Den Structure and Selection of Denning Habitat by Brown Bears in the Romanian Carpathians

Den structure and selection of denning habitat by brown bears in the Romanian Carpathians

Authors: Iosif, Ruben, Pop, Mihai I., Chiriac, Silviu, Sandu, Radu M., Berde, Lajos, et. al. Source: Ursus, 2020(31e5) : 1-13 Published By: International Association for Bear Research and Management URL: https://doi.org/10.2192/URSUS-D-18-00010.1

BioOne Complete (complete.BioOne.org) is a full-text database of 200 subscribed and open-access titles in the biological, ecological, and environmental sciences published by nonprofit societies, associations, museums, institutions, and presses.

Your use of this PDF, the BioOne Complete website, and all posted and associated content indicates your acceptance of BioOne’s Terms of Use, available at www.bioone.org/terms-of-use.

Usage of BioOne Complete content is strictly limited to personal, educational, and non - commercial use. Commercial inquiries or rights and permissions requests should be directed to the individual publisher as copyright holder.

BioOne sees sustainable scholarly publishing as an inherently collaborative enterprise connecting authors, nonprofit publishers, academic institutions, research libraries, and research funders in the common goal of maximizing access to critical research.

Downloaded From: https://bioone.org/journals/Ursus on 18 Apr 2020 Terms of Use: https://bioone.org/terms-of-use Access provided by International Association for Bear Research and Management Den structure and selection of denning habitat by brown bears in the Romanian Carpathians

Ruben Iosif1, Mihai I. Pop2,3, Silviu Chiriac4, Radu M. Sandu4,LajosBerde5, Szilárd Szabó6, Lauren¸tiu Rozylowicz2, and Viorel D. Popescu2,7,8

1Faculty of Natural and Agricultural Sciences, University Ovidius Constanta, 1 Aleea Universitatii, Building B, Constan¸ta, 900470 Romania 2Centre for Environmental Research (CCMESI), University of Bucharest, 1 N. Balcescu, Bucharest, 010041 Romania 3Association for Biodiversity Conservation (ACDB), 12 Ion Creanga, Focsani, 620083 Romania 4Vrancea Environmental Protection Agency, 2 Dinicu Golescu St., Focsani, 620106 Romania 5Covasna Environmental Protection Agency, 10 Grigore Balan St., Sf. Gheorghe, 520082 Romania 6Harghita Environmental Protection Agency, 43 Márton Áron St., Miercurea Ciuc, 530211 România 7Department of Biological Sciences, Ohio University, 107 Irvine Hall, Athens, OH 45701, USA

Abstract: The Romanian Carpathian Mountains provide one of the largest areas suitable for brown bear (Ursus arctos) in Europe, but the long history of logging has reduced old-growth forest to fragments. Continuous timber extraction, along with new recreation opportunities from motorized vehicles, may affect brown bear denning habitat through disturbance. As such, understanding den site selection at the landscape and local levels is important for the conservation and sustainable management of the Romanian brown bear population. We used data on 86 den sites collected between 2010 and 2013 in Southeastern Carpathians and developed Resource Selection Functions for second-order (landscape- level) and third-order (local level) den habitat selection, using habitat structure and topographic attributes of den locations. The altitude of dens ranged between 440 and 1,320 m, with a mean slope of 19.7 ± 0.8%. Aspect was evenly distributed between southwest (22.9% of the dens), east (20.5%), south (18.1%), and southeast (15.7%). Dens stabilized by boulders were dominant (68%), and had maximum mean length = 149.2 ± 5.6 cm, width = 109.1 ± 4.8 cm, and height = 113.8 ± 10.5 cm. At both local and the landscape scales, bears selected for steeper slope and higher percent coverage of mixed (beech–ﬁr–spruce [Fagus sylvatica–Abies alba–Picea abies]) forest. At the landscape scale, bears also selected dens at higher altitude, and with greater coverage of old forest, and away from urban areas and recent clear-cuts. Our spatial predictions have the potential to inform forest management by identifying areas where disturbance of brown bear denning habitat should be avoided or limited, thus contributing to brown bear management and conservation planning in the Romanian Carpathians.

Key words: brown bear, conservation planning, den selection, disturbance, habitat suitability, resource selection function, Romanian Carpathians, Ursus arctos

DOI: 10.2192/URSUS-D-18-00010.1 Ursus 31:article e5 (2020)

Large carnivores are recovering in the European ments (Di Minin et al. 2016). Such spatial predictions human-dominated landscape and efforts are made toward have potential to assist mitigation measures, such as reg- coexistence with humans (Chapron et al. 2014, Zarzo- ulating forestry practices in sensitive areas for conserva- Arias et al. 2018). However, large carnivore habitat con- tion (Pop et al. 2018b), creating green infrastructure to nectivity is threatened by habitat loss and fragmentation, ensure connectivity over barriers (Mateo-Sánchez et al. especially in the context of rapid deforestation (Ripple 2014, Ziółkowska et al. 2016), and finally helping prior- et al. 2014). To assist conservation and management poli- itize conservation decisions from an economic point of cies, conservationists use landscape planning to delineate view while reducing conflicts between humans and large and rank landscape features according to species require- carnivores (Rondinini and Boitani 2007). Large carnivores have high dispersal capabilities and large space-use requirements that are influenced by food 8email: [email protected] availability (Hertel et al. 2016), challenging coexistence

Downloaded From: https://bioone.org/journals/Ursus on 18 Apr 2020 Terms of Use: https://bioone.org/terms-of-use Access provided by International Association for Bear Research and Management r 2BROWN BEAR DEN HABITAT IN THE CARPATHIANS Iosif et al.

with humans in fragmented landscapes (Clevenger et al. denning season); therefore, we expect such activities to 1987, Rozylowicz et al. 2011, Skrbinšek et al. 2012). For exert a high level of disturbance on denning habitat. In hibernating species, such as brown bears (Ursus arctos), addition to logging disturbance, humans are increasingly fragmentation may affect the quality of denning habitat using forest landscape for recreational activities deep into and bear safety during hibernation (Petram et al. 2004, previously undisturbed areas (e.g., off-road vehicles; see Sahlén et al. 2011, Pigeon et al. 2016). Brown bears have Goldstein et al. 2010). high sensitivity to any form of human disturbance during The goal of this study is to identify and map high- this biological stage, and they require contiguous habitats quality denning areas for brown bears in the Southeast- with limited access for humans (Linnell et al. 2000). Dur- ern Carpathians to fill a critical knowledge gap for bear ing pre-denning and denning season, brown bears select conservation, and to provide information for future man- habitat with steeper slopes and remote old-aged forest agement strategies of forest ecosystems in the Carpathi- stands (Pigeon et al. 2014, Pop et al. 2018b). Any distur- ans. Spatial models have the potential to assist managers bance during hibernation may eventually lead to reduced to regulate potential disturbance in highly sensitive den- fitness of individuals (Evans et al. 2016) or cub aban- ning areas. Such models can vary from simple habitat donment and mortality (i.e., pregnant females give birth suitability indices (Kaminski et al. 2013) to frameworks inside the dens, where they feed and shelter cubs until that incorporate species connectivity needs (Pop et al. the spring; Swenson et al. 1997, Linnell et al. 2000). For 2018b), and stochastic predictions that account for sea- example, Swenson et al. (1997) showed that 9% of their sonality in ecology and behavior of species (Nakajima studied dens in Scandinavia were abandoned as a result et al. 2012). We used a data set of 86 den locations col- of human disturbances, with bears forced to move up to lected between 2010 and 2013, along with topographic, 30 km before denning again, thus resulting in a signifi- land cover, and disturbance factors to evaluate denning cant fitness cost. Sahlén et al. (2011) and Hodder et al. habitat selection at landscape and local levels (Johnson (2014) documented avoidance of denning close to roads 1980). Habitat selection is a hierarchical process that in- and areas with high levels of human activity. volves individual behaviors and decisions that ultimately Across the Carpathian Mountains, the transition from maximize individual fitness (Johnson 1980, Johnson et al. a socialist to a capitalist economy was a determinant fac- 2006). Specifically, our objectives were to 1) investigate tor affecting the efficacy of protected areas (Kuemmerle the characteristics of brown bear denning habitat at land- et al. 2008, Müller et al. 2009, Butsic et al. 2017). Ex- scape (second-order selection) and local levels (third- tensive logging that exceeds the maximum allowed patch order selection), 2) develop a spatial model for brown bear size of 3 ha was frequently highlighted through satellite den suitability at the landscape scale, and 3) provide man- image processing (Knorn et al. 2012), threatening valu- agement recommendations regarding brown bear denning able temperate forests in the Polish, Czech, and Romanian habitat in the Southeastern Romanian Carpathians. Carpathians (Griffiths et al. 2013). Although the Roma- nian Carpathians shelter the largest stands of old-growth forest in Europe, suitable for brown bear conservation Study area (Pop et al. 2018b), the long-term effects of habitat loss We conducted the study in the Southeastern Roma- on bear population viability are difficult to assess because nian Carpathian Mountains across 3 administrative units: of severe lack of data. Addressing this knowledge gap Vrancea, Covasna, and Harghita counties (Fig. 1) cover- can have broader positive implication in the conservation ing approximately 15,000 km2. The altitude of the study strategies of the Romanian Carpathians because brown area ranged between 20 m and 1,985 m, with complex bears were documented to act as a conservation surro- topography and land use: extensive forest cover in the gate for other species (Rozylowicz et al. 2011). main mountain ranges (typically above 500 m in eleva- Data on Romanian bear population size are poor, with tion), extensive agricultural lands in the southeastern re- management and conservation decisions relying on unre- gion, high plateau with mixed land use in the western alistic demographic data (Popescu et al. 2016), but there part (Transylvania), and urban development and agricul- is a relatively good understanding of brown bear habitat tural lands within the inter-mountain basins in the cen- requirement, with denning habitat being acknowledged tral part of the study region. This area harbors some as crucial for brown bear conservation in the Romanian of Europe’s largest tracts of old-growth forest, and is Carpathians (Pop et al. 2018b). Logging is a year-round dominated by beech, fir, and spruce (Fagus sylvatica, activity in the Romanian Carpathians, with the highest Abies alba, and Picea abies)—F. sylvatica dominated intensity between October and May (i.e., overlapping the at lower elevations (600–700 m) and coniferous species

Ursus 31:article e5 (2020)

Downloaded From: https://bioone.org/journals/Ursus on 18 Apr 2020 Terms of Use: https://bioone.org/terms-of-use Access provided by International Association for Bear Research and Management r BROWN BEAR DEN HABITAT IN THE CARPATHIANS Iosif et al. 3

Fig. 1. Location of the study area in the Romanian Carpathian Mountains showing the 3 administrative units— Harghita, Covasna, and Vrancea counties—as well as the documented brown bear (Ursus arctos) den locations (2010–2013), and pseudo-absences generated at 2 spatial scales: local (third-order selection) and landscape (second-order selection). Local-scale pseudo-absences were generated within approximately 6-km buffers around den locations. Landscape-scale pseudo-absences were generated within the entire study area. Major cities are highlighted with triangles.

dominated at higher elevations (1,100–1,300 m). Along Romania. The large mammal community is intact and in- the valleys, the landscape structure includes a matrix of cludes the other 2 European large carnivores (wolf [Canis pastures, hayﬁelds, row crops, and orchards, as well as lupus] and Eurasian lynx [Lynx lynx]); several mesocar- small rural settlements. Our study area is thought to har- nivores (red fox [Vulpes vulpes], European wildcat [Felis bor the largest portions of the brown bear population in silvestris], European badger [Meles meles], and several

Ursus 31:article e5 (2020)

Downloaded From: https://bioone.org/journals/Ursus on 18 Apr 2020 Terms of Use: https://bioone.org/terms-of-use Access provided by International Association for Bear Research and Management r 4BROWN BEAR DEN HABITAT IN THE CARPATHIANS Iosif et al.

mustelids); as well as large ungulates (wild boar [Sus Using Johnson (1980) orders of habitat selection, we eval- scrofa], roe deer [Capreolus capreolus], and red deer uated den selection by brown bears at 2 scales: landscape [Cervus elaphus]). During the spring and summer sea- scale (i.e., selection of denning habitat attributes relative sons, brown bears are habitat generalists; in our study to the entire study area; second-order selection), and lo- area, brown bears select for mixed forest, regenerating cal scale (i.e., selection of component of denning habitat forest (transitional woods and shrubs), and coniferous within the pre-denning season home range; third-order forest; deciduous forest is selected during the hyperpha- selection). gia season (autumn), when brown bears feed intensely to build up lipid reserves for the winter season (Pop et al. Binary response variable 2018b). In contrast, during the winter season brown bears For each analysis, the response variable consisted of are highly selective in their denning habitat, which con- 86 documented den locations, and an equal number of sisted mostly of mixed forest for males, and mixed and randomly generated pseudo-absences. For the second- coniferous forests for females (Pop et al. 2018b). Brown order selection analysis, we generated pseudo-absences bear diet is generalist, with main food sources represented throughout the study area, thus providing an evaluation by beech mast (F. sylvatica) and acorns (Quercus spp.), of denning selection throughout the geographic space. invertebrates (ants, bees, wasps), grass and fresh buds Brown bears have large annual home ranges in our study and shoots, roots, and coniferous tree sap (Alm˘a¸san et al. area, with a median 95% minimum convex polygon 1963, Roellig et al. 2014). (MCP) of 629 km2 and no differences between males and females (Pop et al. 2018a); thus, the second-order selection effectively assessed den selection across the entire Methods study area. Den surveys For the third-order selection analysis, we generated Between 2010 and 2013, we identified 86 brown bear pseudo-absences in an aggregated buffer around den loca- dens (Fig. 1); these dens were located opportunistically, tions with a radius of 5,893 m. The area encompassed by a either through back-tracking of bear tracks in snow en- circle with this radius was equal with the average home- countered on forest roads and trails during early spring, range size of brown bears during the pre-denning sea- or opportunistic information on den locations from game son November–December (average home range ± SD = wardens. Many of the areas surveyed were close to for- 109.10 ± 120.71 km2; Fig. 1). We estimated home range est roads and trails (i.e., accessible by foot), so back- size for November–December as 100% MCP based on tracking required accessing difficult topography, often 13 Global Positioning System–tracked bears monitored kilometers away from the initial track encounter. For each in the study area during the denning season (8,822 reloca- den, we measured 1) entrance height and width; 2) cham- tions, Pop et al. 2018a,b; for a similar approach see Pigeon ber height, width, and length; and 3) chamber horizontal et al. 2016). The November–December window is a rele- and vertical depths (we used maximum values of each vant period to capture bear behavior of searching for den measurement because of the complex shapes of the cham- sites and preparing for denning (Pop et al. 2018b). This bers). We also recorded the number of entrances for each approach was used by other studies on carnivore habi- den, and their substrate, as well as local topographic pa- tat selection and abundance (Furnas et al. 2017, Popescu rameters such as exposure and slope. et al. 2017, Pop et al. 2018b). Although it is known that after accumulating fat reserves during hyperphagia, brown Denning habitat suitability bears are searching for a denning site as a strategy for We evaluated denning habitat suitability using resource coping with lack of food and extreme weather conditions selection functions, which allow quantifying the strength (Carey et al. 2003), we acknowledge the difficulties in of animal–habitat selection in the form of suitability identifying the start of the denning period. The denning scores (Johnson et al. 2006). We used logistic regres- period may vary by latitude (Linnell et al. 2000), reaching sion, which relates a binary response (e.g., presence or a maximum length in the extreme North (e.g., 30 Oct–2 absence of dens) to habitat covariates in order to evalu- May in Southeast Alaska; Schoen et al. 1987), with preg- ate the level of selection or avoidance for each specific nant females always entering earlier and emerging later covariate. Habitat selection is a hierarchical process oc- (e.g., Manchi and Swenson 2005). Variation of denning curring at multiple spatial scales that involves a suite of period and length might also be influenced by food avail- behaviors that directly or indirectly relate to acquiring re- ability (Pigeon et al. 2016), thus increasing the variability sources that maximize individual fitness (Johnson 1980). to be accounted for in models.

Ursus 31:article e5 (2020)

Downloaded From: https://bioone.org/journals/Ursus on 18 Apr 2020 Terms of Use: https://bioone.org/terms-of-use Access provided by International Association for Bear Research and Management r BROWN BEAR DEN HABITAT IN THE CARPATHIANS Iosif et al. 5

Table 1. Habitat covariates used to predict denning suitability for brown bears (Ursus arctos) in the Southeast- ern Romanian Carpathian Mountains, from data collected between 2010 and 2013.

Range (at den Category Predictors Deﬁnition locations) Data source Topographic alt Altitude 440–1,320 m ASTER GDEM 30m (https:// slo Slope 3–47° asterweb.jpl.nasa.gov/gdem.asp) asp Aspect 0.1–360.0° tri Terrain ruggedness index 65.7–305.6 m Habitat p.dec Proportion cover of deciduous forest 0–1 Corine Land Cover 2012 p.mix Proportion cover of mixed forest 0–1 (http://land.copernicus.eu/pan- p.con Proportion cover of coniferous forest 0–1 european/corine-land-cover/clc-2012) p.you Proportion cover of young forest 0.00–0.75 De Jong (2012) p.mid Proportion cover of middle-aged forest 0–1 p.old Proportion cover of old-growth forest 0–1 Disturbance urb Distance to urban areas 222.3–14,002.6 Derived from national road and urban m vector data. (http://www.geo-spatial.org) tra Distance to transportation infrastructure 111.2–11,335.4 (paved roads and railroads) m p.past Proportion of canopy cover losses prior to 0.00–0.24 Hansen et al. (2013) the sampling period (2000–2010) p.pres Proportion of canopy cover losses during 0.00–0.26 the sampling period (2010–2013)

Predictor variables ﬂections in the spectral range of light absorbed by the We extracted the predictors used for modeling den foliar chlorophyll (Lillesand et al. 2008). By comparing suitability in Geographic Information Systems (GIS) at images spanning >20 years, De Jong (2012) delineated a resolution of 250 m from various publicly available 7 classes: non-forest, partial forest, clear-cut, shrubland, data sets, and grouped them into topographic, habitat, young, juvenile, and mature forest. We used only the last and disturbance predictors (Table 1). We generated to- 3 forest categories: [p.you] proportion cover of young pographic predictors from a 30-m-resolution digital el- forest, [p.mid] proportion cover of middle-aged forest, evation model [alt] and further transformed them into and [p.old] proportion cover of old-growth forest. Lastly, a slope [slo], an aspect [asp], and a terrain ruggedness we set a series of landscape disturbance layers: [urb]— layer [tri] (see Riley et al. 1999). Habitat predictors pro- Euclidian distance to urban areas, [tra]—Euclidian dis- vided information on the proportion of major forest cat- tance to major transportation infrastructure (paved roads egories and successional forest stages. To understand the and railroads), [p.past]—proportion of canopy cover loss effect of forest composition on den selection, we used prior to the sampling period (2001–2008), and [p.pres]— proportion cover of main forest categories (i.e., [p.dec], proportion of canopy cover loss during the sampling [p.mix] and [p.con] for deciduous, mixed, and coniferous period (2010–2013), both derived from 30-m-resolution forests, respectively) derived from Corine Land Cover global maps of 21st century forest cover change (Hansen 2012 vector data set (https://land.copernicus.eu/pan- et al. 2013; Table 1). Although bears use disturbed areas european/corine-land-cover/clc-2012). Forest composi- (e.g., clear-cuts) during the foraging seasons (Berland tion was documented to be a prerequisite for plan- et al. 2008, Pop et al. 2018b), we expect patches with ning conservation priorities during the winter season for canopy cover loss to have a negative effect on den se- Romanian brown bears, with large stands of old-growth lection. A negative relation between den selection and mixed forest having the highest conservation value (Pop distance to anthropogenic features was also documented et al. 2018b). To evaluate possible effects of forest age in populations from North America and Scandinavia and development on den selection (e.g., Schoen et al. (Ciarniello et al. 2005, Sahlén et al. 2011). 1987, Ciarniello et al. 2005), we used forest succes- We implemented all raster data extractions in Geospa- sional stages as a proxy. Forest successional stages within tial Modeling Environment (Beyer 2012) and further pro- the study area were obtained from De Jong (2012), cessed them in Program R (R Core Team 2018). Prior to who derived forest successional stages from Landsat modeling, the GIS layers were centered on the mean and (https://landsat.gsfc.nasa.gov/) images by analyzing re- transformed using the generic ‘scale’ function available

Ursus 31:article e5 (2020)

Downloaded From: https://bioone.org/journals/Ursus on 18 Apr 2020 Terms of Use: https://bioone.org/terms-of-use Access provided by International Association for Bear Research and Management r 6BROWN BEAR DEN HABITAT IN THE CARPATHIANS Iosif et al.

Table 2. Model selection table for binomial generalized linear models predicting brown bear (Ursus arctos) den selection at the local scale (third-order selection) in the Southeastern Romanian Carpathian Mountains, from data collected between 2010 and 2013. The top model(s) in each category (based on corrected Akaike’s Information Criterion values [AICc]) is marked with an “*”; the variables (see Table 1) in the top models were used to build a global model. For habitat category we selected the second top model as well (marked with “**”) in order to include a larger number of ecologically meaningful habitat covariates in the global model.

a b Category Model K AICc Weight logLik AUC Topographic alt+slo * 2 216.1 0.581 −104.9 0.743 alt+slo+asp 3 218.0 0.222 −104.9 0.743 alt+tri 2 218.9 0.142 −106.4 0.739 alt+tri+asp 3 220.8 0.055 −106.3 0.747 Habitat p.mix+p.dec+p.con * 3 202.1 0.897 −96.9 0.772 p.mix+p.dec+p.con+p.you+p.mid+p.old ** 6 206.6 0.093 −95.9 0.781 p.you+p.mid+p.old 3 211.0 0.010 −101.4 0.747 Disturbance urb+tra * 2 238.6 0.786 −116.2 0.635 p.past+p.pres 2 244.1 0.048 −119.0 0.484 p.past+p.pres+urb+tra 4 241.7 0.166 −115.6 0.647 Global model alt+slo+p.mix+p.dec+p.con+p.you+ 10 210.9 — −93.6 0.809 p.mid+p.old+urb+tra

alogLik is the log-likelihood of the model (the likelihood is a function that associates the probability of observing the given sample to each parameter. The log-likelihood is used for calculating AICc). bAUC is the Area Under the Curve of the Receiving Operator Characteristic; AUC values >0.8 denote good model ﬁt.

in Program R. We also explored all possible correlations bears in the study area using the equation: between predictors, and only retained covariates with cor- relation coefﬁcients >−0.7 or <0.7 to be included in the Pden = same models. exp(intercept + X1 × β1 + X2 × β2 + ... + Xn × βn )

1+ exp(intercept + X1 × β1 + X2 × β2 + ... + Xn × βn ) Resource selection functions For each of the 2 analyses (landscape-level [second- where Pden is the relative probability of bears selecting order] and local-level [third-order]), we fit generalized any pixel of the study area as denning habitat, Xsarethe β linear models with a binomial distribution of errors in covariates retained in the best model, and s are their Program R (R Core Team 2018). For each analysis, we ran corresponding coefficient estimates. several candidate models categorized into topographic models, disturbance models, and habitat models (Tables 2 and 3). We took a model selection approach based Results on second-order Akaike’s Information Criterion (cor- Den measurements rected for small sample sizes; AICc) values (Burnham and Altitude of the denning sites ranged between 440 and Anderson 2002) and identified the top models for each 1,320 m with an average (±SE) of 954.1 ± 20.2 m. Slope category. We then integrated the variables in the top mod- of the terrain around the dens ranged between 3° and 47° els for each category in a global model for each scale, and with an average (±SE) of 19.7° ± 0.8°. Den chambers evaluated second-order and third-order brown bear den- were mainly located on mountain slopes with southwest ning habitat selection (Tables 2 and 3). We used the stan- (22.9% of dens), east (20.5%), south (18.1%), and south- dardized coefficient estimates from each global model to east (15.7%) exposure, with bears showing a greater mag- evaluate denning habitat selection. We evaluated model fit nitude of selection for dens with eastern exposure (i.e., for all models using the Area Under the Curve (AUC) of 28.0%; Table 4). Dens were mainly excavated between the Receiver Operating Characteristic. Area Under Curve boulders (73%), followed by sites excavated directly in values of 0.5 denote a complete lack of fit (similar to an the soil (23%), and sites excavated under tree roots or intercept-only model), while AUC values >0.8 denote leaf litter (4%). Ninety percent of dens had 1 entrance, good model fit. and 10% had 2 entrances. The entrance of dens measured Finally, we used the second-order global model esti- averaged 78.4 ± 4.2 cm in height and 95.3 ± 4.3 cm in mates to map the probability of den selection by brown width. Chamber height was on average (±SE) 113.8 ±

Ursus 31:article e5 (2020)

Downloaded From: https://bioone.org/journals/Ursus on 18 Apr 2020 Terms of Use: https://bioone.org/terms-of-use Access provided by International Association for Bear Research and Management r BROWN BEAR DEN HABITAT IN THE CARPATHIANS Iosif et al. 7

Table 3. Model selection table for binomial generalized linear models predicting brown bear (Ursus arctos)den selection at landscape scale (second-order selection) in the Southeastern Romanian Carpathian Mountains, from data collected between 2010 and 2013. The top model(s) for each category (by corrected Akaike’s Infor- mation Criterion values [AICc]) are marked with an “*”; the variables (see Table 1) in the top models were used to build a global model. For habitat models we selected the second top model as well (marked with “**”) in order to include a larger number of ecologically meaningful habitat covariates in the global model (the model also had better ﬁt than the top model).

a b Category Model K AICc Weight logLik AUC Topographic alt+slo* 2 206.2 0.587 −100.0 0.757 alt+asp 2 206.9 0.413 −101.4 0.683 Asp 1 223.8 0.000 −108.8 0.548 Tri 1 240.9 0.000 −118.4 0.743 Habitat p.mix+p.dec+p.con* 3 192.0 0.828 −91.8 0.809 p.mix+p.dec+p.con+p.you+p.mid+p.old** 6 195.2 0.165 −90.2 0.814 p.you+p.mid+p.old 3 201.6 0.007 −96.6 0.787 Disturbance urb+p.pres* 2 239.9 0.419 −116.8 0.640 urb+p.past 2 240.2 0.363 −117.0 0.644 tra+p.pres 2 243.1 0.083 −118.4 0.584 tra+p.past 2 243.2 0.079 −118.5 0.585 p.past+p.pres 2 243.9 0.056 −118.8 0.535 Global model alt+slo+p.mix+p.dec+p.con+p.you+p.mid+ 10 193.5 — −85.7 0.847 p.old+urb+p.pres

10.5 cm, chamber width was 109.1 ± 4.8 cm, and cham- bance was not different from an intercept-only model: ber length was 149.2 ± 5.6 cm (Table 5). AUC = 0.484; Table 2). Brown bears showed significant selection of several Local and landscape predictors of brown bear habitat, topographic, and disturbance variables at the denning habitat landscape scale (second-order selection). Similar to the At the local scale (third-order selection), brown bears local-scale selection, dens were located in areas with significantly selected den habitats in areas with greater greater proportion of mixed forest (beech–spruce–fir), proportion of mixed forest cover, and steeper slopes and steeper slopes, relative to availability throughout the compared with random locations within 5,893-m buffers study areas (Fig. 2; Table 3). In addition, brown bears around den locations (Fig. 2; Table 2). Brown bears also selected for higher elevation, greater proportion of de- showed an affinity for denning in old forest stands, and ciduous forest, as well as greater proportions of all age at greater distance from the permanent transportation classes (likely an artifact of overall selection of forest network, but these variables were not significant at this habitat over other land-cover types across the study area). scale (95% CIs of standardized coefficients crossed zero; Although not significant, brown bears tended to avoid Fig. 2). Overall, habitat models performed best across dens in areas with greater recent forest loss and close to the models tested and also had the best fit (highest AUC; urban areas (i.e., away from recent disturbance; Fig. 2). Table 2), while disturbance models performed worse Similar to the local-scale analysis, disturbance models (e.g., the model including past and recent forest distur- performed worst across the models tested (Table 3), but

Table 4. Distribution of Romanian brown bear (Ursus arctos) den chambers and entrances by aspect, from data collected between 2010 and 2013 in Southeastern Carpathian Mountains, Romania.

NNEE SES SWWNW Den chamber 1.2% 9.6% 20.5% 15.7% 18.1% 22.9% 8.7% 3.6% Den entrance 2.4% 4.9% 28.0% 11.0% 26.8% 18.3% 6.1% 2.4%

Ursus 31:article e5 (2020)

Downloaded From: https://bioone.org/journals/Ursus on 18 Apr 2020 Terms of Use: https://bioone.org/terms-of-use Access provided by International Association for Bear Research and Management r 8BROWN BEAR DEN HABITAT IN THE CARPATHIANS Iosif et al.

Table 5. Summary measurements of 86 brown bear the eastern part of the study area (i.e., western part of (Ursus arctos) den entrances and chambers col- Ciucului and Hasmas mountains). Inter-mountain basins lected between 2010 and 2013 in Southeastern such as Giurgeu, Ciuc, and Brasov lowlands had lowest Carpathian Mountains, Romania. denning suitability (Fig. 3). Measurement (cm) Min. Max. Mean SE Entrance Height 15 260 78.4 4.2 Width 20 220 95.3 4.3 Discussion Chamber Horizontal depth 10 700 160.1 14.1 Vertical depth 30 250 103.6 5.2 Brown bears in the Southeastern Romanian Carpathi- Height 50 780 113.8 10.5 ans showed hierarchical denning habitat selection, with Width 30 270 109.1 4.8 stronger selection at the landscape scale compared with Length 70 330 149.2 5.6 local scale. Forest type and slope were the strongest predictors at both scales, with positive selection to steep terrain, and mixed forest cover (or beech–fir–spruce). their fit was slightly better compared with the disturbance Contrary to our expectations, the degree of disturbance at the local scale. Models including habitat variables had (past or present forest cover loss, urban and transporta- the best fit (AUC >0.8; Table 3), with the global model tion network) were not strong predictors of denning habi- showing the best fit of all models tested across the 2 scales tat suitability. Brown bears selected slopes with eastern (AUC = 0.847; Table 3). and southern exposure. Our observed selection patterns of The spatial prediction derived from the best fit model denning habitat for forest type, age, and topographic vari- at the landscape scale revealed homogenous and highly ables largely corroborate other studies in Europe, Asia, suitable areas for denning in Buzau and Vrancea moun- and North America at comparable latitudes (Huber and tains (south-central part of the study area), as well as in Roth 1997, Seryodkin et al. 2003, Goldstein et al. 2010, Nemira and Giurgeu mountains (northeastern part of the Ugarković et al. 2014, Krofel et al. 2017). This first as- study area; Fig. 3). On the western part, there is a north– sessment of Romanian brown bear denning habitat at- south strip that alternates highly suitable and moderately tributes at 2 different scales and the resulting spatial suit- suitable denning habitat in Bodoc, Baraolt, Gurghiu, and ability predictions complement and corroborate broader Harghita mountains. Low suitability was predicted in extent (i.e., Eastern Romanian Carpathians) predictions

Fig. 2. Standardized coefﬁcient estimates and 95% conﬁdence intervals for variables included in the global models used to evaluate denning habitat suitability for brown bears (Ursus arctos) in the Southeastern Romanian Carpathian Mountains (from data collected between 2010 and 2013) at 2 spatial scales: local (third- order selection) and landscape (second-order selection). Variable names: alt—altitude; slo—slope; p.dec— proportion cover of deciduous forest; p.mix—percent cover of mixed forest; p.con—proportion cover of coniferous forest; p.you—proportion cover of young forest; p.mid—proportion cover of middle-aged forest; p.old— proportion cover of old-growth forest; urb—Euclidian distance to urban areas; tra—Euclidian distance to major transportation infrastructure; p.pres—proportion of canopy cover loss during the sampling period.

Ursus 31:article e5 (2020)

Downloaded From: https://bioone.org/journals/Ursus on 18 Apr 2020 Terms of Use: https://bioone.org/terms-of-use Access provided by International Association for Bear Research and Management r BROWN BEAR DEN HABITAT IN THE CARPATHIANS Iosif et al. 9

Fig. 3. Predicted suitability for brown bear (Ursus arctos) denning habitats in the Southeastern Carpathian Mountains, Romania (from data collected between 2010 and 2013) based on the global binomial model at landscape scale (second-order selection). The associated suitability probabilities are as follows: low (0.00– 0.33), medium (0.33–0.66), and high (0.66–1.00).

of habitat of high conservation value based on telemetry forest with dominant Picea abies (Elfström et al. 2008). and movement ecology (Pop et al. 2018a, b). These patterns highlight the overall flexibility of brown bears in selecting denning habitat, with major differences Local and landscape predictors for denning in selection occurring even within the same population habitat suitability when forest types and structure change across latitudes We tested denning habitat selection using resource se- and ecosystems (Schoen et al. 1987). In our study, for- lection functions at 2 spatial scales: local, within 5,893-m est age was not a strong predictor at the local scale, but buffers around den sites, and landscape, across the brown bears exhibited positive (but not significant) selec- broader study area. There were differences in the strength tion for areas with greater proportion of older stands at of selection between the 2 scales, but several variables the landscape level. In contrast, the selection of all-aged were consistent across scales: mixed forest cover and stands at the landscape level is likely an outcome of the slope steepness. The probability of den occurrence was overall strong selection for forest habitat over other land- positively related to the proportion of mixed forest cover, cover types in our study area (e.g., agricultural land in the typically with a more complex vertical forest structure. low-elevation areas). This corroborates findings in Croatia (Ugarkovićetal. The other consistent variable selected across scales was 2014), as well as Eastern Romanian Carpathians (Pop slope. Positive selection for steeper slopes was strong, et al. 2018b), where the bears showed strongest selec- which corroborates other studies (e.g., Vroom et al. 1980, tion during the denning season for mixed forests. Selec- Judd et al. 1986, Seryodkin et al. 2003, Hodder et al. tion for complex vertical forest structure and more open 2014), suggesting that brown bears select to hibernate canopy cover for denning was also found in the Scandi- in locations that are less accessible to humans. At the navian boreal forest, where brown bears selected for open landscape scale, we also found a positive selection for canopy habitats with Pinus sylvestris over closed canopy higher altitudes; this could be explained both by the fact

Ursus 31:article e5 (2020)

Downloaded From: https://bioone.org/journals/Ursus on 18 Apr 2020 Terms of Use: https://bioone.org/terms-of-use Access provided by International Association for Bear Research and Management r 10 BROWN BEAR DEN HABITAT IN THE CARPATHIANS Iosif et al.

that in the Romanian Carpathians human disturbance substrates used by Romanian brown bears corroborates and access decrease with increasing altitude (Kuemmerle findings for brown bears in North America, where Cia- et al. 2008), and that higher altitudes are also less acces- rniello et al. (2005) reported the use of boulders and tree sible to human activities during winter (e.g., Seryodkin roots as a primary stabilizing structure of the dens. Sim- et al. 2003). It is likely that other areas of inaccessible ilar patterns were reported for the black bear (U. amer- denning habitat are available, and we underestimated the icanus) in Canada (Hodder et al. 2014), grizzlies in the overall habitat suitability as a result of our rather limited Greater Yellowstone Ecosystem (Judd et al. 1986), and sample size, and very complex topography. brown bears and Asiatic black bears (U. thibetanus)inthe In contrast to our expectations, disturbance predictors Russian Far East (Seryodkin et al. 2003). No den was lo- had the lowest influence at both scales, and the distur- cated in natural caves or crevices, in contrast with Slove- bance models had the worst fit. Distance to the main nian and Croatian populations, where most of the bears road network had a negative (but not significant) ef- were denning in karstic caves because of their wide avail- fect on den occurrence at the local scale, while recent ability (Huber and Roth 1997, Ugarković et al. 2014, Kro- forest loss (2010–2013) and distance to urban centers fel et al. 2017). As expected, brown bears preferred steep also had a negative (but not significant) effect on den slopes and slopes with longer periods of sun exposure occurrence at the landscape scale. Other studies found during the short winter days. In our area, brown bears that brown bears actively avoided disturbed areas during are known to emerge briefly during warmer periods (Pop denning season; for example, Scandinavian brown bears et al. 2018a); thus, selection of sun-exposed slopes may did not den close to paved highways and smaller roads enable them to find food resources. This pattern is consis- that were ploughed during winter (Elfström et al. 2008). tent with other populations in Europe and North America Elfström et al. (2008) also found no avoidance of the (Podruzny et al. 2012, Krofel et al. 2017, Whiteman et al. denser network of un-ploughed forestry roads, suggesting 2017), although other studies found that exposure was they have no disturbance effect on denning in the Scandi- random across denning sites (e.g., Elfström et al. 2008). navian population (due to lack of traffic during denning). This finding corroborates our study, where most roads are forest roads with little to no traffic during winter; Management implications however, further assessments are needed to understand In this study we developed predictive models of po- effect of unregulated off-road vehicle access during the tential denning habitat for brown bears at local and land- denning season, which is known to negatively affect den scape scales in the Southeastern Romanian Carpathians, occurrence (Goldstein et al. 2010). and described den characteristics. Our study adds novel Distance to urban areas had no influence on den se- information that can assist brown bear conservation in lection at the local scale, but was marginally significant Romania in the context of the current reassessment of at the landscape scale, likely because most urban areas its management and conservation planning nationwide are concentrated along the valleys and >6 km (i.e., size (Popescu et al. 2016, Pop et al. 2018b). The informa- of the buffer for the local-scale analysis) away from the tion on predictors of habitat selection at the 2 scales and contiguous mixed forests selected by brown bears at both the spatially explicit landscape-scale predictions has the scales. As such, dens in our study area are less likely to potential to inform forest management in implementing be affected by noise disturbance from human settlements. timber harvesting strategies that address the habitat con- This is particularly important because Linnell et al. (2000) servation needs of brown bears. In addition, our findings highlighted that brown bears do not tolerate noises, even provide guidance for identifying denning hotspots based at low intensity, if they occur in close proximity to dens on local site information, such as slope, aspect, and pres- (e.g., human visitation leading to den abandonment; Sery- ence of boulders and old tree roots as stabilizing materi- odkin et al. 2003). Moreover, the potential negative effect als. of disturbances is increased by the high degree of fidelity Extensive forest harvesting continues in the Romanian adult bears show to a particular denning area within their Carpathians at a rapid pace (Griffiths et al. 2013), and home range (Manchi and Swenson 2005). there is no information available on possible disturbance effects of forestry practices on brown bear ecology, par- Den attributes ticularly for denning habitat. We suggest that further eval- Most dens used by Romanian brown bears are cham- uation and monitoring of the magnitude of forestry activ- bers located between boulders, or dug directly in the ities is needed in areas that were predicted as highly suit- soil, and less frequently under tree roots. The variety of able for brown bear denning. When compared with other

Ursus 31:article e5 (2020)

Downloaded From: https://bioone.org/journals/Ursus on 18 Apr 2020 Terms of Use: https://bioone.org/terms-of-use Access provided by International Association for Bear Research and Management r BROWN BEAR DEN HABITAT IN THE CARPATHIANS Iosif et al. 11

mountain ranges, paved roads and railroads have low den- disturbance in the Carpathian Mountains 1985–2010. Con- sity in the Carpathians, but the network of forestry roads servation Biology 31:570–580. allows access in the core denning habitats for off-road CAREY, H.V., M.T. ANDREWS, AND S.L. MARTIN. 2003. Mam- vehicles; limiting access of off-road vehicles at the be- malian hibernation: Cellular and molecular responses to de- ginning of the denning period is critical because brown pressed metabolism and low temperature. Physiological re- views 83:1153–1181. bears were documented to abandon their disturbed dens CHAPRON,G.,P.KACZENSKY, J.D.C. LINNELL,M.VON ARX, more easily during this period (Swenson et al. 1997, Lin- D. HUBER,H.ANDRÉN,J.V.LÓPEZ-BAO,M.ADAMEC,F. nell et al. 2000). We recommend not only avoidance of ÁLVARES,O.ANDERS,L.BALCIAUSKASˇ ,V.BALYS,P. clear-cutting, when possible, in highly suitable denning BEDO˝,F.BEGO,J.C.BLANCO,U.BREITENMOSER,H. landscape, but also regulating access of vehicles and min- BRØSETH,L.BUFKA,R.BUNIKYTE,P.CIUCCI,A.DUTSOV, imizing other human activities within these habitats dur- T. ENGLEDER,C.FUXJÄGER,C.GROFF,K.HOLMALA,B. ing the sensitive denning period. HOXHA,Y.ILIOPOULOS,O.IONESCU,J.JEREMIC´,K.JERINA, G. KLUTH,F.KNAUER,I.KOJOLA,I.KOS,M.KROFEL,J. KUBALA,S.KUNOVAC,J.KUSAK,M.KUTAL,O.LIBERG,A. Acknowledgments MAJIC´,P.MÄNNIL,R.MANZ,E.MARBOUTIN,F.MARUCCO, D. MELOVSKI,K.MERSINI,Y.MERTZANIS,R.W.MYSŁA- Den surveys were performed under the framework of JEK,S.NOWAK,J.ODDEN,J.OZOLINS,G.PALOMERO,M. the project LIFE08NAT/RO/000500–LIFEURSUS enti- PAUNOVIC´,J.PERSSON,H.POTOCNIKˇ ,P.Y.QUENETTE,G. tled “Best practices and demonstrative actions for the con- RAUER,I.REINHARDT,R.RIGG,A.RYSER,V.SALVATORI, servation of Ursus arctos populations in Central-Eastern T. SKRBINŠEK,A.STOJANOV,J.E.SWENSON,L.SZEMETHY, Carpathians,” financed by the LIFE Nature program, A. TRAJÇE,E.TSINGARSKA-SEDEFCHEVA,M.VÁNAˇ , the European Union’s financial instrument supporting R. VEEROJA,P.WABAKKEN,M.WÖLFL,S.WÖLFL,F. environmental, nature conservation, and climate action ZIMMERMANN,D.ZLATANOVA, AND L. BOITANI. (http://ec.europa.eu/environment/life/). Authors were 2014. Recovery of large carnivores in Europe’s mod- supported by a grant from the Romanian National Author- ern human-dominated landscapes. Science 346:1517– ity for Scientific Research, CNCS–UEFISCDI (Consili- 1519. CIARNIELLO, L.M., M.S. BOYCE,D.C.HEARD, AND D.R. SEIP. ului Na¸tional al Cercet˘ariiStiin¸ ¸ tifice–Unitatea Executiva 2005. Denning behavior and den site selection of grizzly pentru Finantarea Invatamantului Superior, a Cercetarii, bears along the Parsnip River, British Columbia, Canada. Dezvoltarii si Inovarii; http://uefiscdi.gov.ro), project Ursus 16:47–58. PN-II-RU-TE-2014-4-0058. We thank J. Swenson, CLEVENGER, A.P., F.J. PURROY,M.S.DE BURUAGA,F.PURDOY, A. Ordiz, V. Penteriani, and one anonymous reviewer for AND M.S. DE BURGUAGA. 1987. Status of the brown bear in helpful comments on an earlier version of this manuscript, the Cantabrian Mountains, Spain. Bears: Their Biology and and T. Estabrook Boal for edits on the final version. Management 7:1–8. DE JONG, J.F. 2012. Object-based classification of the successional stage of temperate forests using bi-temporal Landsat Literature cited TM Band 3 images. Thesis, Wageningen University, Wa- ALMA˘ SAN¸ ,H.,T.BABUTIA¸ ,V.COTTA, AND C. POPESCU. 1963. geningen, The Netherlands. Contribu¸tii la cunoa¸sterea r˘asp˘andiriisi ¸ biologiei ursului DI MININ,E.,R.SLOTOW, L.T.B. HUNTER,F.M.POUZOLS, (Ursus arctos L.) în R.P.R. Studiisi ¸ Cercet˘ari INCEF 23:51– T. TOIVONEN,P.H.VERBURG,N.LEADER-WILLIAMS,L. 70. [In Romanian.] PETRACCA, AND A. MOILANEN. 2016. Global priorities for BERLAND,A.,T.NELSON,G.STENHOUSE,K.GRAHAM, AND J. national carnivore conservation under land use change. Sci- CRANSTON. 2008. The impact of landscape disturbance on entific Reports 6:23814. grizzly bear habitat use in the Foothills Model Forest, Al- ELFSTRÖM,M.,J.E.SWENSON, AND J.P. BALL. 2008. Selection berta, Canada. Forest Ecology and Management 256:1875– of denning habitats by Scandinavian brown bears Ursus arc- 1883. tos. Wildlife Biology 14:176–187. BEYER, H.L. 2012. Geospatial modelling environment (Version EVANS, A.L., N.J. SINGH,B.FUCHS,S.BLANC,A.FRIEBE, 0.7.3.0). http://www.spatialecology.com/gme. Accessed 5 T.G. LASKE,O.FROBERT,J.E.SWENSON, AND J.M. Jan 2018. ARNEMO. 2016. Physiological reactions to capture in BURNHAM, K.P., AND D.R. ANDERSON. 2002. Model selection hibernating brown bears. Conservation Physiology 4: and multi-model inference. Second edition. Springer, Berlin, 1–11. Germany. FURNAS, B.J., R.H. LANDERS,R.L.CALLAS, AND S.M. BUTSIC,V.,C.MUNTEANU,P.GRIFFITHS,J.KNORN,V.C. MATTHEWS. 2017. Estimating population size of fishers RADELOFF,J.LIESKOVSKÝ,D.MUELLER, AND T. KUEM- (Pekania pennanti) using camera stations and auxiliary data MERLE. 2017. The effect of protected areas on forest on home range size. Ecosphere 8:e01747. Ursus 31:article e5 (2020)

Downloaded From: https://bioone.org/journals/Ursus on 18 Apr 2020 Terms of Use: https://bioone.org/terms-of-use Access provided by International Association for Bear Research and Management r 12 BROWN BEAR DEN HABITAT IN THE CARPATHIANS Iosif et al.

GOLDSTEIN, M.I., A.J. POE,L.H.SURING,R.M.NIELSON, AND LINNELL,J.,J.E.SWENSON,R.ANDERSEN, AND B. BARNES. T.L. MCDONALD. 2010. Brown bear den habitat and win- 2000. How vulnerable are denning bears to disturbance? ter recreation in South-Central Alaska. Journal of Wildlife Wildlife Society Bulletin 28:400–413. Management 74:35–42. MANCHI,S.,AND J.E. SWENSON. 2005. Denning behaviour of GRIFFITHS,P.,D.MÜLLER,T.KUEMMERLE, AND P. HOSTERT. Scandinavian brown bears Ursus arctos. Wildlife Biology 2013. Agricultural land change in the Carpathian ecoregion 11:123–132. after the breakdown of socialism and expansion of the Eu- MATEO-SÁNCHEZ, M.C., S.A. CUSHMAN, AND S. SAURA. 2014. ropean Union. Environmental Research Letters 8:045024. Connecting endangered brown bear subpopulations in the HANSEN, M.C., P.V. POTAPOV,R.MOORE,M.HANCHER,S.A. Cantabrian Range (north-western Spain). Animal Conser- TURUBANOVA,A.TYUKAVINA,D.THAU,S.V.STEHMAN, vation 17:430–440. S.J. GOETZ,T.R.LOVELAND,A.KOMMAREDDY,A.EGOROV, MÜLLER,D.,T.KUEMMERLE,M.RUSU, AND P. GRIFFITHS. L. CHINI,C.O.JUSTICE, AND J.R.G. TOWNSHEND. 2013. 2009. Lost in transition: Determinants of post-socialist crop- High-resolution global maps of 21st-Century forest cover land abandonment in Romania. Journal of Land Use Science change. Science 342:850–853. 4:109–129. HERTEL, A.G., S.M.J.G. STEYAERT,A.ZEDROSSER,A. NAKAJIMA,A.,S.KOIKE,T.MASAKI,T.SHIMADA,C. MYSTERUD,H.K.LODBERG-HOLM,H.W.GELINK,J.KIND- KOZAKAI,Y.NEMOTO,K.YAMAZAKI, AND K. KAJI. 2012. BERG, AND J.E. SWENSON. 2016. Bears and berries: Species- Spatial and elevational variation in fruiting phenology of a specific selective foraging on a patchily distributed food re- deciduous oak (Quercus crispula) and its effect on forag- source in a human-altered landscape. Behavioral Ecology ing behavior of the Asiatic black bear (Ursus thibetanus). and Sociobiology 70:831–842. Ecological Research 27:529–538. HODDER, D.P., C.J. JOHNSON,R.V.REA, AND A. ZEDROSSER. PETRAM,W.,F.KNAUER, AND P. KACZENSKY. 2004. Human 2014. Application of a species distribution model to identify influence on the choice of winter dens by European brown and manage bear den habitat in central British Columbia, bears in Slovenia. Biological Conservation 119:129–136. Canada. Wildlife Biology 20:238–245. PIGEON, K.E., S.D. CÔTÉ, AND G.B. STENHOUSE. 2016. Assess- HUBER,D.,AND H.U. ROTH. 1997. Denning of brown bears in ing den selection and den characteristics of grizzly bears. Croatia. Bears: Their Biology and Management 9:79–83. Journal of Wildlife Management 80:884–893. JOHNSON, C.J., S.E. NIELSEN,E.H.MERRILL,T.L.MCDONALD, ———, S.E. NIELSEN,G.B.STENHOUSE, AND S.D. CÔTÉ. AND M.S. BOYCE. 2006. Resource selection functions based 2014. Den selection by grizzly bears on a managed land- on use–availability data: Theoretical motivation and evalua- scape. Journal of Mammalogy 95:559–571. tion methods. Journal of Wildlife Management 70:347–357. PODRUZNY, S.R., S. CHERRY,C.C.SCHWARTZ, AND L.A. LAN- JOHNSON, D.H. 1980. The comparison of usage and availability DENBURGER. 2012. Grizzly bear denning and potential con- measurements for evaluating resource preference. Ecology flict areas in the Greater Yellowstone Ecosystem. Ursus 61:65–71. 13:19–28. JUDD, S.L., R.R. KNIGHT, AND B.M. BLANCHARD. 1986. Den- POP, M.I., L. BERECZKY,S.CHIRIAC,R.IOSIF,A.NITA,V.D. ning of grizzly bears in the Yellowstone National Park area. POPESCU, AND L. ROZYLOWICZ. 2018a. Movement ecol- Bears: Their Biology and Management 6:111–117. ogy of brown bears (Ursus arctos) in the Romanian Eastern KAMINSKI, D.J., C.E. COMER,N.P.GARNER,I.K.HUNG, AND Carpathians. Nature Conservation 26:15–31. G.E. CALKINS. 2013. Using GIS-based, regional extent habi- ———, R. IOSIF,I.V.MIU,L.ROZYLOWICZ, AND V.D. tat suitability modeling to identify conservation priority ar- POPESCU. 2018b. Combining resource selection functions eas: A case study of the Louisiana black bear in east Texas. and home-range data to identify habitat conservation pri- Journal of Wildlife Management 77:1639–1649. orities for brown bears. Animal Conservation 21:352–336. KNORN,J.,T.KUEMMERLE,V.C.RADELOFF,A.SZABO,M. doi:10.1111/acv.12399. MINDRESCU,W.S.KEETON,L.ABRUDAN,P.GRIFFITHS,V. POPESCU, V.D., K.A. ARTELLE,M.I.POP,S.MANOLACHE, GANCZ, AND P. HOSTERT. 2012. Forest restitution and pro- AND L. ROZYLOWICZ. 2016. Assessing biological realism of tected area effectiveness in post-socialist Romania. Biolog- wildlife population estimates in data-poor systems. Journal ical Conservation 146:204–212. of Applied Ecology 53:1248–1259. KROFEL,M.,M.ŠPACAPAN, AND K. JERINA. 2017. Winter sleep ———, R. IOSIF,M.I.POP,S.CHIRIAC,G.BOUROS¸, AND with room service: Denning behaviour of brown bears with B.J. FURNAS. 2017. Integrating sign surveys and teleme- access to anthropogenic food. Journal of Zoology 302:8–14. try data for estimating brown bear (Ursus arctos)density KUEMMERLE,T.,D.MÜLLER,P.GRIFFITHS, AND M. RUSU. in the Romanian Carpathians. Ecology and Evolution 7: 2008. Land use change in Southern Romania after the col- 7134–7144. lapse of socialism. Regional Environmental Change 9:1–12. RCORE TEAM. 2018. R: A language and environment for sta- LILLESAND,T.,R.W.KIEFER, AND J. CHIPMAN. 2008. Remote tistical computing. R Core Team, Vienna, Austria. ISBN sensing and image interpretation. Sixth edition. University 3-900051-07-0. http://www.R-project.org/. Accessed 1 Dec of Wisconsin, Madison, Wisconsin, USA. 2018.

Ursus 31:article e5 (2020)

Downloaded From: https://bioone.org/journals/Ursus on 18 Apr 2020 Terms of Use: https://bioone.org/terms-of-use Access provided by International Association for Bear Research and Management r BROWN BEAR DEN HABITAT IN THE CARPATHIANS Iosif et al. 13

RILEY, S.J., S.D. DEGLORIA, AND R. ELLIOT. 1999. A terrain SWENSON, J.E., F. SANDEGREN,S.BRUNBERG, AND P. ruggedness index that quantifies topographic heterogeneity. WABAKKEN. 1997. Winter den abandonment by brown bear Intermountain Journal of Sciences 5:1–4. Ursus arctos: Causes and consequences. Wildlife Biology RIPPLE, W.J., J.A. ESTES,R.L.BESCHTA,C.C.WILMERS,E.G. 3:35–38. RITCHIE,M.HEBBLEWHITE,J.BERGER,B.ELMHAGEN, UGARKOVIC´,D.,S.MIKAC, AND N.K. UGARKOVIC´. 2014. Den- M. LETNIC,M.P.NELSON,O.J.SCHMITZ,D.W.SMITH, ning ecology of brown bear (Ursus arctos L.) in the Nature A.D. WALLACH, AND A.J. WIRSING. 2014. Status and eco- Park Velebit, Croatia. Pages 514–518 in S. MarićandZ. logical effects of the world’s largest carnivores. Science Lonˇcarić, editors. 49th Croatian & 9th International Sym- 343:1241484. posium on Agriculture, Dubrovnik, Croatia. Poljoprivredni ROELLIG,M.,I.DORRESTEIJN,H.VON WEHRDEN,T.HARTEL, fakultet Sveuˇcilišta Josipa Jurja Strossmayera u Osijeku, Os- AND J. FISCHER. 2014. Brown bear activity in traditional ijeku, Croatia. wood-pastures in Southern Transylvania, Romania. Ursus VROOM, G.W., S. HERRERO, AND R.T. OGILVIE. 1980. The 25:43–52. ecology of winter den sites of grizzly bears in Banff Na- RONDININI,C.,AND L. BOITANI. 2007. Systematic conservation tional Park, Alberta. Bears: Their Biology and Management planning and the cost of tackling conservation conflicts with 4:321–330. large carnivores in Italy. Conservation Biology 21:1455– WHITEMAN,A.,G.PASSONI,J.M.ROWCLIFFE,D.UGARKOVIC´, 1462. J. KUSAK,S.RELJIC´, AND D. HUBER. 2017. Iden- ROZYLOWICZ,L.,V.D.POPESCU,M.PATROESCU˘ , AND G. tifying key denning habitat to conserve brown bear CHISAMERA¸ . 2011. The potential of large carnivores as con- (Ursus arctos) in Croatia. Wildlife Research 44:309– servation surrogates in the Romanian Carpathians. Biodiver- 315. sity and Conservation 20:561–579. ZARZO-ARIAS,A.,M.DEL M. DELGADO,A.ORDIZ,J.GAR- SAHLÉN, E., O.G. STØEN, AND J.E. SWENSON. 2011. Brown CÍA DÍAZ,D.CAÑEDO,M.A.GONZÁLEZ,C.ROMO,P. bear den site concealment in relation to human activity in VÁZQUEZ GARCÍA,G.BOMBIERI,C.BETTEGA,L.F.RUSSO, Sweden. Ursus 22:152–158. P. CABRAL,R.GARCÍA GONZÁLEZ,J.MARTÍNEZ-PADILLA, SCHOEN, J.W., L.R. BEIER,J.W.LENTFER, AND L.J. JOHNSON. AND V. PENTERIANI. 2018. Brown bear behaviour in human- 1987. Denning ecology of brown bears on Admiralty and modified landscapes: The case of the endangered Cantabrian Chichagof Islands. Bears: Their Biology and Management population, NW Spain. Global Ecology and Conservation 7:293–304. 16:e00499. SERYODKIN, I.V., A.V. KOSTYRIA,J.M.GOODRICH,D.G. ZIÓŁKOWSKA,E.,K.OSTAPOWICZ,V.C.RADELOFF,T.KUEM- MIQUELLE,E.N.SMIRNOV,L.L.KERLEY,H.B.QUIGLEY, MERLE,A.SERGIEL,T.ZWIJACZ-KOZICA,F.ZIEBA˛ ,W. M.G. HORNOCKER,I.VSERYODKIN,A.V.KOSTYRIA,J.M. S´ MIETANA, AND N. SELVA. 2016. Assessing differences in GOODRICH,D.G.MIQUELLE,E.N.SMIRNOV,L.L.KER- connectivity based on habitat versus movement models LEY,H.B.QUIGLEY, AND M.G. HORNOCKER. 2003. Den- for brown bears in the Carpathians. Landscape Ecology ning ecology of brown bears and Asiatic black bears in the 31:1863–1882. Russian Far East. Ursus 14:153–161. SKRBINŠEK,T.,M.JELENCIˇ Cˇ,L.WAITS,I.KOS,K.JERINA, AND P. TRONTELJ. 2012. Monitoring the effective population size Received: April 26, 2018 of a brown bear (Ursus arctos) population using new single- Accepted: April 9, 2019 sample approaches. Molecular Ecology 21:862–875. Associate Editor: A. Ordiz

Ursus 31:article e5 (2020)

Downloaded From: https://bioone.org/journals/Ursus on 18 Apr 2020 Terms of Use: https://bioone.org/terms-of-use Access provided by International Association for Bear Research and Management