Andean bear habitat use in the Oyacachi River Basin, Ecuador

Andean bear habitat use in the Oyacachi River Basin, Ecuador Francisco Cuesta1'3, Manuel F. Peralvo1'4, and Frank T. van Manen2'5 1EcoCiencia,FranciscoSalazar E14-34, P.O. Box 12-17-257, Quito,Ecuador 2U.S. GeologicalSurvey, SouthernAppalachianField Branch,Universityof Tennessee, 274 EllingtonPlant Sciences, Knoxville,TN 37996, USA Abstract: One of the primarythreatsto Andeanbear (Tremarctosornatus)populationsin Ecuadoris conversion of bear habitat to human uses, resulting in habitat loss and fragmentation.To develop science-based conservation plans, informationon the suitability and distributionof Andean bear habitatis criticallyneeded. We studiedhabitatuse in the 721-km2OyacachiRiver Basin in the eastern Andes Mountains.We used bimonthly sign surveys along 1.6-km transects(n = 53) to determine habitatuse. We recorded549 and 202 locations of bear activity during2000 and 2001, respectively; feeding signs were recordedmost frequently(53.3%), followed by scats (19.6%), footprints(13.4%), hair (6.9%),tree marks(4.8%),trails(1.2%), and groundnests (0.8%).The combinedtotal distanceof surveyedtransectsfor both years was 1,018 km with a correspondingsign encounterrate of 0.74/km. Use of the different vegetation types within the study area varied among the bimonthly sampling periods. Habitat suitability was calculated with a geographic informationsystem (GIS) based on Mahalanobisdistance (D2), a multivariatemeasureof dissimilarity,using 8 habitatvariablesand 437 bearlocations.Model validity was confirmedby testing whetherthe D2 values of 61 randomlocations in the Oyacachi River Basin were greater than those associated with 61 test locations. We used a cumulativefrequencycurve based on D2 values associatedwith the 61 independenttest locations to define 5 classes of habitat suitability,ranging from most used to avoided areas. The most suitable habitatclass occupied 86.3 km2 (11.9%) of the study area.The resultsof our study may be appliedon a regionalscale to define priorityconservationareasfor Andeanbearsin the easternAndes Mountains of Ecuador.Ourresultsindicatethe usefulness of field-basedstudiescombinedwith GIS and statistical analyses as a scientific basis for developing conservationstrategiesfor Andean bears on a landscape scale. Key words: Andean bear, conservationplanning, Ecuador, habitat suitability, habitat use, Mahalanobisdistance, Tremarctosornatus Ursus 14(2):198-209 (2003) I danger of extinction in Ecuador (Cuesta and Suarez 2001). The Andean bear is a key species in the conservation and managementof Andeanhabitatsdue to its large spatial requirements,its ecological role (e.g., potential seed disperser),and its profoundcharisma(Yerena and Torres 1994, Young 1999, Cuesta 2000). The Andean bear's wide ecological requirementsand its seasonaluse of differenthabitats,such as extensive paramoand cloud forest areas, make this species an appropriatesubject on which to base conservationplanningto preservethe high biodiversity of these ecosystems (Peyton 1999). The seasonal variabilityin food availabilityin habitats 4peralvomf@mail.utexas.edu used by Andeanbearstriggerswide-rangingmovements 3email:osos@ecociencia.org of the animals within their home ranges, as has been 5vanmanen@utk.edu Habitat loss and fragmentationare 2 of the main challenges in the conservationand managementof large carnivoresin the world (Peyton et al. 1999, Tiriraet al. 2001). Habitat fragmentationcan result in small, isolated populationsthatbecome increasinglyvulnerableto extinction(Diamond 1986, Wilcove 1987). The Andean bear presents a clear example of how habitatfragmentation and illegal huntinghave caused severe population reductions;consequently,this species is now considered threatenedat a global scale (Hilton-Taylor2000) and in 198 ANDEANBEARHABITAT USE * Cuesta et al. documented for other bear species (Schoen 1990). Those movements, however, often are impeded by the loss of cloud forestand pairamoareasbecause of advancing agriculturalfrontiers and expanding infrastructure (e.g., roads). Sierraet al. (1999) estimatedthat 38% of the original cover of paramos and cloud forests of Ecuadorhas been transformedinto agriculturalor urban lands. Schoen (1990) suggested that the broad ecological and spatial requirementsof bears demand management actions on a regionalscale. Because little is known about the habitat relationshipsof Andean bears, it is fundamental to undertakeresearchthat will aid in decisionmaking supporting regional bear conservation. The objective of our study was to determine suitability of Andean bear habitat in the Oyacachi River Basin, an importantareafor Andeanbearson the easternslopes of the Andes Mountainsin Ecuador. Study area Our study area was approximately50 km east of the city of Quito, within the Cayambe-CocaEcological Reserve, in the province of Napo. The study area covered721 km2,mainly in the OyacachiRiverbasin, of which 446 km2 were ancestralterritoriesof the Quichua Communityof Oyacachi (Fig. 1). Elevations within the study area range from 1,600 to 4,500 m with a mean slope of 41? (Cuestaet al. 2001). In the upperportionof the area,the salientgeomorphologic featuresare of glacial origin and include circles, roches moutonnees (streamlined knobs projecting from the land surface),and U-shaped valleys filled with moraine deposits, mudslides, and lahars (volcanic mudflows). The lower portion of the study area is highly dissected and covered with cloud forests; heterogeneous landscapes are predominant.The weather in the Oyacachi River basin varies accordingto elevation. In the upper portion,the mean annualtemperatureis 9? C with a mean rainfall of 1,500 mm. In the lower valley, the mean annualtemperaturereaches 17? C, with a mean rainfall of 2,500 mm (Lopez 1992). Six naturalvegetationtypes occur within the study area: upper montane evergreen forest (BSV-ma), montane cloud forest (BN-m), mixed pairamo forest (BPM), herbaceous paramo (PH), swampy paramo (PA), and alder (Alnus spp.) forests (Baez et al. 1999, Iturraldeet al. 2000; Fig. 2). The only humanpopulationin the study area lives in the village (600 residents)of Oyacachi,reachablefrom the town of Cayambe via a road built in 1995. A road south of the study area crosses the paramostowardthe Ursus 14(2):198-209 (2003) 199 town of Papallacta.A footpathconnects Oyacachi with the town of El Chaco, located 33 km east and crossing cloud forests within the study area (Fig. 1). The main human activities include extensive livestock management, subsistence agricultureat a range of altitudes, breeding of domestic animals, and handicraftmanufacturing (Morales and Schjellerup 1997). The community's relationshipwith the Andean bear dates back 500 years;the bearused to be considereda divinity (Andrade Marin 1952, Camachoet al. 1999). Presently,a relationship of mutual acceptance seems to have developed; hunting is prohibited by rules of the community (ComunidadQuichuade Oyacachi 2001). Methods Field surveys for bear sign Because of logistical challenges and inaccessibility of the terrain, we chose sign surveys, rather than radiotelemetry,as the basic data collection method for our study. Monitoring of wildlife populations through sign recordshas been used in many studies to determine population abundanceand to quantify habitat use and availability (Nams 1989, Clevenger et al. 1997). Six local researchersfrom Oyacachi surveyed for sign of bearactivityduring2000 and 2001 (x = 22 days/month). Bear sign informationwas gatheredalong 53 transects with a length of 1.6 km each. Placement of transects within the study areas was stratifiedaccording to the area representedby each vegetation type in the study area (Kendall et al. 1992). Given the poor accessibility within the study area, starting locations for most transectswere placed near the "horsetrail"to El Chaco along the OyacachiRiver or near the unpavedroad that connects Oyacachi village with Papallacta in the southwesternportion of the study area (Fig. 1). Once we located the start of each transect, we followed an upslope direction for those transects starting near the Oyacachi River and a random direction for highelevation transects.Transectroutes were placed in such a mannerthat various microhabitatconditions (such as differentaspects and slopes) were sampled within each vegetation formation.Each transectroute was marked with colored flagging at =10-m intervalsto ensure that subsequentsurveys followed the same route. Detectability of mammal sign may vary among different habitats and among types of sign (Wemmer et al. 1996). To minimize biases due to varying detectability rates, we used a fixed-widthtransectsurvey and only consideredbear sign within 2 m on either side of the transects (4-m transectwidth). This limited search 200 * Cuesta et al. ANDEAN BEAR HABITAT USE Elevation < 1900 m I ::-. 1900200 3LII3000- m - 3600 m] m 3600 - 4200 m-- gj I 2400 2400- 3000 m I > 4200 m . Community boundary N 0 3 6 Km Unpaved road Path Ursus 14(2):198-209 (2003) ANDEAN BEAR HABITATUSE * distanceincreasedthe probabilityof sign detection.Furthermore,we increasedsign detectabilityby using local field personnel, who were extremely skilled at locating bear sign. Finally, based on the results of a pilot study, Cuesta et al. (2001) found that the proportionof each type of sign was consistent among vegetation types, suggesting that the fixed-width surveys were effective in reducing detectabilitybias due to different types of sign. The 53 transectswere surveyedonce every 2 months. For each site with bearsign, field personnelcollected (1) global positioning system (GPS) coordinates of the location, (2) the type of sign, and (3) additional field measurementsto characterizethe site. The GPS coordinates were then used in combination with GIS to measure topographic, ecological, and anthropogenic variables selected to assess bear habitatuse within the study area. We acquireddigital cartographicdata from the Ecuadorian Instituto Geografico Militar's (IGM; Quito, Ecuador)topographiccharts(scale 1:50,000) and Landsat 5 Thematic Mapper (December 1998; EROS Data Center, Sioux Falls, South Dakota, USA) satellite imagery (Table 1). Each data layer was generated with TNT Mips GIS software (MicroImages, Inc., Lincoln, Nebraska, USA) in a raster format based on 30 X 30-m pixels. Use of vegetation types The recordsobtainedduringfield monitoringallowed us to definehabitatuse patternsand selection throughout 2000 and 2001. Data were groupedfor each bimonthly survey period and differentialuse of vegetation types was tested with Friedman's method for randomized blocks (Sokal and Rohlf 1981) using ao= 0.05. Habitat modeling Habitatmodels based on GIS technology are suitable tools to predict the presence and relative use of bear habitat across large landscapes (Clark and van Manen 1992), particularlybecause such models are appropriate for generalistspecies (Donovan et al. 1987). We used 8 GIS variables(Table 1) to measurehabitatconditionsfor the bear sign locations sampledduringthe field surveys. Those habitatconditionswere used as the trainingset to determinehabitatuse of Andean bears in the Oyacachi Fig. 1. Location of the Oyacachi River Basin study area in Ecuador for a study of habitat use by Andean bears, 2000-01. Ursus 14(2):198-209 (2003) Cuestaet al. 201 study area.We chose the habitatvariablesbased on our field observationsand a review of Andean bear literature. Of the 751 bear locations, we used 437 to develop the habitat model; 203 locations were combined with other locations that were within a distance of 5 m, and 111 locationswere excludedbecauseof largeGPS errors (>100 m). Such large errors usually were caused by poor satelliteacquisitiondue to dense forest canopy and rugged terrain. We calculated a multivariatestatistic, Mahalanobisdistance,to develop a habitatmodel (Clark et al. 1993, van Manen et al. 2002): D2= (X - )' -l(x - ), where x is the vector of habitat features at any given point, ii is the mean vector of habitat features at the locations sampled during the field surveys, and E-1 is the inverse of the variance-covariancematrix,calculated from the sampledpoints. D2 is a statistical measure of dissimilaritybetween the given point and the mean for all locations, and is expressed as a distance. We calculated D2 with GIS programs based on the values of digital map layers (habitatvariables)and "ideal" values of those variables associated with bear sign locations. Low D2 values indicatethata location has habitatfeaturessimilarto the "ideal" conditions sampled at the bear sign locations, whereas greaterD2 values indicate increasinglydissimilar conditions. Mahalanobisdistance is dimensionless because it is a functionof standardizedvariables,despite the different measurement scales among the original variables.There is no one best combinationof variables that results in the lowest D2 values; a variety of habitat combinations can result in identical distance values (Clarket al. 1993). We tested the habitatmodel based on 61 observations of bear sign collected independentlyfrom the bimonthly field surveys during2000 and 2001 (test locations). We divided the distance values associated with the 61 test locations into 5 range classes. These classes were based on discerniblediscontinuitiesof percentilesof a cumulative frequency curve and essentially represented 5 suitability classes (Boitani et al. 1999). We used the frequency data analysis function in SPSS 9.0 software (SPSS 1998) to identify percentiles at discontinuities among the 61 test locations. We further assessed model validity by generating a null model based on 61 randomlocations within the OyacachiRiver Basin for comparisonwith D2 values of the 61 test locations. We used those observationsto test the hypothesisthatD2 values for the test locations were 202 ANDEAN BEAR HABITATUSE * Cuesta et al. Table 1. Variables selected to determine Andean bear habitat availability, Oyacachi River Basin, Ecuador, 2000-01. Variable(unitof measure) Elevation(m) Slope (degrees) Rangein study area Datasource and processing 1,584-4,312 Base cartographyof InstituteGeograficoMilitar, Ecuador;scale 1:50,000. Derivedfromelevationusing the Slope functionin TNT MipsGIS. Vegetationcharacterizationsby Baez et al. (1999) and Iturralde et al. (2000) based on superviseddigital classificationof a LandsatTMsatelliteimage (December1998) and fieldtransects. Base cartographyof InstituteGeograficoMilitar, Ecuador,using the Distance Rasterfunctionin TNT MipsGIS. Percentageof disturbedarea (villages,farms)present in a circulararea of 10 km2,centeredon the processing pixel, using the Focalvarietyfunctionin TNTMipsGIS. Densityindex based on the lengthof roads present in a circulararea of 10 km2,centeredon the processing pixel, using the Focalvarietyfunctionin TNTMipsGIS. Calculatedfromelevationbased on McNab(1989). Fourcategories of use intensityderivedfromthe communitymanagementplan and fromthe vegetation characterizationsby Baez et al. (1999) and Iturralde et al. (2000) based on superviseddigitalclassification of a LandsatTMsatelliteimage (December1998) and fieldtransects. 0-76 Vegetationcover Distance to water (m) 0-4,623 Percentageof humandisturbance 0-44.4 Road density 0-4.9 Terrainshape index Landuse intensity -22.4-25.4 1-4 lower comparedwith the 61 randomsites using a MannWhitney U test. Results showed a greaterconcentrationof bear use in montane cloud forests, whereas September-Octoberand November-December reflected a greater use of herbaceous paramoand mixed pairamoforest. Field surveys for bear sign We recorded751 locations with signs of bearactivity; 549 during2000 and 202 during2001. Signs of feeding activity were most frequently observed (53.3%), followed by scats (19.6%),footprints(13.4%),hair (6.9%), tree marks (4.8%), trails (1.2%), and ground nests (0.8%). The combined total distance traversedduring both years of monitoring was 1,018 km with an encounter rate of observed sign of 0.74 records per kilometerof transect. Use of vegetation types The transect surveys showed differences in bear use among the vegetation types between years and among the bimonthly periods (x2 = 34.6, P = 0.001). Throughout the 6 bimonthly monitoring periods, the greatest intensity of use occurred in montane cloud forest and herbaceous pairamo(Table 2). Our results indicated a relatively continuous use of all vegetation types throughoutthe year except for swampy paramo andmixed paramoforest,whereuse was irregular(Table 2). The periods January-Februaryand July-August Habitat use We calculated D2 values for each pixel in the Oyacachi River Basin based on the original 437 locations of bear sign (Fig. 3); D2 values within the study arearangedfrom 2.6 to 10,991.1, with a mean of 320.8 (SD = 1,069.9). D2 values correspondingto the 437 model input positions ranged from 2.7 to 286.3 ( = 11.7, SD = 23.0), whereas the test locations (independent locations of bear sign) had D2 values ranging from 2.9 to 5,012.3 (x = 131.6, SD = 681.3; Fig. 4). Based on the cumulativefrequency curve of the D2 values for the 61 test locations, we identified 5 percentiles (23.0%, 47.6%, 86.9%, 96.7%, and 100%) to define classes of habitatsuitability:0 < D2 < 7.2 for class 1, 7.2 < D2 < 10.1 for class 2, 10.1 < D2 < 23.7 for class 3, 23.7 < D2 < 200.0 for class 4, and D2 > 200.0 for class 5 (Table3, Figs. 3 and 4). Ninety percent of the independentbear sign locations were below a D2 value of 26; in other words, when we encounteredbear sign, the probabilityof an associatedD2 value >26 was <10%. Ursus 14(2):198-209 (2003) ANDEANBEARHABITAT USE * Cuestaet al. 203 Table 2. Percent of Andean bear sign locations found in vegetation types by bimonthly period, Oyacachi River basin, Ecuador, for 2000 (n = 549 locations) and 2001 (n = 202 locations). Blank fields indicate absence of Andean bear sign. Year Vegetation typea % of locations Jan-Feb Mar-Apr May-Jun Jul-Aug Sep-Oct 2000 BNM BPM BSV-ma PA PH Total 47.6 3.6 21.5 7.1 20.2 100.0 30.8 4.1 20.0 39.6 45.5 42.5 100.0 17.0 5.7 37.7 100.0 27.2 2.3 25.0 100.0 14.0 32.0 18.0 2.0 34.0 100.0 2001 BNM BPM BSV-ma PA PH Total 44.7 13.2 18.4 5.3 18.4 100.0 55.6 2.2 24.4 2.2 15.6 100.0 22.5 30.0 20.0 58.6 6.9 3.4 25.0 28.6 32.1 27.5 100.0 31.1 100.0 14.3 100.0 2.6 Nov-Dec 22.0 14.6 34.1 29.3 100.0 13.6 13.6 31.8 4.5 36.5 100.0 aBNM= montanecloudforest, BPM= mixedparamoforest, BSV-ma= uppermontaneevergreenforest, PA= swampypramo, PH herbaceous paramo. The spatial distributionof the first 4 D2 classes in the study area was fairly homogeneous (Fig. 3). The class 1 area occupied 86.3 km2, equivalentto 12.0% of the study area extent (Table 3, Fig. 3). Pixels in that class were distributed between 1,955 and 4,391 m elevation (x = 3,219, SD = 669) with slopes ranging from 0 to 78 degrees (x = 50.5, SD = 14.8). Almost half (38.0 km2) of the class 1 zone was associated with montane cloud forests and 38.3 km2 with herbaceous paramo.The distributionof class 1 areas in relation to road density and disturbedareas indicated a negative association of these elements with bear habitat use. Areas with a high percentage of human disturbance tended to be associated with the 2 lowest suitability classes (Fig. 3). The D2 values associated with the 61 random locations were greater than those associated with the test locations, with the greatest differences occurring in the lowest range of D2 values (Table 3, Fig. 5). The Mann-Whitney U-test indicated that the differences between the 2 distributionsof D2 values were significant (Z = -2.78, P = 0.006). Discussion Use of vegetation types Members of the Ursidae family tend to concentrate theiruse of the landscapein the most productivehabitats (Schoen 1990); this may partiallyexplain the observed Ursus 14(2):198-209 (2003) variationin use intensity of the vegetation types within the OyacachiRiver Basin. Ourresults showed a distinct patternof seasonal use, suggesting seasonal variationof food resourcesavailable within each vegetation type in the study area. Peyton (1980, 1984) and Suarez (1988) also observed marked seasonal use patterns among different habitatsin other portions of the range of the Andean bear. In our study, alder forests (Alnus acuminata) representedthe only vegetation type for which we had no recordof bearuse. We speculatethatthis low use was due to the lack of primarybear foods and scant cover provided by alder forests (Baez et al. 1999, Iturralde et al. 2000). The bear use that we observed for the remaining vegetation types seemed to be associated with food resources.One of the primaryfood sources of Andean bears is the meristematic tissue of various terrestrialand epiphyticbromeliads(Davis 1955). Troya (2001) identified several giant bromeliads (Puya spp.) and the terrestrialbromeliads Greigia vulcanica and Greigia mulfordii as the most frequently consumed species in the Oyacachi study area. Terrestrialbromeliads are most common in the pairamoareas,whereas epiphytic bromeliads occur in the upper montane evergreenforest. Bear use of upper montaneevergreen forest and the montane cloud forest was particularly seasonal (Jan-Feb and Jul-Aug; Table 2), which may reflect the importanceof fruit trees, such as Hyeronima spp. Indeed, fruiting rates in these forests tend to be greatestduringthose periods (Skov 1997, Troya 2001). 204 ANDEAN BEAR HABITAT USE * Cuesta et al. Vegetation type Elder Forest Water | Grassland _ Herbaceous Paramo Swampy Paramo Mixed Paramo Forest '[ E 0S ] Cloud Forest ] Community boundary Upper M. Evergreen Forest Landslide --... 3 0 Unpaved road Burned Paramo _ Path . Transect 6 Km M Fig. 2. Vegetation types of the Oyacachi River Basin, Ecuador, for a study of habitat use by Andean bears, 2000-01. N 0 3 6Km A, Fig. 3. Mahalanobis distance (D2) values for a study of Andean bear habitat use in the Oyacachi River Basin, Ecuador, 2000-01. Five habitat suitability classes were defined; class 1: 0 < D2 < 7.2; class 2: 7.2 < D2 < 10.1; class 3: 10.1 < D2 < 23.7; class 4: 23.7 < D2 < 200; and class 5: D2 > 200. Ursus 14(2):198-209 (2003) ANDEAN BEAR HABITATUSE * 60 A A 50- > ? 40 o I , - -30- ' I1/ 20 10 O i ! 0 , tt 5 lI I 10 15 I , l 20 25 30 35 40 Mahalanobisdistance , 45 l E 50 55 Fig. 4. Cumulative frequency of 61 test locations (independent locations of bear sign) and associated Mahalanobis distance values for a study of Andean bear habitat use in the Oyacachi River Basin, Ecuador, 2000-01. Four outlying observations were excluded for scaling purposes. Arrows indicate discontinuities used to define classes of Mahalanobis distance values. Results from studies in Venezuela also indicate that Andeanbearspreferto feed in paramoareaswith a high concentration of giant bromeliads (Puya spp.) and forested areas where trees have a high concentrationof epiphyticbromeliads(Gusmaniaspp.) (Goldstein 1992). Although food is only one of many resources that large mammalsselect for, it is particularlyimportantfor bearsbecause of theirlarge size and relativelyinefficient digestive system (Pritchardand Robbins 1990). Our study suggests that selection of vegetationtypes may be associatedwith food abundance,but it will be necessary to conduct additional studies to determine availability and quality of food resourcesthroughoutthe year. That informationshould be helpful to betterinterpretseasonal movement patternsof Andean bears among vegetation types. Habitat suitability Our analysis indicates that the Mahalanobisdistance statistic can be an effective measureto define potential suitability of Andean bear habitat. For the Oyacachi River Basin, bearlocations were associatedwith low D2 values (mode = 3.6, n = 437). Of the independenttest locations, 23.0% occurredin areaswith D2 values <7.2, 47.6% occurred with D2 values <10.1, and 86.9% occurredwith D2 values <23.7 (Fig. 4). The distribution of D2 values associatedwith the 61 null model (random) locations was different from those associated with the test locations (Table 3, Fig. 5); this difference was mostly evident for the low range of D2 values, which Ursus 14(2):198-209 (2003) Cuestaet al. 205 representsthe best bear habitat. Many bear locations, however, were in areas with intermediateD2 values (Table 3), which resultedin convergenceof the cumulative D2 values for the null model (randomlocations) and the test locations (Fig. 5). This convergence may indicatethat a large portionof the study arearepresents habitatsfor which model predictionsare marginal.We speculatethat a largersamplethan 61 test locationsmay be needed to improve the power to test the model. Despite the probablelack of power to properlytest all aspects of the Mahalanobisdistance model, we found that D2 values successfully predictedareas that receive frequentuse by Andean bears. Elevationand vegetationtype seemed to have a strong influenceon habitatsuitabilityin the study area(Fig. 3). The 2 variables are intimately linked to the type and concentrationof food resourcesand protectionprovided by the structureof certainvegetationtypes. Such habitat components are importantfor all bear species, and the Andean bear is no exception (Peyton 1980, Clevenger et al. 1992). The montanecloud forest has been defined by several authors (Peyton 1980, Yerena and Torres 1994) as a criticalecosystem to supportviable bearpopulations. The high ac-diversityof these forests regarding herbs, vines, and epiphytes (Jorgensenet al. 1995), in conjunction with the high P-diversityof Andean ecosystems (Jorgensenet al. 1999), likely provides many of the resources needed by Andean bears. To better understandthe seasonal movement patternsof Andean bears, it is importantto study the variability of food availabilityand abundanceduringdifferentseasons and among the vegetation types used by this species. Othertopographicvariablesalso seemed to influence model predictions. Andean bears in our study area seemed to use a wide range of landscapefeatures,from steep ravinesto flat areas.The slope associatedwith bear sign locations ranged from 0? to 70? (x = 20.8?, SD = 14.9). However, elevations above 4,300 m with slopes >60? were not used; these areas generally are covered with a small shrub, Loricaria thuyoides, which is not used by Andean bears for cover or food. The influence of anthropogenic variables on bear habitat use was evident as well (Fig. 3), but seemed to occur on a larger scale compared with vegetation type. Some human activities, such as road construction or agriculturalfrontierexpansion, have been correlated with substantial range and population reduction of Europeanbrown bears (Ursus arctos) (Elgmork 1978, Clevengeret al. 1992) and Americanblack bears(Ursus americanus)(McLellanand Shackleton 1988). Because similarprocesses are occurringin Andean bear habitat, 206 ANDEANBEARHABITAT USE * Cuesta et al. Table 3. Five classes of Mahalanobis distance values (D2),their area, frequency of occurrence of test locations (independent locations of Andean bear sign), and frequency of null model locations for a study of Andean bear habitat in the Oyacachi River Basin, Ecuador, 2000-01. D2 class Class 1 0 < D2 < 7.2 Class 2 7.2 < D2 < 10.1 Class 3 10.1 < D2 < 23.7 Class 4 23.7 < D2 < 200 Class 5 2 > 200 Total Area (km2) Percent of area (%) Frequency of test locations Percent test locations (%) Frequency of null model locations Percent of null model locations (%) 86.3 12.0 14 23.0 7 11.5 172.8 23.9 15 24.6 19 31.2 308.3 42.7 24 39.3 24 39.3 46.5 6.4 6 9.8 3 4.9 108.6 15.0 100.0 2 3.3 8 13.1 61 100.0 61 100.0 722.5 Our method of defining habitat suitability has a numberof desirablequalities to promoteunderstanding of wildlife habitatuse and apply this understanding to conservationmanagement.Biological, topographical, and anthropogenicvariablesall are partof Andean bear habitatand were includedin the model. Furthermore,all variables are also present in landscapes of the eastern Andes in northernEcuador to allow model extrapolation. Applicationof the model at a regional scale would allow the definition of conservation priority areas in Ecuador that could help guide conservationof critical habitatsto maintainviable populationsof the species in the future. Predictions from the habitat model can be used in conjunctionwith other data to identify potential sites for conservationpurchases,to establishor preserve movement corridors,and to mitigatenegative effects of certain land managementpractices, such as construction of roadsor increasingintensityof humanuse (Clark et al. 1993). The informationobtainedduringthis study, however, should not be seen as absolute criteriafor the definitionof Andean bear conservationareas. It will be Management and research implications necessary to conduct further model testing and deThe temporalvariationin bear use that we observed define conservation priorities, and to find among the different vegetation types may indicate velopment, to achieve a better integrationof mechanisms effective continuous movements of bears within an altitudinal the socio-economic, political, and legal issues to ensure gradient. Thus, the integrity of habitats along this effectiveness and viability of defined conservation the altitudinalrange should be considered when defining in the long term. areas conservationareasfor Andeanbears so thatthese moveother studies have alreadyshown (e.g., Clevenger As ments can be maintained(Yerenaand Torres 1994). For and Purroy 1996), biological monitoring by means of example, expansion of the agriculturalfrontier (low sign records can be highly effective to generate basic elevations) into montanecloud forests (high elevations) ecological dataon habitatuse, diet, and even population may reduce bear access to importantfood sources, such that as tree-bornefruits, that are not present in the pairamos trends.For example, the encounterrate of bear sign be it that indicates may obtained we transect) (0.74/km (highest elevations). the inclusion of anthropologicalvariables in the model was importantto identify bear habitatsthat are within humaninfluence zones. We used 5 classes of D2 values (Table 3) to define broad categories of bear habitat suitability (Fig. 3). Class 1 representedfeaturesclosest to the "ideal"habitat of the Andean bear, as measured from known bear locations. The second and third classes included areas that have increasingly different habitat features from those ideal conditions;those areasmay be consideredof lower habitatsuitabilitybut still importantbear habitat within the study area. The fourth class included areas that generally surroundedthe 3 previous classes, representing marginal habitat that received only occasional use by bearsand may not be suitablefor permanentbear presence. However, class 4 areas may be crucial to connect importantclass 1 areas (Boitani et al. 1999). Finally, the fifth class comprised areas that generally were not used by Andean bears. Ursus 14(2):198-209 (2003) 208 ANDEANBEARHABITAT USE * Cuesta et al. EcoCiencia, Ministerio del Ambiente y Uni6n Mundial parala Conservaci6nde la Naturaleza,Quito, Ecuador.(In Spanish.) AND D. SANCHEZ. 2001. Metodos para , M. PERALVO, investigarla disponibilidaddel habitat del oso andino: el caso de la cuenca del rio Oyacachi, Ecuador. Serie Biorreserva del C6ndor No. 1. EcoCiencia y Proyecto Biorreservadel C6ndor, Quito, Ecuador.(In Spanish.) DAVIS,D. 1955. Masticatoryapparatusin the spectacledbear, Tremarctosornatus. FieldianaZoology 37(2):25-46. J. 1986. The design of a naturereserve system for DIAMOND, Indonesian New Guinea. Pages 485-503 in M. Soule, editor. Conservationbiology: the science of scarcity and diversity. Sinauer Associates, Sunderland,Massachusetts, USA. DONOVAN,M., D. RABE, AND C. OLSON. 1987. Use of geographic information systems to develop habitat suitability models. Wildlife Society Bulletin 15:574-579. K. 1978. Human impact on a brown bear popuELGMORK, lation. Biological Conservation13:81-103. 1995. Local communities FIALLO,E.A., AND S.K. JACOBSON. and protected areas: Attitudes of rural residents towards conservation and Machalilla National Park, Ecuador. EnvironmentalConservation22(3):241-249. I. 1992. Spectacled bear predation and feeding GOLDSTEIN, behavioron livestock in Venezuela. Studies in Neotropical Fauna and Environment27(1):1-5. C. 2000. IUCN red list of threatened species. HILTON-TAYLOR, The IUCN Species Survival Commission/IUCN. Gland, Switzerlandand Cambridge,United Kingdom. ITURRALDE,J.A., N. OLEAS, F. CUESTA, AND M. PERALVO.2000. Caracterizaci6nvegetal de la cuenca baja del rio Oyacachi, Reserva Ecol6gica Cayambe-Coca. EcoCiencia, Quito, Ecuador.(In Spanish.) JORGENSEN, P.M., C. ULLOA-ULLOA, J. MADSEN, AND R. VALENCIA.1995. A floristic analysis of the high Andes of Ecuador.Pages 221-237 in S.P. Churchill,H. Balslev, E. Forero,and L. Luteyn, editors. Biodiversity and conservation of neotropical montane forest. The New York Botanical Garden,New York, USA. D. NEILL,AND S. LE6N-YANEZ.1999. Introduction. Pages 1-108 in P.M. Jorgensen, and S. Leon-Yanez, editors. Catalogue of the vascular plants of Ecuador. Herbarium de la Universidad Catolica del Ecuador, Herbariumof the Museo Ecuatorianode Ciencias Naturales, and Departmentof Systematic Botany of Aahrus University, Quito, Ecuador. KELLERT, S. 1994. Public attitudes toward bears and their conservation.Interational Conference on Bear Research and Management9(1):43-50. KENDALL,K.C., L.H. METZGAR, D.A. PATTERSON,AND B.M. STEELE.1992. Power of sign surveys to monitorpopulation trends.Ecological Applications2:422-430. L6PEZ,R.S. 1992. Diagn6stico socioecon6mico de la Reserva Ecologica Cayambe-Coca.Estudios en aireasprotegidas/ FundacionNatura.Volume 2. Quito,Ecuador.(In Spanish.) 1988. Grizzly bears MCLELLAN,B., AND D.M. SHACKLETON. and resource extraction industries: effects of roads on behaviour,habitatuse and demography.Journalof Applied Ecology 25:451-460. McNAB,W.H. 1989. Terrainshapeindex: quantifyingeffect of minorlandformson tree height. Forest Science 35:91-104. 1997. The people and their MORALES, P., ANDI. SCHJELLERUP. culture.Pages 28-65 in Oyacachi-peopleand biodiversity. Technical Report No. 2. Centre for Research on Cultural and Biological Diversity of Andean Rainforests, Ronde, Denmark. NAMS,V. 1989. Effects of radiotelemetryerroron sample size and bias when testing for habitat selection. Canadian Journalof Zoology 67:1631-1636. B. 1980. Ecology, distributionand food habits of PEYTON, spectacled bears, Tremarctosornatus, in Peru. Journalof Mammalogy61:639-652. . 1984. Spectacled bear habitat use in the historical Sanctuaryof Machu Picchu, Peru. Thesis, University of Montana,Missoula, Montana,USA. . 1999. Spectacledbearconservationactionplan. Pages 157-198 in C. Servheen, S. Herrero, and B. Peyton, compilers. Bears. Status survey and conservation action plan. IUCN/SSC Bear and Polar Bear Specialist Groups. IUCN, Gland, Switzerlandand Cambridge,United Kingdom. . C. SERVHEEN,AND S. HERRERO.1999. An overview of bear conservationplanning and implementation.Pages 824 in C. Servheen, S. Herrero,and B. Peyton, compilers. Bears. Status survey and conservationaction plan. IUCN/ SSC Bear and PolarBear SpecialistGroups.IUCN, Gland, Switzerlandand Cambridge,United Kingdom. G.T., AND C.T. ROBBINS.1990. Digestive PRITCHARD, and metabolicefficiencies of grizzly and black bears. Canadian Journalof Zoology 68:1645-1651. J.W. 1990. Bear habitatmanagement:a review and SCHOEN, future perspective. International Conference on Bear Researchand Management8:143-154. 1999. Areas SIERRA,R., F. CAMPOS,AND J. CHAMBERLIN. prioritariaspara la conservaci6nde la biodiversidaden el Ecuadorcontinental.Un estudiobasadoen la biodiversidad de ecosistemas y su omitofauna.Ministeriodel Ambiente, Proyecto INEFAN/GEF/BIRF, EcoCiencia y Wildlife ConservationSociety, Quito, Ecuador.(In Spanish.) SKOV, F. 1997. Physical environment. Pages 1-12 in Oyacachi-people and biodiversity. Technical Report No. 2. Centrefor Researchon Culturaland Biological Diversity of Andean Rainforests,Ronde, Denmark. SOKAL,R., ANDJ. ROHLF.1981. Biometry-The principles and practiceof statisticsin biological research.Second edition. W.H. Freemanand Company,New York,New York,USA. Ursus 14(2):198-209 (2003) BEARHABITAT ANDEAN USE * Cuesta et al. SPSS. 1998. SPSS for Windows. Release 9.0. Standard version.SPSS,Chicago,Illinois,USA. L. 1988. Seasonal distributionand food habits of SUAREZ, spectacledbears (Tremarctosornatus) in the highlandsof Ecuador. Studies of Neotropical Fauna and Environment 23(3):133-136. TIRIRA, D., F. CUESTA, AND L. SUAREZ.2001. Introducci6n. Pages 1-7 in D. Tirira,editor. Librorojo de los mamfferos del Ecuador. Serie libros rojos del Ecuador, Tomo 1. Publicaci6n Especial sobre los mamiferos del Ecuador4. SIMBIOE/EcoCiencia/ Ministerio del Ambiente/IUCN, Quito, Ecuador.(In Spanish.) TROYA,V. 2001. Habitos alimentarios del oso andino (Tremarctosornatus) en diferentes formacionesvegetales de la cuenca del rio Oyacachi, Reserva Ecologica Cayambe-Coca, Ecuador. Thesis, Pontificia Universidad Catolica del Ecuador (PUCE), Quito, Ecuador. (In Spanish.) VAN MANEN, F.T., J.D. CLARK,S.E. SCHLARBAUM,K. JOHNSON, ANDG. TAYLOR.2002. A model to predict the occurrence of surviving butternuttrees in the southern Blue Ridge Mountains.Pages 491-497 in J.M. Scott, P.J. Heglund,and Ursus 14(2):198-209 (2003) 209 M.L. Morrison,editors. Symposium on predictingspecies occurrences: issues of scale and accuracy. Island Press, Covelo, California,USA. AND W.J. WEMMER,C., T.H. KUNZ, G. LUNDIE-JENKINS, MCSHEA. 1996. Mammalian sign. Pages 157-176 in D.E. Wilson, F.R. Cole, J.D. Nichols, R. Rudran, and M.S. Foster, editors. Measuringand monitoringbiological diversity. Standard methods for mammals. Smithsonian InstitutionPress, Washington,D.C., USA. D.S. 1987. From fragmentation to extinction. WILCOVE, NaturalAreas 7:23-29. 1994. Spectacled bear conserYERENA,E., ANDD. TORRES. vation and dispersal corridorsin Venezuela. International Conferenceon Bear Researchand Management9(1):169172. YOUNG,K. 1999. Dispersal of Styrax ovatus seeds by the spectacledbear (Tremarctosornatus). Vida Silvestre Neotropical2(2):68-69. Received: 30 October2002 Accepted:9 June 2003 Associate Editor: R.B. Harris