance on temperate rain forest structure and dynamics Davis, H. 1996. Characteristics and selection of winter dens of Southeast Alaska. USDA Forest Service, General Tech- by black bears in coastal British Columbia. Thesis. Simon nical Report PNW-GTR-421. Fraser University, Burnaby, British Columbia, Canada. Nowacki, G. J., M. Shephard, P. Krosse, W. H. Pawuk, G. Erickson, A. W., B. M. Hanson, and J. J. Brueggeman. 1982. Fisher, J. Baichtal, D. Brew, E. J. Kissinger, and T. Brock. Black bear denning study, Mitkof Island, Alaska. Univer- 2001. Ecological subsections of Southeast Alaska and sity of Washington, School of Fisheries, Fisheries Research neighboring areas of Canada. USDA Forest Service, Tech- Institute, Seattle, Washington, USA. nical Publication R10-TP-75. Eyre, F. H. 1980. Forest cover types of the United States and Oli, J. K., H. A. Jacobson, and B. D. Leopold. 1997. Denning Canada. Society of American Foresters, Washington, D.C., ecology of black bears in the White River National Wildlife USA. Refuge, Arkansas. Journal Wildlife Management 61:700± Fox, J. F. 1989. Bias in estimating forest disturbance rates 706. and tree lifetimes. Ecology 70:1267±1272. Oliver, C. D., and B. C. Larson. 1996. Forest stand dynamics. Franklin, J., T. A. Spies, R. Van Pelt, D. Carey, D. Thorn- Updated edition. McGraw-Hill, New York, New York, burgh, D. R. Berg, D. Lindenmayer, M. Harmon, W. Kee- USA. ton, and C. G. I. Shaw. 2002. Disturbances and structural Pawuk, W. H., and E. J. Kissinger. 1989. Preliminary forest development of natural forest ecosystems with silvicultural implications, using Douglas-®r forests as an example. For- plant association of the Stikine Area, Tongass National For- est Ecology and Management 155:399±423. est. USDA Forest Service, Technical Publication R10-TP- Gaillard, J. M., P. Duncan, D. Delorme, G. Van Laere, N. 72. Pettorelli, D. Maillard, and G. Renaud. 2003. Effects of Pickett, S. A., and P. S. White. 1985. The ecology of natural hurricane Lothar on the population dynamics of European disturbance and patch dynamics. Academic Press, New roe deer. Journal of Wildlife Management 67:767±773. York, New York, USA. Hamilton, R. J., and R. L. Marchinton. 1980. Denning and Pojar, J., and P. S. White. 1994. Plants of coastal British related activities of black bears in the coastal plain of North Columbia. Lone Pine Publishing, Edmonton, Alberta, Can- Carolina. International Conference on Bear Research and ada. Management 4:121±126. Robertsen, P. A. 1998. Resource report for the Canal Hoya Hansen, A. J., R. Patten, E. J. DeGayner, and B. L. Marks. Timber SaleÐAnan Bears. USDA Forest Service, Wrangell 1996. Simulating forest and habitat change in south-east Ranger District, Wrangell, Alaska, USA. Alaska within landscape model PAYSAGE. Transactions Schwartz, C. C., S. D. Miller, and A. W. Franzmann. 1987. in GIS 1:119±136. Denning ecology of three black bear populations in Alaska. Ecological Applications 1316 EUGENE J. DEGAYNER ET AL. Vol. 15, No. 4

International Conference on Bear Research and Manage- Veblen, T. T., and P. B. Alaback. 1996. A comparative review ment 7:281±292. of forest dynamics and disturbance in the temperate rain- Shumacher, U. D., and J. D. Wilson. 1986. On the atmo- forests of North and South America. Pages 173±213 in R. spheric and oceanic environment of the Gulf of Alaska. G. Lawford, P. B. Alaback, and E. Fuentes, editors. High- Pages 135±178 in Proceedings of a workshop on compar- latitude rainforests and associated ecosystems of the west ative biology, assessment, and management of Gadoids coast of the Americas. Ecological studies. Springer-Verlag, from the North Paci®c and Atlantic Oceans. National Oce- anic and Atmospheric Administration, Paci®c Marine En- New York, New York, USA. vironmental Laboratory, Seattle, Washington, USA. Weaver, K. M., and M. R. Pelton. 1994. Denning ecology of Smith, T. R. 1985. Ecology of black bears in the bottomland black bears in the Tensas River Basin of Louisiana. Inter- hardwood forest in Arkansas. Dissertation. University of national Conference on Bear Research and Management 9: Tennessee, Knoxville, Tennessee, USA. 427±433.