Delineating Priority Habitat Areas for the Conservation of Andean Bears in Northern Ecuador

Delineating priority habitat areas for the conservation of Andean bears in northern Ecuador

Manuel F. Peralvo1'3, Francisco Cuesta1 4, and Frank van Manen2'5

1EcoCiencia,Francisco Salazar E14-34, PO Box 12-17-257,Quito, Ecuador 2US GeologicalSurvey, SouthernAppalachian Field Branch,University of Tennessee, 274 EllingtonPlant Sciences, Knoxville,TN 37996, USA

Abstract: We sought to identify priority areas for the conservation of Andean bear (Tremarctos ornatus) habitatin the northernportion of the easternAndean cordillerain Ecuador.The study area included pairamoand montane forest habitats within the Antisana and Cayambe-Cocaecological reserves, and unprotectedareas north of these reserveswith elevations rangingfrom 1,800 to 4,300 m. We collected data on bear occurrencealong 53 transectsduring 2000-01 in the OyacachiRiver basin, an areaof indigenouscommunities within the Cayambe-CocaEcological Reserve. We used those data and a set of 7 environmentalvariables to predictsuitability of Andean bear habitatusing Mahalanobis distance, a multivariatemeasure of dissimilarity.The Mahalanobisdistance values were classified into 5 classes of habitatsuitability and generalizedto a resolutionof 1,650-m x 1,650-m grid cells. Clusters of grid cells with high suitabilityvalues were delineatedfrom the generalizedmodel and defined as importanthabitat areas (IHAs) for conservation.The IHAs were rankedusing a weighted index that includedfactors of elevationrange, influence from disturbedareas, and currentconservation status. We identified 12 IHAs, which were mainly associated with paramoand cloud forest habitats;2 of these areashave high conservationpriorities because they are outside existing reserves and close to areas of human pressure.The distributionof the IHAs highlighted the role of human land use as the main source of fragmentationof Andean bear habitat in this region, emphasizing the importance of preserving habitat connectivity to allow the seasonal movements among habitat types that we documentedfor this species. Furthermore,the existence of areas with high habitatsuitability close to areas of intense humanuse indicates the importanceof bear-humanconflict managementas a critical Andean bear conservation strategy. We suggest that a promising conservationopportunity for this species is linked to its occurrencein highland habitats,which play a key role in the maintenanceof long-termwater supplies.

Key words: Andean bear, Andes Mountains,conservation planning, Ecuador, geographical information system, GIS, habitatsuitability, Mahalanobis distance, Tremarctosornatus

Ursus16(2):222-233 (2005)

Habitat loss is the single largest cause of species biodiversityloss in the shortterm necessitate prioritizing of 2 main criteria: endangermentworldwide (Hilton-Taylor2000, Pitman conservation efforts by means et al. 2002). Defining effective conservation areas for selection of large tracts of relatively undisturbed associated with biodiversity protection is a fundamental necessity in vegetation and selection of habitats countries where an accelerated degradation and de- species thatrequire large areas(Poiani et al. 2001). Such structionof naturalhabitats is fragmentingand isolating landscape species often occur in ecologically diverse in time and wildlife populations. However, the lack of biological areas, and their habitatrequirements space informationand difficulties in addressingthe causes for make them particularly vulnerable to land-use and resource-harvestingpractices of people. Thus, identifying ecological requirements of landscape species is of of Texas 3Presentaddress: Department Geography,University useful to aid the and managementof landscapes USA; design at Austin,Austin, TX, 78712-1098, peralvomf@mail. conservation Sander- utexas.edu for biodiversity (Simberloff 1999, 4F.CuestaCamachostudent.uva.nl [email protected] son et al. 2002).

222 DELINEATINGANDEAN BEAR CONSERVATION AREAS * Peralvo et al. 223

Andean bear distribution in Ecuador is fragmented among many habitatpatches (subpopulations),with the largest patches distributedwithin the eastern Andean range (n = 10, x = 4,340 km2). Many of the remaining habitatpatches (n = 15) in Ecuadorare small (x = 1,140 Location in Ecuador km2; EcoCiencia/WorldWildlife Fund, Quito, Ecuador, Legend unpublisheddata). Although no reliablepopulation data * Humansettlement are available, the Cayambe-Coca Ecological Reserve Mainroad and Sangay National Park likely contain the largest /aProtected area limit a es Andean bear populationsin Ecuador(Suarez 1999). River The Andeanbear's broadecological requirementsand Oyacachi study area its seasonal use of different habitats make it an appropriate species on which to base conservation planning to preserve the high biodiversity of these (Yerena 1998, 1999). The conserva- N ecosystems Peyton tion of Andean bears in Ecuador depends on proper Darkershading protectionand managementof their habitat. Studies to representsareas determine the distribution of the Andean bear and 0 10 20 km ti between E 'R .1,8004,300 m availability of suitable habitats can be important to inelevation delineate conservationunits. Knowledge of the extent, shape, and spatialdistribution of those units is important Fig. 1. Study area in the northeastern Ecuadorian to help reduce the effects of fragmentation,thereby Andes to identify priority conservation areas for improving the long-term viability of Andean bear Andean bears. populations (Yerena and Torres 1994, Suarez 1999). However, importanthabitat areas for Andean bearshave Piramo and cloud forests represent the primary not been delineatedfor northernEcuador. Therefore, we habitats of Andean bears (Tremarctosornatus) in the sought to delineate important habitat areas in the northernAndes (Ecuador, Colombia, and Venezuela). northern portion of the eastern Ecuadorian Andean The seasonal variabilityin food availability in habitats cordilleraand ranktheir relative importancefor Andean used by Andean bears may result in wide-ranging bear conservation. movements, which are increasingly obstructed by the loss of favorable habitat due to advancing agricultural frontiers and expanding infrastructure.Peyton et al. Study area (1998) estimated that approximately 18% of potential Our study area included the Cayambe-CocaEcolog- bearrange in the Andes Mountainswas containedwithin ical Reserve, the northern section of the Antisana 58 protectedareas. However, many of those areas were Ecological Reserve, and adjacent paramo and cloud small, particularlyin the northernAndes. The median forest areas in the Sucumbios and Carchi provinces in size of 43 parks in Venezuela, Colombia, and Ecuador the northeasternAndes Cordilleraof Ecuador (Fig. 1). was 1,250 km2, which may be insufficient to ensure The study area covered 6,048 km2, and elevations long-termspecies persistence (Peyton et al. 1998). ranged from 1,800 to 4,300 m. The Cayambe-Cocaand In Ecuador,the Andean bear inhabits approximately Antisana ecological reserves made up 47.7% (2,884 58,000 km2 of paamo and cloud forest habitats, of km2)and 3.2% (194 km2),respectively, of the study area which almost 19,000 km2 are within the Ecuadorian (Fig. 1). Due to the geographicalposition of the area,its system of protected areas and remaining areas are altitudinalrange, and the differentclimatic strataof the undevelopedbut not legally protected(Rodriguez et al. region, the flora and fauna are extremely diverse 2003). Those habitats have been reduced by approxi- (Paredeset al. 1999). These ecosystems are known for mately 40% from their original distribution, thereby high alpha diversity and endemism and for their fragile creatinga series of "islands" in the regional landscape soils and topography(World Wildlife Fund 2001); the (Sierra et al. 1999). The conversion of those areas to area is part of the Tropical Andes biome defined by agricultural uses has fragmented bear habitat and Myers et al. (2000) as one of the 10 worldwide isolated populations (Peyton 1999, Suarez 1999). biodiversity "hotspots".

Ursus 16(2):222-233 (2005) 224 DELINEATINGANDEAN BEAR CONSERVATION AREAS * Peralvo et al.

In the higherelevations of the studyarea, the prominent Table 1. Altitudinal range and area of land-cover geomorphologicfeatures are of glacial origin and include types within the northeastern Ecuadorian Andes roches moutonnees knobs study area. Adapted from a study of Andean bear cirques, (smooth projecting habitat in the River basin et al. from the land and filled with Oyacachi (Cuesta surface), U-shaped valleys 2003). morainedeposits, mudslides, and lahars(volcanic mud- Altitudinal Area of flows). The terrain of the lower elevations is highly Proportion Land-cover type range (m) (km2) study area (%) dissected and covered with cloud forests;heterogeneous Humid 30.4 landscapesare predominant(Cuesta et al. 2003). paramo 3,400-4,300 1,837 Highevergreen upper Four major watersheds begin inside the Cayambe- montaneforest 3,000-3,600 1,428 23.6 Coca Ecological Reserve: the Aguarico River (1,395 Montanecloud forest 1,800-3,000 1,979 32.7 km2), the Quijos River (2,504 km2), the Mira River (67 Disturbedareas 804 13.3 km2), and the EsmeraldasRiver (2,624 km2).Because of Total 6,048 100.0 these hydrological resources, this protected area repre- sents the most importantwater reserve of the northern accompanying study in the Oyacachi River basin to portion of the country (Paredes et al. 1999). The area predict bear habitat suitability on a regional scale consists of 4 majorland-cover types (Valenciaet al. 1999; (Cuesta et al. 2001, 2003). The 721-km2 Oyacachi EcoCiencia, Quito, Ecuador, unpublishedreports): (1) study area was located in the southern portion of the humid paramo(includes herbaceouspairamo, frailejones Cayambe-CocaEcological Reserve (Fig. 1). The field [Espeletia pycnophylla] pairamo,cushion paramo, and data for that model were based on bear sign locations mixed-forestpairamo), (2) high evergreenupper montane that were collected along 53 transectsduring 2000-01 forest, (3) montanecloud forest, and (4) disturbedareas (Cuesta et al. 2001, 2003). Habitat suitability was (Table 1). The disturbedland-cover type correspondedto determinedusing the multivariatestatistic Mahalanobis landscapesdominated by a matrixof anthropogenicland distance (Clarket al. 1993) and a set of environmental uses such as agricultureand urbanareas. This land-cover variablesconsidered to be the most relevant to Andean type is predominantin the inter-Andeanvalleys located bearhabitat use (Cuestaet al. 2003). We used thatmodel west of the study area,which have been used by humans to extrapolatethe Mahalanobisdistance values to the for millennia. In contrast,most of the disturbedareas in regional study area. Mahalanobisdistance is a multivar- the eastern slopes of the Andes correspond to recent iate measure of dissimilarity (Clark et al. 1993). As deforestationand colonization frontiersassociated with such, our extrapolationrepresents how dissimilar (or roads built to connect the Ecuadorianhighlands with the similar) habitats are compared with those used by lowlands in the Amazon basin (Young 1998). Human Andean bears in the Oyacachi River basin. Similar activities (such as extensive cattle breeding,hunting) in Mahalanobisdistance values can suggest similarhabitat non-disturbedland-cover types (paramo and montane potential, despite differences in habitat configurations forests) are limited and have relativelylow impacton the (Knick and Rotenberry 1998). An advantage of this structureof those naturalhabitats. However, the intensity statistic is that it accounts for correlationamong vari- of these activities varies spatially with accessibility and ables and avoids the assumptionof multivariatenormality proximityto populatedareas. (Clark 1993, Knick and Rotenberry1998). We used geographic information system (GIS) software(TNT Mips, V6.3, MicroImages,Inc., Lincoln, Methods Nebraska, USA) to build a raster GIS database for 7 the We rankedareas for the conservationof Andean bear environmentalvariables (30- x 30-m grid cells) for terrain index, habitatin the study areausing a 3-stage approachby (1) entire region: elevation, slope, shape road of a statistical model to predict suitability of distanceto rivers, density, percentage developed applying the Andean bear habitatbased on occurrencedata, (2) using area,and land-cover type (Table2). Despite relatively coarse scale of the source data, we chose the predictions of that model to delineate important topographic of 30 m to maintain with the habitat areas, and (3) ranking and assessing the con- a resolution compatibility Mahalanobismodel for the Oyacachi water- servationimportance of those areas. generated shed acrossthe whole region. In addition,high resolution have been Model application digital elevation models (DEM) successfully informationin A quantitative model is important to objectively extracted from small-scale topographic delineate areas of suitablebear habitat.We relied on an areas where large elevation gradients and extreme

Ursus 16(2):222-233 (2005) DELINEATINGANDEAN BEAR CONSERVATION AREAS * Peralvo et al. 225

Table 2. Geographic information system (GIS)variables used in this study to determine suitability of Andean bear habitat for the northeastern Ecuadorian Andes. Variable Description Source Elevation (m) Elevation(m) Digitalelevation model (DEM) interpolatedfrom contour lines (1:250,000 scale); Geographic Instituteof the Armyof Ecuador(IGM) Slope (degrees) Slope steepness (degrees) Calculatedfrom elevation using the Slope function(TNT Mips) Terrain index shape Meandifference in elevationbetween the Calculatedbased on McNab(1989) centralpixel and its 24 neighborsin a windowof 5 x 5 pixels Land-covertype One disturbedand three natural Superviseddigital classification land-covertypes (see text) of a LandsatTM satellite image (Path 10, Row 60; December20, 1998; 30-m resolution) Distance to rivers Distance to the nearest stream (m) Calculatedfrom a stream coverage (1:50,000 scale) producedby the IGMwith the Distance function (TNTMips) Road density Ratioof roadpixels to the totalnumber of Calculatedfrom a rasterlayer of pixels withina 10-km2circular window roads (1:50,000 scale) Percent of developed area Ratioof disturbedarea pixels to the total Calculatedusing the disturbed numberof pixels withina areas class of the land-cover 10-km2circular window type rasterlayer

relief topographic generate high planimetricdensity of (30- x 30-m grid cells). We calculatedthe most frequent data contour topographic (i.e., lines; Zomer et al. 2002). class value (mode) inside each square to assign that Mahalanobis distance was calculated for the regional value to the entire 2.72-km2 window area. Because of area that study using GIS database.The model results the lack of baseline ecological dataon Andeanbears, we were to areas with extrapolated only biophysical defined an area of 2.72 km2 based on averagedaily area characteristicssimilar to the Oyacachistudy area(Cuesta requirements of adult female American black bears et al. 2001, Because elevation is 2003). the primary (Ursus americanus;Clark et al. 1993). That process was we ecological gradient, restrictedmodel applicationto designed to reduce the spatial resolution of the habitat areas between 1,800 m and 4,300 which was the m, data, thereby facilitating the identificationof IHAs by elevation monitoredin the range Oyacachi study. delineating contiguous areas of >10 grid cells (2.72- km2) with class 1 values. We consideredgrid cells to be adjacentwhen cells with class 1 values shared at least Delineating important habitat areas (IHAs) 1 entire with a cell of We used the results from the Mahalanobisdistance edge neighboring class 1. No reliable informationexists on minimum model to identify IHAs for the conservationof Andean viable populations for Andean bears, so we defined this bears by delineating zones with high habitat suitability arbitrary minimumarea as a means to define IHAs with values (low Mahalanobisdistance values). Cuesta et al. favorable habitatthat would potentiallysupport a small (2003) tested the habitat model using independent population of bears. data locations and divided the associated distance values Preliminarypopulation from the Oyacachi study area that 6 or 7 into 5 range classes. These classes were based on suggest females may occupy a contiguous area of 27 km2 discerniblediscontinuities of percentilesof a cumulative (F. Cuesta, EcoCiencia, unpublisheddata). Until more detailed frequency curve (23.0%, 47.6%, 86.9%, 96.7%, and ecological data become available,these IHAs a 100%),representing 5 habitatsuitability classes (Boitani simply represent starting point for conservation et al. 1999). Areas with Mahalanobis distance values prioritizing efforts. <7.2 were considered the most suitable habitat (class 1). We used those 5 classes for our subsequent anal- Ranking of important habitat areas yses by overlaying a squareGIS "window" (2.72-km2; After identifyingand delineatingthe IHAs, we ranked 1,650 x 1,650 m) with the of Mahalanobis grid values their importance based on 3 criteria: (1) altitudinal Ursus 16(2):222-233 (2005) 226 DELINEATINGANDEAN BEAR CONSERVATION AREAS * Peralvo et al. range, (2) human influence, and (3) current conser- area outside the zone of influence. A value of 0 was vation status. We assigned scores between 0 and 1 for assigned to IHAs entirely outside this zone. each of these factors, except for the altitudinalrange Status of protection. By means of this criterion, criterion, which was weighted by assigning values we gave priorityto IHAs that were entirely or partially between 0 and 2. We used the overall scores to rank outside of the Cayambe-Cocaand Antisana ecological the areas and to identify preliminary strategies for reserves (Fig. 1). We assigned values to each IHA by their conservation.The scores were additive on a scale calculatingthe percentof their extent outside ecological of 0-4, with higher scores indicating greater priority reserves. Areas completely outside reserves were given for conservation. a value of 1, with diminishingvalues toward0 for IHAs Altitudinal ranges. Cuestaet al. (2003) found that with increasingarea inside the protectedareas. Andean bears use pairamos at high elevations and montane forest at lower elevations more intensely than Qualitative assessment of other vegetation types throughoutthe year. They sug- conservation importance gested this may be due to a greaterconcentration and These methods provided a quantitativeand objective year-roundavailability of food items (such as palmetto approachto delineate and rank IHAs. However, strictly trees [Arecaceae family] in the cloud forest and quantitative criteria cannot capture all important as- terrestrialbromeliads [Puya spp. and Gregia spp.] in pects of potentialconservation areas. Therefore, we also the pairamo)in these areas. Following that study, we used qualitative considerations: watershed and bio- defined 4 altitudinalranges: 1,800-2,400 m (range 1), diversity protection are 2 important biological and 2,400-3,000 m (range 2), 3,000-3,600 m (range 3), and economical incentives for resource protection in Ecua- 3,600-4,200 m (range 4). We determinedthe percent dor (Josse 2000). covered by each altitudinal range in each IHA and Hydrological resources. One of the most impor- multiplied those percents by 2 for ranges 1 and 4 and tant indices in assessing the conservation potential of 1 for ranges2 and 3. Ranges 1 and4 generallycorrespond largeecosystems in the Andeanmountains is the presence with cloud forest and pairamoareas, respectively, and of intactwatersheds (Olson and Dinerstein1994, Peyton were assigned a greater weight to emphasize their 1999). The relativelyundisturbed, high-elevation water- and sur- importance.These elevation belts have been identified sheds in the Cayambe-CocaEcological Reserve as the best bear habitatwithin the range of the species roundingareas provide one of the majorwater sources in define the (Suarez 1985, Yerena and Torres 1994, Peyton 1999, Ecuador (Paredeset al. 1999). Therefore,to for water re- Troyaet al. 2004). potential importanceof IHAs protecting Influence of disturbed areas. Natural habitats sources, we delineated watersheds in the study area used the DEM for surrounded by anthropogenic influences are more (WatershedTool, TNT Mips GIS). We of flow direc- likely to be exposed to edge effects and isolation, the areato derive hydrologicalparameters on which watersheds affecting the quality of habitat and ultimately leading tion and flow accumulationbased and total to loss of biodiversity due to local extinctions were delineated.We then calculatedthe number with each of the IHAs. (Diamond 1986, Wilcove 1987). The degree of areaof watershedsthat intersected Effective conser- connectivity of remaining habitat patches depends on Number of vegetation types. conservation of functional the pattern and distributionof human disturbances.A vation planning requires units that access to high degree of internal fragmentationoften leads to landscape guarantee adequate bears the The intensity greater human-caused mortality because of more resources by throughout year. different interactions (such as crop and cattle predation by with which Andean bears use vegetation types Cuesta bears and their subsequent killing by farmers). We changes seasonally (Peyton 1980, Suarez 1985, we used the land- ranked the potential for human impacts by calculating et al. 2003). Therefore, digital data to determinethe extent and numberof vege- the percent of each IHA within 6 km from human cover indicators of disturbancezones. This distance is an approximation, tation types included in each IHA as We examined based on own observations in the field, of how far vertical connectivity and beta diversity. the and extent of 3 vegetation types people in the Oyacachi village travel to check on their pattern primary in the area: cattle. Areas completely inside this zone were given used by Andean bears study upper montane cloud and a value of 1, indicating the greatest need for con- montane evergreen forest, forest, Ecuador, servation or management measures. This value di- humid paramo (EcoCiencia, Quito, unpub- minished proportionally with increasing amount of lished report). Ursus 16(2):222-233 (2005) "

DELINEATINGANDEAN BEAR CONSERVATION AREAS * Peralvo et al. 227

Legend Humansettlement ^/ Mainroad A Protectedarea limit MahalanobisDistance Classes cClass 1 EClass 2 class 3 Class 4 Class 5

o 10 2,0 km

Legend c) VA////Impoatant HabitatAreas Land-covertypes _I Highevergreen uppermontane forest I Lowermontane evergreen forest I Montanecloud forest I Humidpiramo Water _II Disturbedareas N

0 10 20 km

Fig. 2. Delineation of important habitat areas (IHAs)for Andean bears in the northeastern Ecuadorian Andes. (a) original model of Andean bear habitat suitability (30-m x 30-m grid cells), (b) generalized model of suitability of Andean bear habitat (1,650-m x 1,650-m grid cells), (c) IHAsdefined as clusters of >10 cells with class 1 values based on the generalized model. The areas are superimposed to the main land-cover types in the study area.

Results tween the slopes of the Cayambevolcano and the upper Habitat suitability watershedof the OyacachiRiver (areaA in Fig. 2a), the Areas with high habitat suitability (class 1 areas) paramo in the southern portion of the Cayambe-Coca covered 27.3% of the study area and were concentrated Ecological Reserve (area B), the paramo surrounding in zones with a predominanceof paramoor cloud forests Lake Puruhanta(area C), the mosaic of upper montane (Table 3, Fig. 2a). Contiguousclass 1 areasoccurred be- forest and paramosin the northernportion of the study

Ursus 16(2):222-233 (2005) 228 DELINEATINGANDEAN BEAR CONSERVATION AREAS * Peralvo et al.

Table 3. Mahalanobis distance classes used in this habitat classes to define IHAs. We identified 12 study as an index to suitability of Andean bear contiguous zones consisting of at least 10 grid units Andes. habitat within the northeastern Ecuadorian to habitatclass 1 The total extent of Mahalanobis distance classes were defined based belonging (Fig. 2c). these areas was km2, 30% of the on a 2000-01 field study in the Oyacachi River basin 1,820 representing (Cuesta et al. 2003). study area. Individually,the area of the 12 IHAs ranged from 30.2 to 397.7 km2 (Table IHAs 3, 6, 7, and 9 Mahalanobis 4). Proportion were small km2) and distance (D2) Area of study relatively (<44 together repre- class Value range (km2) area (%) sented only 2.4% of the study area.IHAs 1, 2, 5, 11, and I 12 between 76 and 161 km2 and covered Class 1 0 < D2 < 7.2 1,650 27.3 were together Class 2 7.2 < D2 < 10.1 1,133 18.7 9% of the study area. Finally, the 3 largest IHAs (4, 8, Class 3 10.1 < D2 < 23.7 1,338 22.1 and 10) had extents greater than 340 km2 and 4 23.7 < D2 < 200 516 8.6 Class 18.6% of the study area (Table 4, Fig. 2c). Class 5 D2 > 200 1,411 23.3 represented Total 6,048 100.0 Ranking of IHAs The scores assigned to the IHAs based on the area (areaD), and in heterogeneousareas of cloud forest altitudinalcoverage criterionranged from 1.35 to 1.90 scores were in the lower portion of the study area (area E). Class 2 (Table 4). Areas that received the 3 highest those areas were in and 3 areascomprised 40.8% of the study area,and their 7, 12, and 8 because mostly 4 which was spatialconfiguration was distinctive.Class 3 areas were elevation range (3,600-4,200 m), weighted more abundant(22.1%) across the landscape, but their more (Table 4). In contrast,IHAs 2, 4, and 10 received 60% of their extents distributionwas heterogeneous, forming a background low scores because approximately received lower for class 1 and 2 areas (Table 3, Fig. 2a). Class 2 areas were within altitudinal ranges that most IHAs had substantial were slightly less common (18.7%) and mainly defined weights (Table 4). Overall, connection zones between class 1 areas. Class 4 and 5 portionsof their areaswithin the 6-km disturbancezone. score for areas covered 31.9% of the study area; the distribution IHA 7 received a high (0.95) anthropogenic influencebecause 94.8% of its area was inside the zone of these areasgenerally resembled the spatialpatterns of influence. we observed variation the road network and agriculturalland uses (Table 3, of human Finally, high of scores for the conservationstatus criterion. IHA 1 was Fig. 2a). located completely outside protected areas and thus and 11 Delineating important habitat areas received a score of 1. Most of IHAs 6, 9, 10, Reserve in The generalized model was based on 2.72-km2 grid were within the Cayambe-CocaEcological disturbances cells (Fig. 2b). We used the spatial patterns of the 5 a zone isolated from major anthropogenic

Table 4. Important habitat areas (IHAs)for Andean bears in the northeastern Ecuadorian Andes and scores used to rank their conservation value. IHAs were defined as areas with >10 contiguous grid cells with Mahalanobis distance values in class 1 based on the generalized model of suitability of Andean bear habitat (1,650-m resolution). Rankingscore Portion of Altitudinal Influence of Conservation Sum of IHA Area (km2) study area (%) coverage disturbed areas status scores 1 101.0 1.67 1.68 0.83 1.00 3.51 84.4 1.4 1.35 0.86 0.91 3.12 2 2.41 3 35.5 0.59 1.57 0.64 0.20 2.48 386.6 6.39 1.36 0.40 0.72 4 2.84 5 161.0 2.66 1.62 0.69 0.53 43.6 0.72 1.68 0.12 0.00 1.80 6 3.09 38.1 0.63 1.90 0.95 0.24 7 2.62 340.3 5.63 1.87 0.49 0.26 8 2.36 9 30.2 0.5 1.63 0.68 0.05 6.58 1.38 0.66 0.00 2.04 10 397.7 2.36 11 125.3 2.07 1.82 0.54 0.00 12 76.2 1.26 1.89 0.45 0.28 2.62 Total 1,820.0 30.1

Ursus 16(2):222-233 (2005) DELINEATINGANDEAN BEAR CONSERVATIONAREAS * Peralvo et al. 229

Table 5. Number and surface area of watersheds within their boundaries (Table 6, Fig. 2c). Similarly, associated with 12 important habitat areas (IHAs) >90% of areas8 and 12 were in paramovegetation. Areas for Andean bears in the northeastern Ecuadorian 4 and 10 wereprimarily associated with cloud forests.The Andes identified in this study. remainingIHAs had varyingpercents of pairamoand high IHA Number of watersheds Area (km2) evergreenupper montane forests; the percentof paramoin 1 41 435 those areas was consistently greaterthan the percent of 2 41 339 high evergreenupper montane forests (Table 6). 3 19 235 4 161 1,099 5 56 524 6 22 313 Discussion 7 17 153 8 109 850 The establishmentof naturereserves often is based on 9 16 210 general biodiversity patterns. However, those nature 10 138 1,178 reserves do not always incorporatethe full ecological 11 54 357 of and 12 62 172 requirements large carnivores, many protected areas would not viable of these Total 736 5,865 support populations species in the long term (Peyton 1999, Suarez 1999). Ourapproach was designed to addressthose concernsby (Table 4, Fig. 2c). Consequently,those areas obtained identifying areas that may complement the current lower scores or 0 for the conservationstatus criterion. system of protected areas and provide long-term After summing scores correspondingto the 3 criteria, protection for viable populations of Andean bears on IHAs 1, 2, and 7 received the highest scores (3.51, 3.12, a regionalscale. Therefore,we focused on the role of the and 3.09, respectively; Table 4). Most of the 12 IHAs Andean bearnot only as an umbrellaspecies, but also as had total scores distributedin a relatively small range an indicator species of the integrity of paramo and between 2.04 to 2.84 (areas 3, 4, 5, 8, 9, 10, 11, 12). montane forest landscapesin the NorthernAndes. IHA 6 received the lowest score (1.80; Table 4). Digital map layers and GIS allowed us to predict suitability of Andean bear habitat on a regional scale Hydrological resources and vegetation types using field-based data and multivariate statistical A total of 1,560 watersheds were derived from the techniques. The model results should be interpreted DEM within the study area. Of those, 736 were at least within the context of the environmental and socio- partially associated with one or more of the 12 IHAs. economic conditions of the area where the field data Total area covered by these watershedswas 5,865 km2, were collected (Oyacachi River basin; Cuesta et al. representing97% of the study area (Table 5). 2003). Extrapolationof the model to a larger area was The majority of IHAs were associated with 2 conducted on the assumption that observed habitat vegetationtypes. Areas 7 and 9 includedmostly paramo relationshipsof Andean bears were constantthrough the

Table 6. Vegetation types associated with the 12 important habitat areas (IHAs) identified in this study for Andean bears in the northeastern Ecuadorian Andes.

Humid paramo High evergreen upper montane forest Montane cloud forest Numberofvegetation Numberof vegetation IHA Area (km2) % Area (km2) % Area (km2) % types representeda 1 84.8 84.0 15.9 15.8 2 2 56.4 66.8 28.0 33.1 2 3 26.2 73.9 9.2 25.9 2 4 0.3 0.1 21.0 5.6 353.8 93.8 2 5 127.5 79.2 33.2 20.6 2 6 35.1 80.6 8.4 19.3 2 7 37.8 99.3 0.3 0.7 1 8 314.8 92.5 23.9 7.0 2 9 29.9 98.9 0.2 0.7 1 10 0.5 0.1 10.6 2.7 386.2 97.1 2 11 99.6 79.5 24.9 19.9 0.5 0.4 2 12 67.9 90.4 7.2 9.6 2 aThe minimumsurface consideredfor a vegetationtype to be representedinside a IHAwas 2.72 km2.

Ursus 16(2):222-233 (2005) 230 DELINEATINGANDEAN BEAR CONSERVATION AREAS * Peralvo et al.

region. At the scale of analysis of our study this assump- effects of roads on habitatconnectivity may depend on tion seems valid, because environmental conditions human use of those roads. For example, the road in the study area were similar to those in the Oyacachi between the Cayambe-CocaEcological Reserve and the study. However, effective on-the-groundconservation AntisanaEcological Reserve (Fig. 1) is associated with strategiesrequire consideration of differentsocio-economic an area of dynamic land-use changes (conversion of conditions throughoutthe study area. forest and paramosto agriculturaluses) and also coin- We observed 2 general trends in the distributionof cides with the routeof 2 oil pipelines. Consequently,this suitable habitatwithin the study area. First, most areas road likely is an effective barrierto bear movements belonging to the highest suitabilityclass (class 1) were between the 2 ecological reserves (Figs. 2b and 2c). In associatedwith largeand contiguous tracts of paramoand contrast,the road from Oyacachi to Papallacta(Figs. 1, cloud forest present at the highest and lowest elevation 2b, 2c) may be less of a barrierto bear movements ranges, respectively (Fig. 2a). The distributionof areas because this road has regulatedand restricteduse and it correspondingto lower suitabilitycategories resembles is not associated with human colonization of new areas a backgroundmatrix (class 3) in which class 2 areascon- within the Cayambe-CocaEcological Reserve. nect patches of the most suitablehabitat (Fig. 2a). Thus, one importantconservation strategy may be to protect high-qualitypairamo and cloud foresthabitats while main- Area ranking tainingconnectivity between these elevationalzones. The level of threat to the integrity of ecosystems in A second pattern that we observed is that habitat northernEcuador is not homogeneousand, consequently, suitabilitystrongly corresponded to distributionof roads there is a need to protect those areas that are most anddisturbed areas, delineating unsuitable habitat (class 5 vulnerablebefore they are degraded(Stoms 2000). Also, area;Fig. 2a). However,some class 1 areaswere adjacent in practice only a portion of relatively undisturbed to roads, producing a sharp boundarybetween suitable habitats can be managed for conservation in the long andnon-suitable habitats (Fig. 2a). Because open paramo term,which emphasizesthe need to identifyconservation habitatsprovide less protectionfor wildlife than cloud areasthat maximize biodiversity gains (Woodhouseet al. forests, the presenceof roads therehas more potentialto 2000). In addition to the need to complement existing increase hunting pressure.In addition, the existence of protectedareas, a rankingto focus conservationefforts is potentially high-quality bear habitat near agricultural needed. Because conservationplanning involves more areas and pastures may result in crop or livestock thanjust biologicalfactors, we also integratedcriteria with depredation. Thus, conservation strategies alleviating measures of altitudinalcoverage, human influence, and these conflictsshould consider the spatialconfiguration of level of protection.Our ranking based on altitudinalrange Andean bear habitat.The existence of areas with high was intended to emphasize the conservation value of habitat suitability close to areas of intense human use paramo(3,600-4,200 m) and cloud forest (1,800-2,400 suggests the importanceof bear-humanconflict manage- m) areas (Cuesta et al. 2003). However, the elevation ment as a criticalAndean bear conservationstrategy. criterionhad relatively little influence on final ranking The distributionof Andean bear habitatin the study because some areas alreadywere protected(area 10) or areacorroborates the patternsdescribed for Andeanbear were subject to relatively low human pressure (area 4; distributionin the northernAndes. At a regional scale, Table 4). In terms of the level of threatdue to human the roads and associated areas of human use reduce activities, the 2 extremes of human influence are rep- habitat connectivity (Kattan et al. 2004). Agricultural resentedby areas6 and7 (Fig. 2c). Area6 only has 11% of areas near the Colombian border to the north and its surfacewithin the 6-km bufferarea of humaninfluence northeast,fragmented inter-Andean valley to the west, becauseit is locatedwithin the core of the Cayambe-Coca and disturbed areas along the road to Amazonian Ecological Reserve.In contrast,area 7 is on the marginof lowlands to the south and southeast of the study area the reserveand borderedby agriculturalareas (Fig. 2c). define the macro configurationof bear habitat,creating The conservationstatus criterionallowed us to rank large tracts of suitable habitat that are mostly discon- areas not in the currentsystem of protectedareas. Only nected (Fig. 2a). areas 1, 2, and 4 had substantial portions outside Roads may be an importantbarrier to bearmovements protectedareas (Fig. 2c). Although the remainingareas and population connectivity. There is evidence that mostly were located in the Cayambe-CocaEcological bears avoid roads, and that roads pose a mortalitythreat Reserve or the Antisana Ecological Reserve, their (Peyton 1999, Rodriguez et al. 2003). However, the delineationlikely is importantfrom a parkmanagement

Ursus 16(2):222-233 (2005) DELINEATINGANDEAN BEAR CONSERVATIONAREAS * Peralvo et al. 231

perspective.For example, areas 3, 5, 7, 8, 9, and 12 all importance to the conservation planning process for contain potentially importanthabitat for Andean bears the preservation of hydrological resources, but also and representcontiguous areas with proportionsinside facilitates delineation of conservationunits in the field an existing reserve, but are subject to a high level of and thereforeits protectionfrom externalpressures and human pressure extending up from the inter-Andean implementation of management actions (Peres and valley (Fig. 2c). These areas could be used to design Terborgh1995). In addition, the preservationof water- comprehensivemanagement strategies aimed at mitigat- sheds would explicitly promote the achievement of ing human impacts on reserves. vertical connectivity between high- and low-elevation The final rankingof IHAs showed that areas 1 and 2 habitats while facilitating connections among IHAs. were most importantfor additionalprotection of Andean Finally, the integration of hydrological criteria at bear habitatin the study area (Table 4; Fig. 2c). Those a landscape scale allows special managementof areas areas are outside the Cayambe-CocaEcological Reserve importantnot only for Andean bears, but also other bi- and cover importantpairamo areas north of the reserve. otic communities associated with aquatic environments Degradationof those areas is likely in the near future (headwaters,wetlands), such as neotropical amphibian because of their proximity to highly dynamic agricul- communities(e.g. Atelopus spp.). turalborders that threatenthe remnantareas of montane Although IHAs for bear conservationrepresent 30% cloud forest nearthe disturbedinter-Andean valley. Area of the study area,the associatedwatersheds influenced by 4 representedthe highest-rankedIHA covered mainly by managementof these areas cover 97% of the study area cloud forest (rank 7; Tables 4 and 6; Fig. 2c). Higher (Table 5). Also, most of IHAs are located in paramo levels of human influence in the highlandsmay explain zones, which arethe headwaterareas for manywatersheds why pairamo-dominatedareas generally were ranked that drain towardthe inter-Andeanregion. Because any higher than cloud forest areas. Despite the apparent managementstrategy will affect the quantityand quality dichotomy in conservation priorities between paramo of waterresources downstream, the distributionof IHAs and cloud forest areas, conservation strategies for the in such highlandareas could representan opportunityto Andean bear within the study area should remainbased promotetheir conservation and sustainablemanagement on maintaininglandscape diversity to ensure seasonal with the goal of preservingwater resources in the long access to resources in both paramo and forest areas term. This is particularlyimportant in the highlands, (Cuesta et al. 2003, Kattanet al. 2004). where land tenureissues would restrictthe establishment of conservationareas for Andean bears alone. The link between bear habitat and research conservationand managementof Management waterresources may provide a useful frameworkfor the implications developmentof effective conservationstrategies because We identified several issues relevant to conservation it involves many differentstakeholders, from the central of Andean bearsin northeasternEcuador and elsewhere. governmentto local people interestedin the conservation For example, the delineation and ranking of IHAs of these resources. indicatedthe of importance maintainingconnectivity of The conservation priorities identified in our study bearhabitat both within and between differentaltitudinal should be considered an initial step in defining con- ranges (such as between areas 6 and 8; Fig. 2c). The servation areas for Andean bears in norther Ecuador. habitat areas we delineated important were associated This is a hierarchicalprocess, and the next step is to either with and pairamo high montaneevergreen forest at furtherevaluate informationregarding different threats elevations or with cloud forests at high lower elevations and conservationopportunities in each priorityarea. The (Table 4, Fig. 2c). Given the ecological requirementsof definition and analysis of these local criteriashould be the Andean bear, effective conservation strategies rooted in a participatoryprocess involving local people, should conservation areas that emphasize promote government officials, and conservation organizations. between and forest connectivity paramo ecosystems. In Furthermore,Andean ecosystems are highly threatened this the IHAs we context, delineatedcould be considered due to an expanding agricultural frontier and the focal areas in a core-corridorstrategy for conservation development of new infrastructuresuch as roads and of bear habitatat finer scales. dams. Therefore, future research should specifically For our we study area, suggest that watersheds examine landscapepermeability between IHAs, similar useful conservation represent planning units. The use to studies on grizzly bears (Ursus arctos) in North of watershedsas planningunits not only adds functional America (Singleton et al. 2004).

Ursus 16(2):222-233 (2005) 232 DELINEATINGANDEAN BEAR CONSERVATION AREAS * Peralvo et al.

Acknowledgments OLSON, D.M., AND E. DINERSTEIN.1994. Evaluacion de Potenciales de Conservacion Grados de Amenaza Funding for this study was provided by the De- y para las de America Latina el Caribe: Un partment for International Development program of the Ecorregiones y Analisis Utilizando Ecologia de Paisaje. Nota de Divulga- British Embassy in Quito, The Nature Conservancy, ci6n 10. Banco Mundial, Departamento Tecnico para United States Agency for International Development, America Latina, Division Ambiental.(In Spanish.) and the Wildlife Conservation We extend Society. PAREDES,R., J. CALVOPINA,AND A. BUITRON. 1999. Plan de a thanks to P. H. C. special Aigaje, Aigaje, Aigaje, Manejo de la ReservaEcol6gica Cayambe-Cocay su zona L. L. V. and all the Parion, Aigaje, Aigaje, people de amortiguamiento.Instituto Ecuatoriano Forestal de Areas from Oyacachi. We thank L. Suarez, R. Hofstede, Naturalesy Vida Silvestre, Quito, Ecuador.(In Spanish.) C. Valle, S. Benitez, J. Jorgenson, G. Medina, I. PERES,C.A., AND J.W. TERBORGH.1995. Amazonian nature Goldstein, and R. Cedeio for their continuing support reserves: an analysis of the defensibility status of existing of this research conservationunits and design criteriafor the future. Con- servationBiology 9:34-46. PEYTON,B. 1980. Ecology, distributionand food habits of Literaturecited Andean bears, Tremarctos ornatus, in Peru. Journal of 61:639-652. BOITANI,L.P., P. CIUCCI,F. CORSI,AND E. DUPRE.1999. Mammalogy E. D.I. J. AND J.E. Potential and corridorsfor brown bears in the Eastern , YERENA, RUMIZ, JORGENSON, range OREJUELA.1998. Status of wild Andean bears and Ursus 11:123-130. policies Alps, Italy. for their Ursus 10:87-100. CLARK,J.D., J. DUNN,AND K. SMITH.1993. A multivariate management. . 1999. Spectacledbear conservation action plan. Pages model of female black bear habitat use for a geographic 157-198 in C. Servheen, S. Herrero, and B. Peyton, informationsystem. Journalof Wildlife Management57: Bears. Status and conservation action 519-526. compilers. survey Survival Commission Bear and Polar CUESTA,F., M. PERALVO,AND D. SANCHEZ.2001. Metodos para plan. IUCN/Species Bear Specialist Groups. IUCN, Gland, Switzerland and investigar la disponibilidaddel hdbitatdel oso andino: el UK. caso de la cuenca del rno Oyacachi, Ecuador. Serie Cambridge, PITMAN,N.C., P.M. JORGENSEN,R. WILLIAMS,S. LEON-YANEZ, Biorreserva del C6ndor 1. EcoCiencia y Proyecto Bio- AND R. VALENCIA.2002. Extinction-rate estimates for rreservadel C6ndor, Quito, Ecuador.(In Spanish.) a modern flora. Conservation Biology 16: , AND F.T. VAN MANEN.2003. Andean bear neotropical 1427-1431. habitat use in the Oyacachi River basin, Ecuador. Ursus AND K.A. CHAPMAN.2001. 14:198-209. POIANI K., M.D. MERRILL, conservation areas in a fragmented DIAMOND,J. 1986. The design of a naturereserve system for Identifying priority Minnesota based on the umbrella concept IndonesianNew Guinea.Pages 485-503 in M. Soule, editor. landscape species and selection of of natural vegetation. Conservationbiology-the science of scarcityand diversity. large patches Conservation 15:513-522. Sinauerand Associates, Sunderland,Massachusetts, USA. Biology RODRIGUEZ,D., F. CUESTA,I. GOLDSTEIN,A.E. BRACHO,L.G. HILTON-TAYLOR,C. 2000. The 2000 IUCN red list of ANDO.L. HERNANDEZ,EDITORS. 2003. Estrategia threatened species. IUCN-World Conservation Union, NARANJO, la conservacion del oso andino en los Gland, Switzerland. Ecorregional para Andes del Norte. World Wildlife Fund Colombia, Funda- JOssE, C., EDITOR.2000. La Biodiversidad del Ecuador- Conservation Informe2000. Ministeriodel Ambiente,EcoCiencia, Union cion Wii, EcoCiencia, and Wildlife Society, Colombia. Internacional para la Conservacion de la Naturaleza (In Spanish.) SANDERSON K.H. REDFORD, C.B. CHETKxWIZ, R.A. (UICN-Sur),Quito, Ecuador.(In Spanish.) E.W., MEDELLIN,A.R. RABINOwrrz,J. ROBINSON,AND A. TABER. KATTAN,G., O.L. HERNANDEZ,I. GOLDSTEIN,V. ROJAS, O. to save a the as a model. MURILLO,C. G6MEZ, H. RESTREPO, AND F. CUESTA.2004. 2002. Planning species: jaguar Conservation 16:58-72. Range fragmentationin the spectacled bear Tremarctos Biology AND J. CHAMBERLIN.1999. Areas ornatus in the northernAndes. Oryx 38:155-163. SIERRA,R., F. CAMPOS, la de la biodiversidaden el KNICK, S.T., AND J.T. ROTENBERRY.1998. Limitations to prioritariaspara conservaci6n Ecuadorcontinental: un studio basado en la diversidadde mappinghabitat use areasin changinglandscapes using the Ministerio del Mahalanobis distance statistic. Journal of Agricultural, ecosistemas y su omitofauna. EcoCiencia, INEFAN/GEF-BIRF Wildlife Con- Biological, and EnvironmentalStatistics 3:311-322. Ambiente, Proyecto y Ecuador. MCNAB,H.W. 1989. Terrainshape index: quantifyingeffect of servationSociety, Quito, (In Spanish.) bears-A conservation minor landformson tree height. Forest Science 35:91-104. SIMBERLOFF,D. 1999. Biodiversity and MYERS,N., R.A. MTrERMEIER,C.G. MITERMEIER,G.A. DA paradigmshift. Ursus 11:21-28. AND J.F. LEHMKUHL.2004. FONSECA,AND J. KENT.2000. Biodiversity hotspots for con- SINGLETON,P.H., W.L. GAINES, for bear movements in servationpriorities. Nature 403:853-858. Landscape permeability grizzly

Ursus 16(2):222-233 (2005) DELINEATINGANDEAN BEAR CONSERVATION AREAS * Peralvo et al. 233

Washington and southwestern British Columbia. Ursus WOODHOUSE,S., A. LOVETT, P. DOLMAN, AND R. FULLER. 2000. 15:90-103. Using a GIS to select priority areas for conservation. STOMS,D.M. 2000. GAP management status and regional Computers,Environment and Urban systems 24:79-93. indicators of threats to biodiversity. Landscape Ecology WORLDWILDLIFE FUND. 2001. Complejo Ecoregional de los 15:21-33. Andes del Norte: Hacia una visi6n de su biodiversidad. SUAREZ,L. 1985. Habitos alimenticios y distribuci6nestacio- World Wildlife Fund, Fundaci6nNatura, FUDENA, Cali, nal del oso de anteojos (Tremarctosornatus) en el paramo Colombia. surorientaldel Volcan Antisana,Ecuador. Thesis, Pontificia YERENA, E., AND D. TORRES. 1994. Andean bear conservation Universidad Catolica del Ecuador, Quito, Ecuador. (In and dispersalcorridors in Venezuela. InternationalConfer- Spanish.) ence on Bear Researchand Management9(1):169-172. . 1999. Status and managementof the Andean bear in . 1998. Protectedareas for the Andean bear in South Ecuador. Pages 179-182 in C. Servheen, S. Herrero,and America. Ursus 10:101-106. B. Peyton, compilers. Bears. Status survey and conserva- YOUNG,K.R. 1998. Deforestation in landscapes with humid tion action plan. IUCN/Species Survival CommissionBear forests in the Central Andes. Pages 75-99 in K.S. and Polar Bear Specialist Groups. IUCN, Gland, Zimmererand K.R. Young, editors. Nature's geography: Switzerlandand Cambridge,UK. new lessons for conservationin developing countries.The TROYA,V., F. CUESTA,AND M. PERALVO.2004. Food habits of Universityof Wisconsin Press, Madison, Wisconsin, USA. Andeanbears in the OyacachiRiver Basin, Ecuador.Ursus ZOMER,R., S. USTIN, AND J. IVES.2002. Using satellite remote 15:59-62. sensing for DEM extraction in complex mountainous VALENCIA,R., C. CERON, W. PALACIOS, AND R. SIERRA.1999. terrain:landscape analysis of the Makalu Barun National Los sistemas de clasificaci6n de la vegetaci6n propuestos Park of eastern Nepal. InternationalJournal of Remote parael Ecuador.Pages 19-28 in R. Sierra,editor. Propuesta Sensing 23:125-143. preliminarde un sistema de clasificaci6n de la vegetaci6n para el Ecuadorcontinental. Proyecto INEFAN/GEF-BIRF y EcoCiencia. Quito, Ecuador.(In Spanish.) Received: 2 August 2004 WILCOVE,D. 1987. From fragmentationto extinction. Natural Accepted: 9 May 2005 Areas 7:23-29. Associate Editor: P. McLoughlin

Ursus 16(2):222-233 (2005)