Environmental factors influencing guanaco distributionand abundance in central Patagonia Argentina

Julieta PedranaABF Alejandro TravainiA Juan Ignacio ZanoacutenC Sonia Cristina ZapataAAlejandro RodriacuteguezD and Javier Bustamante E

ACentro de Investigacioacuten Puerto Deseado Universidad Nacional de la Patagonia Austral CONICETAvenida Prefectura Naval SN 9050 Puerto Deseado Santa Cruz Argentina

BPresent address Recursos Naturales y Gestioacuten Ambiental Instituto Nacional de Tecnologiacutea AgropecuariaEEA Balcarce CC 276 7620 Balcarce Argentina

CInstituto de Ciencias de la Tierra y Ambientales de La Pampa (INCITAP-CONICET) Universidad Nacionalde La Pampa Avenida Uruguay 151 6300 Santa Rosa Argentina

DDepartment of Conservation Biology Estacioacuten Bioloacutegica de Dontildeana CSIC CAmeacuterico Vespucio 2641092 Sevilla Espantildea

EDepartment of Wetland Ecology amp Remote Sensing and GIS lab (LAST-EBD) Estacioacuten Bioloacutegica de DontildeanaCSIC CAmeacuterico Vespucio 26 41092 Sevilla Espantildea

FCorresponding author Email pedranajulietaintagobar

AbstractContextThe guanaco is the largest wild herbivore inhabiting the Patagonian steppes Since the end of the 19th Century

it has suffered a progressive decline in numbers owing to poaching and unregulated hunting because of on an assumedcompetition with sheep Unfortunately there has never been a management program for guanaco populations inArgentine Patagonia Consequently the guanaco is still considered a pest species by ranchers and has never beenconsidered profitable in the range management model implemented in Patagonia

Aims The present article updates the distribution limits of guanaco and estimate its abundance across Chubut a largeprovince of Patagonia Argentina The relative effects of several environmental and anthropogenic factors on guanacodistribution are also assessed

MethodsRoad surveys (7010 km) and species distribution modelling were used to build a habitat suitability model anda distribution map A distance sampling method was used to estimate guanaco population densities and size The surveyeffort required to monitor population trends in this region was also calculated

Key results According to the best habitat suitability model guanaco distribution decreased with altitude and primaryproductivity as measured by Normalised Difference Vegetation Index (NDVI) and increased with the distance to thenearest urban centre and oil field Guanaco distribution showed a clear geographical pattern in Chubut with low tomediumoccurrence probability towards the west and higher values towards the east Guanaco population size was estimatedas 657 304 individuals (95 CI 457 437 to 944 059) with a mean density of 297 guanacos kmndash2 Finally throughsimulations of guanaco monitoring it was estimated that an annual survey effort of 10 to thirty 30-km road transects isneeded to detect with confidence a significant population decrease or increase over the next 6 or 10 years

Conclusions The habitat suitability map presented herein highlights areas with high guanaco densities in Chubutwhere it would be possible to identify ranches suitable for performing profitable herding and shearing experiences

Implications The maps of guanaco distribution and density as well as the survey effort required to monitorpopulation trends may be used to inform decisions concerning the sustainable use of this species

Additional keywords habitat models Lama guanicoe monitoring predictive cartography population density

Received 7 May 2018 accepted 5 October 2018 published online 16 January 2019

Introduction

Patagonian pastoral systems were uniquely based on sheepranching from the very initial land occupation by Europeansettlers (1880ndash1930 Soriano and Paruelo 1990) This was

followed by the increase and stabilisation of sheep stocksand capital accumulation (1930ndash80) up to full landoccupation Consequently the carrying capacity of grasslandsprogressively began to decrease while steppe degradation

CSIRO PUBLISHING

Wildlife Researchhttpsdoiorg101071WR18085

Journal compilation CSIRO 2019 wwwpublishcsiroaujournalswr

increased (Golluscio et al 1998) As a result there has beena rural depopulation process that started in the 1980s moreevident nowadays in the southern provinces of Argentina (egSanta Cruz and Chubut Borrelli and Cibils 2005) Throughoutthese 135 years wildlife has never been considered asan economically profitable option (eg wool meat and eventourism) by the government of Argentina nor a complementaryincome to domestic livestock production by ranchers (Caroet al 2017) Now national or regional agricultural guidelinesshould acknowledge that arid and semiarid Patagonianrangelands cannot be sustainably developed using onlyconventional meat and wool production systems Thus animalproduction cannot generally be increased or even sustainedwithout further degrading the natural capital ie the grasslandsThe consequences of overgrazing on the productivity andprofitability of rangelands have been found to be severe(Ares et al 1990) In contrast wildlife might be potentiallyamenable to community-based sustainable managementwhich is an increasingly desirable exploitation model (Rothand Merz 1996) This would allow diversification of thelocal economy that with the addition of services andmarketing would raise incomes without over-utilising naturalcapital This scenario gives wildlife use a comparativeadvantage in arid rangelands and allows for sustainableeconomic development (Child 1988) It also provides theopportunity to test the ecological advantages of multi-speciessystems over traditional livestock production systems basedon a single species In South Africa for example there arenow 10 000 to 14 000 private ranchers that promote wildlifeenterprises alone or in combination with domestic livestock(Child et al 2012) This is an example of an importantconservation success that has been accompanied by theimprovement of social wellbeing through economic growthand employment creation

The guanaco (Lama guanicoe Muumlller 1776) is the largestherbivore of the Patagonian steppe (Redford and Eisenberg1992) and has an extraordinary potential for a sustainable use(Franklin et al 1997) something that could be enhanced ifranchers and others social actors (eg politicians economistsgovernors) were aware of the added value of guanaco productssuch as fibre leather and meat (Lichtenstein and Carmanchahi2012) Until recently guanacos as well as other wildlifespecies (eg the lesser rhea Rhea pennata) have never beenconsidered economically profitable in the range managementmodel implemented in Argentine Patagonia (Von Thuumlngen andLanari 2010) In addition ranchers from Santa Cruz provincecategorise the guanaco as a pest species and have a negativeattitude towards it (Caro et al 2017) Guanaco wool hasa particularly fine fibre that can be obtained through live-shearing of wild individuals (Montes et al 2006 Saccheroet al 2006) using shearing machines that guarantee a residualfibre of appropriate length to avoid excessive heat loss ina cold environment (Gerken 2010) Sheep ranchers fromnorthern Patagonia are gaining expertise at shearing guanacoswhich is currently allowed by Argentinean legislation(Baldi et al 2010) Unfortunately the guanaco is stillconsidered and consequently treated as a pest species byranchers in the southern provinces (Baldi et al 2010Schroeder et al 2014)

Guanacos are distributed in an extensive range of arid andsemiarid habitats from sea level to 4500m from northern Peruto central Chile and across Argentinean and Chilean PatagoniaAlthough this species has been considered a highly adaptablecamelid with a broad distributional range (Franklin 1983) theguanaco has experienced a progressive drop in numbers andalso a parallel reduction of its geographic range (Franklin et al1997) This decline has been attributed to unregulated huntingand poaching (Franklin 1983 Donadio and Burskik 2006)prompted by an assumed competition with sheep for waterand food (Franklin 1983 Pedrana et al 2010) Althoughthe guanaco is not considered a threatened species at thecontinental level (Baldi et al 2016) some populations couldbe at risk of disappearing (Wheeler 2006) or may be locallyextinct as a result of hunting and habitat loss (Baldi et al 2010)These conservation problems may be relatively local andapparently hardly apply to all of Patagonia (Zanoacuten Martiacutenezet al 2012) In addition most populations are restrictedto low-quality habitats because areas with high primaryproductivity are occupied by sheep (Pedrana et al 2010)

Any activity based on the guanaco that is economicallyprofitable will generate a more positive perception of guanacosby sheep ranchers (Nabte et al 2013) and less conflict by wayof competition for food and water (Baldi et al 2001) Theunresolved situation between the guanaco and sheep ranchingactivities in Argentine Patagonia has led the Santa CruzProvincial Agricultural Council (CAP) to consider declaringguanacos as a pest and to develop mitigation actions to reduceguanaco populations such as encouraging indiscriminate huntingacross the entire province and to execute culling as part of theirmanagement plan (Schroeder et al 2014)

The humanndashguanaco relationship including commercialuse and conflict in the Patagonian steppes makes it necessaryto update its present distribution and evaluate its abundanceas well as to understand the underlying environmental andanthropogenic drivers on their regional distribution Thusthis information should be essential to generate detailedknowledge for the sustainability of guanaco exploitation Theaims of the present study were (1) assessing the relative effectsof environmental and anthropogenic factors on guanacodistribution in a large region of Patagonia (2) generatingpredictive distribution maps of guanaco at a regional scaleusing habitat suitability models (3) estimating guanacodensity and population size for this region and (4) based onguanaco abundance at the time of the present study evaluatingthe sampling effort needed to monitor guanaco populations undertwo possible scenarios namely its sustainable exploitationand the unlikely but not impossible success in the initiative todeclare the species as potentially detrimental for sheep husbandryWe also discuss and compare the results obtained in the presentstudy with those of similar surveys in the adjacent Santa Cruzprovince (Pedrana et al 2010 Travaini et al 2015)

Materials and methodsStudy areaChubut with a total area of 224 686 km2 represents 6 ofArgentina and is the second largest Patagonian province(42ndash46S 63ndash72W Fig 1) This region has a cool-temperate

B Wildlife Research J Pedrana et al

and dry climate with strong and predominantly west winds andwith a mean annual temperature that usually ranges between 6and 12C (Paruelo et al 1998) Annual precipitation of ~200mm(Paruelo et al 1998) concentrates during winter and is stronglyinfluenced by general circulation patterns the Pacific air massesand the Andes topographic barrier parallel to the Pacific coastresulting in a strong westndasheast gradient (Barros et al 1979)The dominant vegetation type is shrub steppe composed ofmedium height to dwarf shrubs (Nassauvia glomerulosaN ulicina Chuquiraga aurea C avellanedae) with variabletussock grass cover (Stipa humilis S ibari) Extensive sheepgrazing is the predominant agricultural practice in PatagoniaMerino sheep of varying quality are kept in the field throughoutthe year herded accordingly to a summerndashwinter rotationscheme and fed with a varying seasonal forage dependingon landscape and altitude (Ares et al 1990) Ecosystemdegradation has been attributed to human managementfactors such as overestimation of carrying capacity of therangelands inadequate distribution of animals in extremelylarge heterogeneous paddocks and year-long continuousgrazing (Golluscio et al 1998)

Study speciesGuanacos are sexually monomorphic camelids that aggregatein social groups during the summer breeding seasonAccording to Franklin (1983) and Young and Franklin (2004)

the spatial distribution of populations is influenced by a matingsystem of resource defence polygyny (a territorial systemwherein males compete for access to resources required byfemales) Guanacos can be found in three basic social unitsfamily groups (a male with a group of females and their progenyyounger than one-year old called lsquochulengosrsquo) non-reproductive male groups and solitary males A fourthpotential social unit female groups was only describedduring intensive field studies in Torres del Paine NationalPark Chile (Ortega and Franklin 1995) but this social unitwas not considered in our guanaco censuses because it wasuncommon even in the study reporting it (8 of all socialunits observed Ortega and Franklin 1995) and might includeseveral females with young making female groupsindistinguishable from family groups Pedrana et al (2009)empirically demonstrated that only family groups withchulengos can be unequivocally identified during surveys byvehicle and that the identification of male groups was alsochallenging because these could be mistaken for familygroups without or with undetected chulengos During thepresent study we used four observable social units as a formof stratification to produce separate density estimates breedinggroups (family groups with detected chulengos) non-breedinggroups (groups without chulengos) solitary animals andundetermined (groups that could not be clearly assigned toany of the first three categories Pedrana et al 2009)

60deg W

30deg S

50deg S

Chubut

Santa Cruz

100 km

Survey Tracks November 2006

Flat ndash Low NDVI

N

Flat ndash Medium NDVI

Flat ndash High NDVI

Rugged ndash Low NDVI

Rugged ndash Medium NDVI

Rugged ndash High NDVI

Survey Tracks December 2006

Road network

Fig 1 Study area stratification based on productivity (three Normalised Difference Vegetation Index (NDVI) categories low mediumand high) and topography (two categories flat and rugged terrain) and roads surveyed in central Patagonia Argentina

Guanaco abundance distribution and monitoring Wildlife Research C

Sampling unit selection and field workTo account for the expected regional variation in guanacodensities associated with major environmental variability weused a stratified random design to select the road segments tobe surveyed For this reason we categorised the study area intosix geographical regions based on the combination of twohabitat characteristics that affect guanaco detection anddensity (1) Normalised Difference Vegetation Index (NDVI)and (2) mean slope (Fig 1 Travaini et al 2007) Indeedboth variables influenced guanaco distribution in Santa Cruz(Pedrana et al 2010)

Fieldworkwas conducted during one austral springndashsummerwhich overlaps with guanaco breeding season Using a vectordata of road coverage we randomly selected 110 survey tracks(road segments of 5 to 80 km) that summed to 4440 km oftransects during the first survey (9ndash27 November 2006Fig 1) During our second survey (7ndash17 December 2006) werandomly selected another 54 survey tracks that added up to2570 km (Fig 1) Almost 90 of the survey tracks were dirtand secondary roads with very low traffic density (ie fewerthan 10 vehicles per day)

Because of their body size colouration and behaviourguanacos are easily observable from vehicles driving on roadsand secondary trails (Puig 1995) in the open steppe or shrub-steppe These conditions are excellent for estimating presenceand even abundance by vehicle survey using distance samplingmethodology (Buckland et al 2001) Road surveys were carriedout from a vehicle driving mostly at a speed of 20 kmh1 witha driver and one observer during one breeding season (Fig 1)following recommendations by Pedrana et al (2010) Whena guanaco was detected we stopped the vehicle recorded groupsize and assigned the group to one of our observable socialgroups We measured the distance to the animal or the groupcentre with a laser range finder (Leica LRF 1200 RangemasterGermany) We also recorded the azimuth of the observerrsquostrajectory obtained from the inertial compass of a GPS unit(Garmin GPS mapR 76S USA) and the angle of the animalrelative to our bearing (using a protractor) Distance and azimuthallowed us to obtain the actual positions of guanacos andtheir perpendicular distance to the survey line as required bydistance-sampling methods (Buckland et al 2001) All datawere collected in a personal digital assistant (PDA TungstenT3 USA) synchronised with the GPS which was used to recordthe trajectory and location of the survey track

Habitat suitability maps for guanacoWe selected eight potential predictors that summarised themost relevant environmental and landscape gradients as wellas anthropogenic features needed to assess which factors mightinfluence the regional distribution of guanacos (Table 1 forfurther details see Travaini et al 2007 Pedrana et al 2010)Distribution modelling requires defining units in whichpresence or absence is recorded For this purpose we overlaidthe surveyed tracks on top of a 1-km grid following the spatialresolution of NDVI data Then the original 457 sightingsregistered during field surveys were also overlaid and gridcells with at least one guanaco sighting were considered as apresence cell and all remaining cells traversed by census tracks

were considered as absences As cells with guanaco presence(n= 310) were outnumbered by absence cells (n= 15 606)we used a resampling scheme to obtain a presenceabsencebalanced sample (Travaini et al 2015) We randomly selected310 out of the 15 606 cells with guanaco absences repeatingthis procedure 100 times Each time we generated two datasetsa random sample of 80 of the cells for calibrating theoccurrence models and the remaining 20 for evaluating themodels (hereafter referred as the construction and evaluationdatasets)All the cellswith presencewere used in each repetitionwhile cells with absence were sampled without replacement

To identify which predictors were most likely to affectguanaco occurrence model fitting was done on the constructiondataset with Generalised Additive Models (GAMs Hastieand Tibshirani 1990) using a binomial error and a logit linkPredictors for the models were selected by a backwardndashforwardstepwise procedure (Sakamoto et al 1986) The AkaikeInformation Criterion (AIC) was used to retain a term(Burnham and Anderson 2002) We considered as competingmodels those for which the differences between AIC and theAIC of the best candidate model (the one with the smallestAIC) was D 2 (Burnham and Anderson 2002)

Model evaluation was made using the validation dataset bycomparing predicted and observed values using a threshold-independent metric ie the area under the curve (AUC) ofthe Receiver Operating Characteristic (ROC) plot which wascomputed for each of the 100 fitted models (Murtaugh 1996)The AUC ranges from 0 (when model discrimination isnot better than random) to 1 (perfect discriminatory ability)Predictive models were considered usable if AUC 07

Finally predictions of the resulting distribution models wereused to create habitat suitability maps (HSM) Predictions forthe entire study area were calculated using R version 322 (RDevelopment Core Team 2016) and transformed to probabilitymaps using IDRISI Selva GIS software (Clark Labs WorcesterMA USA) We categorised and mapped the HSM of guanacooccurrence probability in three classes (low lt033 medium033ndash066 and high gt066)

Table 1 Description of the variables used in thepredictive distributionmodels developed for guanaco in central Patagonia Argentina

NVDI Normalised Difference Vegetation Index

Predictor Predictor description

MeanNDVI Mean Normalised Difference Vegetation Indexcalculated by using the VGT-S10 product thatis a 10-day maximum composite value from theVEGETATION sensor on board of Spot-4 satellitefrom April 1999 to March 2005

SeasonMAX Month at which the NDVI reaches its annualmaximum value

CVNDVI Coefficient of variation of NDVIDistancestream Distance (km) to the nearest stream or riverAltitude Mean elevation in meters above sea level of the

focal cell obtained from the Shuttle RadarTopographic Mission Elevation (SRTM)

Slope Terrain slope in degrees in a 1 km pixelfrom the SRTM

Distanceurban Distance (km) to the nearest urban areaDistanceoil Distance (km) to the nearest oil camp

D Wildlife Research J Pedrana et al

Guanaco density estimationLine transect data were analysed using the DISTANCEprogram (version 60 Buckland et al 2001 Thomas et al2010) Estimates of guanaco densities were done by fitting adetection function from the perpendicular distances of guanacosto the survey line (Buckland et al 2001) We pooled distancedata across all transects within each HSM class of predictedguanaco occurrence to estimate the detection functions g(x) foreach social unit in order to attain the required minimum numberof observations (Buckland et al 2001) Although detectabilityvaries among transects the property of lsquopooling robustnessrsquoensures the reliability of the abundance estimates (Bucklandet al 2001 Thomas et al 2010) Pooling also helps to even outminor fluctuations in object density across surveyed strips andto even out chance fluctuations in object distribution (Fewsteret al 2005) The DISTANCE program allows for carefuldata exploration facilitating the identification of any outlyingobservation at extreme distances or sighting clumping andsuggests appropriate levels of truncation and grouping fordetection function fitting We used the smallest value of AICto choose the best competing model (Key function + adjustmentterms) Density estimates were calculated for each social unitand for all guanaco contacts pooled together and for each HSMcategory and for all of Chubut Considering the extent of eachHSM category density estimates were converted into estimatesof population size

Monitoring programWe estimated and simulated the sampling effort in terms ofnumber of 30 km transects of road survey needed to monitorguanaco population in Chubut and also to detect declines frompresent status in guanaco population using MONITOR (version1100 Gibbs and Ene 2010) We modelled the encounter ratewhich was calculated as the number of guanacos detected persampled kilometre over time and to generate detection ratesderived from route-regression analyses (Gibbs and Melvin1997) This simulation procedure is useful for evaluating thetrade-offs between sampling effort logistical constraints andpower to detect trends (Field et al 2005) To estimate powerwe supplied the programwith initial estimates of encounter ratesand its variance derived empirically from our own results Wearbitrarily selected 30 km as the length of sampling units Basedon our experience a 30-km-long transect usually takes abouttwo hours to complete the interval we used to exchangeobservers during field work to guarantee observer attention

Additionally in the study area random 30-km transects canbe selected from the vector data of road coverage more easilythan longer segments To select suitable sampling efforts wefixed the minimum acceptable power at (1-b) = 080 (Di Stefano2001) ie the highest probability of failing to detect a realtrend was 020 Type I error was fixed at 005 To define theeffect size component we selected two alternative scenariosboth highly probable given the guanaco current situationpreviously described The first scenario is under the hypothesisof a sustainable use of guanaco populations to detect early a 40decrease in the next six to 10 years An alternative scenariothe second is if the guanaco is considered as a pest species inorder to early detect a 40 increase in the next six and 10 yearsAs we had sampled each one of the 164 survey tracks oncerandom resampling without replacement of 30-km transectswas used to estimate the mean and the standard deviation ofguanaco encounter rates as a surrogate of spatial variabilityAs an estimate of power variability we estimated the standarddeviation from 20 simulations for each monitoring scenarioand 2000 iterations for each simulation

Results

We surveyed 7010 km almost 76 of available national andprovincial roads in Chubut We recorded 456 guanaco socialgroups which added up to 2893 individuals (Table 2) Almost20 of individuals were registered as belonging to breedinggroups 30 as non-breeding groups and 4 as solitaryindividuals (Table 2) The remaining 47 of individuals wereclassified as undetermined (Table 2) The high percentageof undetermined individuals resulted from our interest inminimising the chances of social group misclassification Weopted for considering family groups without juveniles orwith undetected juveniles as undetermined rather thanmisclassifying them as non-breeding family groups (Pedranaet al 2009) Family groups (n) averaged 8 (sd 322) individualswhere the mean number of females and chulengos per groupwere 5 (sd 268) and 2 (sd 117) respectively (Table 2)Non-breeding and undetermined groups averaged 28 (sd1581) and 6 (sd 268) individuals per group respectively(Table 2)

Habitat suitability maps for guanaco

Themost parsimonious GAMof guanaco presence incorporatedas predictors themeanNDVI altitude the distance to the nearesturban centre and to the closest oil field (Table S1 available

Table 2 Number of guanaco groups and individuals (within brackets) and mean group size recorded during one breeding season(NovemberndashDecember 2006) in the three areas defined by Habitat Suitability Maps (HSM) and for the total area covered in central

Patagonia Argentinasd standard deviation Und Undetermined groups

HSM classes Area Effort (km) Guanaco social groups (individuals) Total Mean group size (sd)(km2) ( transects) Solitary Breeding Non

BreedingUnd Breeding Females

per groupChulengosper group

NonBreeding

Und

Low 76 741 1824 (46) 14 6 (42) 5 (159) 13 (91) 38 (306) 7 (300) 5 (289) 1 (103) 32 (2272) 7 (438)Medium 68 145 3686 (82) 46 26 (216) 16 (444) 120 (710) 208 (1416) 8 (331) 5 (283) 2 (127) 28 (143) 6 (427)High 76 205 1500 (36) 65 35 (294) 10 (259) 100 (553) 210 (1171) 8 (325) 6 (258) 2 (121) 16 (2590) 6 (359)

Total 221 091 7010 (164) 125 67 (552) 31 (862) 233 (1354) 456 (2893) 8 (322) 5 (268) 2 (117) 28 (1581) 6 (268)

Guanaco abundance distribution and monitoring Wildlife Research E

as Supplementary Material to this paper) The probability ofguanaco occurrence decreased with primary productivity andwith mean altitude whereas it increased with the distance tothe nearest urban centre and oil field (Fig 2)

The mean ( se) AUC of the best model ranged from080 (002) to 082 (002) for the evaluation dataset(Table S1) These results indicate that models have anacceptable discrimination ability and are useful for predictingguanaco distribution

The HSM of guanaco occurrence (Fig 3) corresponding tothe predictions of the best model (Table S1) showed that the

probability of guanaco occurrence was positively associatedwith a gradient of increasing aridity from west to east Areaswith low probability of guanaco occurrence predominatedthroughout the western third of Chubut within the sub-Andean and occidental physiographic districts (Soriano 1956)from the transition from forests to short grasslands steppesand scrublands to the centre of the Province (Ares et al1990) From there the probability of guanaco occurrencesteeply increased eastward up to the coastline and theValdeacutes Peninsula throughout the driest lands with the mostxeromorphic floristic composition (Fig 3)

0

ndash4 ndash2

ndash10ndash10

0505

20

20

ndash5

1

ndash8

60 120 180

Mean NDVI

Par

tial e

ffect

Altitude Distance urban Distance oil

200 000100 000120 00060 000014007000 0

Fig 2 Partial effects of the predictors included in the most-parsimonious model about the variables influencing guanaco occurrence (Table S1)Dashed lines represent 95 confidence intervals for the mean effect NDVI Normalised Difference Vegetation Index

60deg W

30deg S

50deg S

100 km

Low

Medium

High

Sightings

N

Fig 3 Habitat suitability map of guanaco constructed from the most parsimonious model (Table S1) in central Patagonia Argentina Valuesrepresent the probability of guanaco contact in a 1-km cell and are categorised into three classes (low lt033 medium 033ndash066 high gt066)Guanaco sightings are indicated by circles Areas in white correspond to regions without predictions sea lakes forested areas or outside themodelrsquos environmental space

F Wildlife Research J Pedrana et al

Guanaco density estimation

Predictions of guanaco occurrence were not made for waterbodies forests and other habitats unsuitable for guanaco sothe total area for which abundance was estimated included221 091 km2 which represents 98 of the extent of Chubut

After considering competing detection functions thesmallest AIC value indicated that the hazard-rate key withthree-parameter cosine adjustment term should be selected forabundance estimation for each guanaco social group and one-parameter cosine adjustment term for abundance estimationat each HSM category of guanaco occurrence We truncateddistance data at 200m in all models in order to attain a robustline transect analysis after the elimination of outliers (Bucklandet al 2001) Our truncation distance closely follows g(x) = 015where g is the detection function and x is the perpendiculardistance

For the entire Chubut we estimated a total guanacopopulation of 657 082 individuals (95 CI 457 437 to944 059) which represents a mean density of 297 individualsper km2 (Table 3) Mean density varied considerably betweenguanaco breeding groups and non-breeding groups (048 and021 individuals per km2 respectively Table 3) In areasclassified as having low habitat suitability (HS) for guanacothe estimated density was 049 individuals per km2 whereasin the medium and high HS categories densities were 274and 393 individuals per km2 respectively The CV variedbetween 41 and 20 (Table 3) As expected mean densitypositively increased in relation to the HS categories (Table 3)

Guanaco monitoring

Assuming annual monitoring surveys simulations showed thatat least thirty 30-km transects are needed to detect confidentlyany decrease (sustainable use scenario) or increase (alternativescenario) over the next 6 or 10 years in the low HS category(Fig 4a) Required power (080) was attained for most scenariosand HS categories based on a quite reasonable sampling effortHowever power over 090 was attained only with 10 30-kmtransects for any monitoring scenario for the medium andhigh HS categories (Fig 4b c)

Discussion

The sustainable use of any wildlife species requires a thoroughknowledgeof its distribution and abundanceAlso it is important

to develop a scientifically soundmonitoring protocol to evaluatepopulation trends under management Here we provide thesethree elements for the guanaco a valuable wildlife species thatinhabits one of the Neotropical ecosystems most severelyaffected by desertification

Guanaco distribution in Chubut showed a clear geographicalpattern with a gradient from low occurrence probability in thewest and to higher values in the east (Fig 3) The areas withhigher habitat suitability for the guanaco are coincident withthePatagonian semi-deserts (Paruelo et al 1998) and the ʻCentralDistrictrsquo delimited by Soriano (1956) and described by Leoacutenet al (1998) This is the less productive area in Chubut althoughit could be considered one of the most productive semi-deserts(450 kg handash1 yearndash1) in the world (Paruelo et al 1998) Theguanaco distribution pattern observed here has a certainsimilarity to those described for Tierra del Fuego (Raedeke1979 1982) and Santa Cruz (Pedrana et al 2010) provinceswhere higher habitat suitability for the species was found in theless productive and more arid areas in remarkable coincidencewith those areas where sheep husbandry is absent or sheepoccur at a very low density (Pedrana et al 2010) The westof Chubut where guanaco density is low is characterised bya mosaic of shrub-grass steppes with an annual net primaryproductivity higher than that of the eastern semi-deserts (650 kghandash1 yearndash1 Paruelo et al 1998) Habitat suitability modelshighlighted the western region as a low-quality habitat forguanaco Throughout this area the higher productivity hasfavoured more intense and sustained sheep ranching (Baldiet al 2001) Probably as a consequence of interference fromthis activity guanacos have become scarce although they seemto have remained in sheep-free less productive areas Guanacooccurrence in Chubut was positively associated with lowproductivity areas characterised by low NDVI values and lowelevation and far away from oil exploitation and urban areasOur results can be interpreted as a guanaco preference inducedby anthropogenic factors for this arid habitat however it alsoseems plausible that this pattern may not reflect a true habitatpreference but an indirect response to exogenous factors(competition with sheep and a response to direct persecutionby ranchers or poaching) We also found a positive associationof guanaco with remote areas far away from centres of humaninfluence such as cities and oil camps which is consistentwith a guanaco avoidance of areas that could have intensivepoaching These relationships could suggest the hypothesis that

Table 3 Estimates of guanaco density (D) and total population size (N) for each social group and in each class of Habitat Suitability Map (HSM)in central Patagonia Argentina

Upper and lower limits of 95 confidence of intervals for density estimates are given CV coefficient of variation of density estimates

Group kmndash2 D (ind kmndash2) Lower limit Upper limit CV N

Social Groups Breeding 008 048 023 098 3834 105 239Non Breeding 004 021 011 037 3071 45 324Solitary 015 085 046 157 3194 187 706Undetermined 025 144 088 236 2547 318 813Pooled 051 297 270 427 1861 657 082

HSM classes Low 009 049 022 110 4102 38 217Medium 047 274 165 456 2622 186 649High 068 393 265 582 2016 299 105Pooled 046 269 192 377 1736 523 971

Guanaco abundance distribution and monitoring Wildlife Research G

human activities have a direct and negative effect on guanacooccurrence suggesting either a behavioural avoidance of siteswhere mortality risk by humans is high and recurrent or atrue decline of guanaco local population due to overhuntingAccording to Pedrana et al (2010) the same habitat selectionpattern was found in Santa Cruz where high-suitability habitatfor guanacos corresponded with the less productive areas

The extent to which the strong inverse correlation betweenvegetation productivity and guanaco density depends onavoidance by guanacos of high sheep density areas theirillegal culling by ranchers or because they actively selectmore arid vegetation communities remains an unresolved

question (Pedrana et al 2010) One way to address thisquestion is estimating guanaco density and modelling habitatin protected areas with different vegetation communitieswhere sheep and poaching has been excluded (or reduced)Comparing guanaco population and density estimates betweenChubut and other published guanaco figures is challengingsince stratification of the study area survey effort andestimation methods are quite different It has been debatedthat road transects are not an adequate tool for estimatingguanaco population size and occurrence (Schroeder et al2018) because roads rarely cross a region in a random patternand because guanacos could be attracted or repelled by roadsIn Patagonia besides vehicle survey there are two alternativemethods that could be used to calculate guanaco abundancebut it is unclear that these methods would allow extensivedistance sampling over large regions In the case ofhorseback distance sampling has several disadvantagesrugged terrain could hardly be sampled surveying areas faraway from human habitation would be unfeasible (with theconsequent bias) and it would be difficult to cover largeregions during a short session Aircraft distance samplingcould solve some of these problems but has its own uniquedisadvantages such as unreliable counting of individuals ordetermination of social units Although sampling away fromroads would be desirable what we propose here is a cost-effective method of density estimation because aerial surveysare much more expensive than road surveys (Travaini et al2007) and hence rarely undertaken in practice Indeed nopublished guanaco population estimates exist for large areasbased on aerial surveys We covered three times more distancethan the total length of aerial transects reported by Amaya et al(2001) Testing empirically which of the two methods givesmore accurate and cost-effective estimates for guanaco densityand population size in large areas is an important topic to beaddressed in future studies

According to the International Union for Conservationof Nature (IUCN) most of the worldrsquos remaining guanacosoccur in Argentina Argentina is estimated to be home to12ndash19million guanacos (between 81ndash86 of the entirepopulation) most of them concentrated in the Patagonianregion (Baldi et al 2016) Even though its distribution rangeis extended in almost all of Argentine Patagonia guanacopopulations seem to be more fragmented in the northern andcentral provinces (Chubut Riacuteo Negro and Neuqueacuten) comparedwith the southern ones (Santa Cruz and Tierra del Fuego Baldiet al 2016) Guanaco local density estimations for Patagoniaranged from 040 individuals per km2 in some areas of Tierradel Fuego (Montes et al 2000) to more than 10 individuals perkm2 in Neuqueacuten (Rey et al 2009) Riacuteo Negro (Rey 2010) andChubut (Saba and Battro 1987) Our guanaco density estimatescan be comparedwith those in Santa Cruz which were producedusing the same methods (Pedrana et al 2010 Travaini et al2015) In the present study we found that guanaco densitiesin Chubut (2ndash4 individuals per km2) were lower than thoseestimated in Santa Cruz (3ndash7 individuals per km2 table 2 inTravaini et al 2015) While both provinces are similar in size(Chubut 221 866 km2 Santa Cruz 223 117 km2) we estimatethat the Chubut guanaco population is only half of that in SantaCruz (Travaini et al 2015)

100 (a)

(b)

(c)

095

090

085

080

075

070

Pow

er plusmn

sd

065

060

055

050

100

098

096

094

092

090

40 change over thenext 6 years

Annualsurveys Bi-annual

surveys Tri-annualsurveys

Annualsurveys

Bi-annualsurveys

Tri-annualsurveys

Annualsurveys

Annualsurveys

Annualsurveys

Annualsurveys

Bi-annualsurveys

Bi-annualsurveys

Bi-annualsurveys

Bi-annualsurveys

Tri-annualsurveys

Tri-annualsurveys

Tri-annualsurveys

Tri-annualsurveys

40 change over thenext 10 years

088

086

084

100

098

096

094

092

090

088

086

084Monitoring scenarios

Fig 4 Power (sd) for the detection of a decrease in the Chubut guanaco population (solid diamonds) or an increase (solid squares) under different scenarios 40 change over the next 6 or 10 years and survey frequency(a) low-density strata and thirty 30-km transects of road survey (b) medium-density and 10 30-km transects of road survey and (c) high-density strataand 10 30-km transects of road survey

H Wildlife Research J Pedrana et al

Finally if guanaco populations are to be sustained adaptivemanagement would require the monitoring of these populationsfor the early detection of any increase or unwanted decrease intheir numbers Traveling 600 km every 3 years is adequate todetect quite large changes within the two most dense guanacostrata in Chubut where any extractive enterprise might startBased on our experience 30-km transects should be distributedthroughout the strata area which we estimate would take sixcomplete working days by two or three technicians (Travainiet al 2015) However if the low HS category is included inthe follow-up the effort needed would be greater about ninefull-time working days per year and 15 fieldwork days every3 years Ideally an articulated association between Chubutand its neighbouring Santa Cruz could result in the bestbenefits both for the monitoring and conservation of guanacowild populations Moreover these two provinces should evendevelop common strategies for the sustainable use andregulation in the trade of guanaco products

Conservation and management implications

Patagonia holds some 700 000 km2 of grasslands and scrublandsthat still maintain healthy guanaco populations (Baldi et al2010) despite having been deeply modified by humanactivities Both ecosystem structure and functioning werealtered by the introduction of domestic herbivores at thebeginning of the 20th Century (Aguiar et al 1996 Bisigatoand Bertiller 1997) Grazing mining (Radovani et al 2014)and more recently global climate change (Pinto et al 2008)are human-related disturbances acting at different spatial andtemporal scales that have degraded Patagonian vegetationCurrent conditions are very different from the ones encounteredby the first European settlers who were unable to maintaina traditional sheep ranching system that was environmentallysustainable and economically profitable (Texeira and Paruelo2006) Traditional sheep ranching involved keeping sheepgrazing in extensive paddocks (requiring several thousandhectares) with limited attention a few occasions a year fordeworming or shearing (Garciacutea Brea et al 2010) Restoringthe Patagonian shrub-steppe as an ecologically functionalecosystem (Chardonnet et al 2002) would benefit fromefficient and modernised sheep husbandry complemented byactivities such as the sustainable use of wildlife and ecotourismIn this framework guanaco could play a role as an importanteconomic resource for landowners and local communities(Franklin et al 1997) In the absence of alternatives economicincentives could be the most effective driver for the conservationof the species (Fritz and Franklin 1994)

The greatest utility of our results is that they provide anexcellent overview of guanaco distribution and abundance in2006 and the big picture for the entire Chubut from whichthe government could start planning the sustainable use of thespecies Considering our HSM it is possible to identify theareas of the territory with higher densities where it will beeasier to select ranches with adequate guanaco abundance(Baldi et al 2010) to develop profitable herding and shearingexperiences Our HSM for Chubut together with otherenvironmental and human spatial data (eg land use roadcover and cadastral) should be used to generate a decision

tool that guarantees informed sustainable use of the speciesFinally our results need to be implemented in accordance withenvironmental education programs to change the perceptionof ranchers in Chubut province on native wildlife

Conflicts of interest

The authors declare no conflicts of interest

Acknowledgements

This work was primarily funded by the BBVA Foundation through a grantunder the Conservation Biology Program to A Rodriacuteguez and also bythe Agencia Nacional de Promocioacuten Cientiacutefica y Tecnoloacutegica (PICTON 30723) Additional support was provided by Universidad Nacional dela Patagonia Austral CONICET (PEI-6065) and CONAE

References

Aguiar M R Paruelo J M Sala O E and Laurenroth W K (1996)Ecosystem consequences of plant functional types changes in semiaridgrassland Journal of Vegetation Science 7 381ndash390 doi1023073236281

Amaya JN von Thuumlngen J and de Lamo DA (2001) Relevamiento ydistribucioacutendeguanacosen laPatagoniaComunicacioacutenTeacutecnicano109(Instituto Nacional de Tecnologiacutea Agropecuaria Bariloche Argentina)[In Spanish]

Ares J O BeeskowAM BertillerM B Rostagno CM IrisarriM PAnchorena J Defosseacute G E and Meroni C (1990) Structuraland dynamic characteristics of overgrazed grasslands of northernPatagonia Argentina In lsquoManaged Grasslands Regional Studiesrsquo(Ed A Breymeyer) pp 149ndash175 (Elsevier Amsterdam)

Baldi R Albon S D and Elston D A (2001) Guanacos and sheepevidence for continuing competition in arid Patagonia Oecologia 129561ndash570 doi101007s004420100770

Baldi R Novaro A Funes M Walker S Ferrando P Failla M andCarmanchahi P (2010) Guanacomanagement in Patagonian rangelandsa conservation opportunity on the brink of collapse In lsquoConservingWildlife while Maintaining Livestock in Semi-Arid Ecosystemsrsquo (EdsJ Du Toit R Kock and J Deutsch) pp 266ndash290 (Blackwell PublishingOxford UK)

Baldi R Acebes P Cueacutellar E Funes M Hoces D Puig S andFranklin W L (2016) Lama guanicoe The IUCN Red List ofThreatened Species 2016 Available at httpwwwiucnredlistorg[Accessed 6 April 2018]

BarrosVR ScianBV andMattioHF (1979)Camposdeprecipitacioacutende la provincia de Chubut (1931ndash1960) Geoacta 10 175ndash192

Bisigato A J andBertillerMB (1997) Grazing effects on patchy drylandvegetation in northern Patagonia Journal of Arid Environments 36639ndash653 doi101006jare19960247

Borrelli P and Cibils A (2005) Rural depopulation and grasslandmanagement in Patagonia In lsquoGrasslands Development OpportunitiesPerspectivesrsquo (Eds S G Reynolds and J Frame) pp 461ndash488 (FAOEnfield NH)

Buckland S T Anderson D R Burnham K P Laake J L BorchersD L and Thomas L (2001) lsquoIntroduction to Distance SamplingEstimating Abundance of Biological Populationsrsquo (Oxford UniversityPress Oxford)

Burnham K P and Anderson D R (2002) lsquoModel Selection andMultimodel Inference a Practical Information-theoretic Approachrsquo(Springer-Verlag New York)

Caro J Zapata S C Zanoacuten J I Rodriacuteguez A and Travaini A (2017)Ganaderiacutea ovina y usos alternativos del suelo en la Patagonia australArgentina Multequina (Mendoza) 26 33ndash50

Guanaco abundance distribution and monitoring Wildlife Research I

and dry climate with strong and predominantly west winds andwith a mean annual temperature that usually ranges between 6and 12C (Paruelo et al 1998) Annual precipitation of ~200mm(Paruelo et al 1998) concentrates during winter and is stronglyinfluenced by general circulation patterns the Pacific air massesand the Andes topographic barrier parallel to the Pacific coastresulting in a strong westndasheast gradient (Barros et al 1979)The dominant vegetation type is shrub steppe composed ofmedium height to dwarf shrubs (Nassauvia glomerulosaN ulicina Chuquiraga aurea C avellanedae) with variabletussock grass cover (Stipa humilis S ibari) Extensive sheepgrazing is the predominant agricultural practice in PatagoniaMerino sheep of varying quality are kept in the field throughoutthe year herded accordingly to a summerndashwinter rotationscheme and fed with a varying seasonal forage dependingon landscape and altitude (Ares et al 1990) Ecosystemdegradation has been attributed to human managementfactors such as overestimation of carrying capacity of therangelands inadequate distribution of animals in extremelylarge heterogeneous paddocks and year-long continuousgrazing (Golluscio et al 1998)

Study speciesGuanacos are sexually monomorphic camelids that aggregatein social groups during the summer breeding seasonAccording to Franklin (1983) and Young and Franklin (2004)

the spatial distribution of populations is influenced by a matingsystem of resource defence polygyny (a territorial systemwherein males compete for access to resources required byfemales) Guanacos can be found in three basic social unitsfamily groups (a male with a group of females and their progenyyounger than one-year old called lsquochulengosrsquo) non-reproductive male groups and solitary males A fourthpotential social unit female groups was only describedduring intensive field studies in Torres del Paine NationalPark Chile (Ortega and Franklin 1995) but this social unitwas not considered in our guanaco censuses because it wasuncommon even in the study reporting it (8 of all socialunits observed Ortega and Franklin 1995) and might includeseveral females with young making female groupsindistinguishable from family groups Pedrana et al (2009)empirically demonstrated that only family groups withchulengos can be unequivocally identified during surveys byvehicle and that the identification of male groups was alsochallenging because these could be mistaken for familygroups without or with undetected chulengos During thepresent study we used four observable social units as a formof stratification to produce separate density estimates breedinggroups (family groups with detected chulengos) non-breedinggroups (groups without chulengos) solitary animals andundetermined (groups that could not be clearly assigned toany of the first three categories Pedrana et al 2009)

60deg W

30deg S

50deg S

Chubut

Santa Cruz

100 km

Survey Tracks November 2006

Flat ndash Low NDVI

N

Flat ndash Medium NDVI

Flat ndash High NDVI

Rugged ndash Low NDVI

Rugged ndash Medium NDVI

Rugged ndash High NDVI

Survey Tracks December 2006

Road network

Fig 1 Study area stratification based on productivity (three Normalised Difference Vegetation Index (NDVI) categories low mediumand high) and topography (two categories flat and rugged terrain) and roads surveyed in central Patagonia Argentina

Guanaco abundance distribution and monitoring Wildlife Research C

Sampling unit selection and field workTo account for the expected regional variation in guanacodensities associated with major environmental variability weused a stratified random design to select the road segments tobe surveyed For this reason we categorised the study area intosix geographical regions based on the combination of twohabitat characteristics that affect guanaco detection anddensity (1) Normalised Difference Vegetation Index (NDVI)and (2) mean slope (Fig 1 Travaini et al 2007) Indeedboth variables influenced guanaco distribution in Santa Cruz(Pedrana et al 2010)

Fieldworkwas conducted during one austral springndashsummerwhich overlaps with guanaco breeding season Using a vectordata of road coverage we randomly selected 110 survey tracks(road segments of 5 to 80 km) that summed to 4440 km oftransects during the first survey (9ndash27 November 2006Fig 1) During our second survey (7ndash17 December 2006) werandomly selected another 54 survey tracks that added up to2570 km (Fig 1) Almost 90 of the survey tracks were dirtand secondary roads with very low traffic density (ie fewerthan 10 vehicles per day)

Because of their body size colouration and behaviourguanacos are easily observable from vehicles driving on roadsand secondary trails (Puig 1995) in the open steppe or shrub-steppe These conditions are excellent for estimating presenceand even abundance by vehicle survey using distance samplingmethodology (Buckland et al 2001) Road surveys were carriedout from a vehicle driving mostly at a speed of 20 kmh1 witha driver and one observer during one breeding season (Fig 1)following recommendations by Pedrana et al (2010) Whena guanaco was detected we stopped the vehicle recorded groupsize and assigned the group to one of our observable socialgroups We measured the distance to the animal or the groupcentre with a laser range finder (Leica LRF 1200 RangemasterGermany) We also recorded the azimuth of the observerrsquostrajectory obtained from the inertial compass of a GPS unit(Garmin GPS mapR 76S USA) and the angle of the animalrelative to our bearing (using a protractor) Distance and azimuthallowed us to obtain the actual positions of guanacos andtheir perpendicular distance to the survey line as required bydistance-sampling methods (Buckland et al 2001) All datawere collected in a personal digital assistant (PDA TungstenT3 USA) synchronised with the GPS which was used to recordthe trajectory and location of the survey track

Habitat suitability maps for guanacoWe selected eight potential predictors that summarised themost relevant environmental and landscape gradients as wellas anthropogenic features needed to assess which factors mightinfluence the regional distribution of guanacos (Table 1 forfurther details see Travaini et al 2007 Pedrana et al 2010)Distribution modelling requires defining units in whichpresence or absence is recorded For this purpose we overlaidthe surveyed tracks on top of a 1-km grid following the spatialresolution of NDVI data Then the original 457 sightingsregistered during field surveys were also overlaid and gridcells with at least one guanaco sighting were considered as apresence cell and all remaining cells traversed by census tracks

were considered as absences As cells with guanaco presence(n= 310) were outnumbered by absence cells (n= 15 606)we used a resampling scheme to obtain a presenceabsencebalanced sample (Travaini et al 2015) We randomly selected310 out of the 15 606 cells with guanaco absences repeatingthis procedure 100 times Each time we generated two datasetsa random sample of 80 of the cells for calibrating theoccurrence models and the remaining 20 for evaluating themodels (hereafter referred as the construction and evaluationdatasets)All the cellswith presencewere used in each repetitionwhile cells with absence were sampled without replacement

To identify which predictors were most likely to affectguanaco occurrence model fitting was done on the constructiondataset with Generalised Additive Models (GAMs Hastieand Tibshirani 1990) using a binomial error and a logit linkPredictors for the models were selected by a backwardndashforwardstepwise procedure (Sakamoto et al 1986) The AkaikeInformation Criterion (AIC) was used to retain a term(Burnham and Anderson 2002) We considered as competingmodels those for which the differences between AIC and theAIC of the best candidate model (the one with the smallestAIC) was D 2 (Burnham and Anderson 2002)

Model evaluation was made using the validation dataset bycomparing predicted and observed values using a threshold-independent metric ie the area under the curve (AUC) ofthe Receiver Operating Characteristic (ROC) plot which wascomputed for each of the 100 fitted models (Murtaugh 1996)The AUC ranges from 0 (when model discrimination isnot better than random) to 1 (perfect discriminatory ability)Predictive models were considered usable if AUC 07

Finally predictions of the resulting distribution models wereused to create habitat suitability maps (HSM) Predictions forthe entire study area were calculated using R version 322 (RDevelopment Core Team 2016) and transformed to probabilitymaps using IDRISI Selva GIS software (Clark Labs WorcesterMA USA) We categorised and mapped the HSM of guanacooccurrence probability in three classes (low lt033 medium033ndash066 and high gt066)

Table 1 Description of the variables used in thepredictive distributionmodels developed for guanaco in central Patagonia Argentina

NVDI Normalised Difference Vegetation Index

Predictor Predictor description

MeanNDVI Mean Normalised Difference Vegetation Indexcalculated by using the VGT-S10 product thatis a 10-day maximum composite value from theVEGETATION sensor on board of Spot-4 satellitefrom April 1999 to March 2005

SeasonMAX Month at which the NDVI reaches its annualmaximum value

CVNDVI Coefficient of variation of NDVIDistancestream Distance (km) to the nearest stream or riverAltitude Mean elevation in meters above sea level of the

focal cell obtained from the Shuttle RadarTopographic Mission Elevation (SRTM)

Slope Terrain slope in degrees in a 1 km pixelfrom the SRTM

Distanceurban Distance (km) to the nearest urban areaDistanceoil Distance (km) to the nearest oil camp

D Wildlife Research J Pedrana et al

Guanaco density estimationLine transect data were analysed using the DISTANCEprogram (version 60 Buckland et al 2001 Thomas et al2010) Estimates of guanaco densities were done by fitting adetection function from the perpendicular distances of guanacosto the survey line (Buckland et al 2001) We pooled distancedata across all transects within each HSM class of predictedguanaco occurrence to estimate the detection functions g(x) foreach social unit in order to attain the required minimum numberof observations (Buckland et al 2001) Although detectabilityvaries among transects the property of lsquopooling robustnessrsquoensures the reliability of the abundance estimates (Bucklandet al 2001 Thomas et al 2010) Pooling also helps to even outminor fluctuations in object density across surveyed strips andto even out chance fluctuations in object distribution (Fewsteret al 2005) The DISTANCE program allows for carefuldata exploration facilitating the identification of any outlyingobservation at extreme distances or sighting clumping andsuggests appropriate levels of truncation and grouping fordetection function fitting We used the smallest value of AICto choose the best competing model (Key function + adjustmentterms) Density estimates were calculated for each social unitand for all guanaco contacts pooled together and for each HSMcategory and for all of Chubut Considering the extent of eachHSM category density estimates were converted into estimatesof population size

Monitoring programWe estimated and simulated the sampling effort in terms ofnumber of 30 km transects of road survey needed to monitorguanaco population in Chubut and also to detect declines frompresent status in guanaco population using MONITOR (version1100 Gibbs and Ene 2010) We modelled the encounter ratewhich was calculated as the number of guanacos detected persampled kilometre over time and to generate detection ratesderived from route-regression analyses (Gibbs and Melvin1997) This simulation procedure is useful for evaluating thetrade-offs between sampling effort logistical constraints andpower to detect trends (Field et al 2005) To estimate powerwe supplied the programwith initial estimates of encounter ratesand its variance derived empirically from our own results Wearbitrarily selected 30 km as the length of sampling units Basedon our experience a 30-km-long transect usually takes abouttwo hours to complete the interval we used to exchangeobservers during field work to guarantee observer attention

Additionally in the study area random 30-km transects canbe selected from the vector data of road coverage more easilythan longer segments To select suitable sampling efforts wefixed the minimum acceptable power at (1-b) = 080 (Di Stefano2001) ie the highest probability of failing to detect a realtrend was 020 Type I error was fixed at 005 To define theeffect size component we selected two alternative scenariosboth highly probable given the guanaco current situationpreviously described The first scenario is under the hypothesisof a sustainable use of guanaco populations to detect early a 40decrease in the next six to 10 years An alternative scenariothe second is if the guanaco is considered as a pest species inorder to early detect a 40 increase in the next six and 10 yearsAs we had sampled each one of the 164 survey tracks oncerandom resampling without replacement of 30-km transectswas used to estimate the mean and the standard deviation ofguanaco encounter rates as a surrogate of spatial variabilityAs an estimate of power variability we estimated the standarddeviation from 20 simulations for each monitoring scenarioand 2000 iterations for each simulation

Results

We surveyed 7010 km almost 76 of available national andprovincial roads in Chubut We recorded 456 guanaco socialgroups which added up to 2893 individuals (Table 2) Almost20 of individuals were registered as belonging to breedinggroups 30 as non-breeding groups and 4 as solitaryindividuals (Table 2) The remaining 47 of individuals wereclassified as undetermined (Table 2) The high percentageof undetermined individuals resulted from our interest inminimising the chances of social group misclassification Weopted for considering family groups without juveniles orwith undetected juveniles as undetermined rather thanmisclassifying them as non-breeding family groups (Pedranaet al 2009) Family groups (n) averaged 8 (sd 322) individualswhere the mean number of females and chulengos per groupwere 5 (sd 268) and 2 (sd 117) respectively (Table 2)Non-breeding and undetermined groups averaged 28 (sd1581) and 6 (sd 268) individuals per group respectively(Table 2)

Habitat suitability maps for guanaco

Themost parsimonious GAMof guanaco presence incorporatedas predictors themeanNDVI altitude the distance to the nearesturban centre and to the closest oil field (Table S1 available

Table 2 Number of guanaco groups and individuals (within brackets) and mean group size recorded during one breeding season(NovemberndashDecember 2006) in the three areas defined by Habitat Suitability Maps (HSM) and for the total area covered in central

Patagonia Argentinasd standard deviation Und Undetermined groups

HSM classes Area Effort (km) Guanaco social groups (individuals) Total Mean group size (sd)(km2) ( transects) Solitary Breeding Non

BreedingUnd Breeding Females

per groupChulengosper group

NonBreeding

Und

Low 76 741 1824 (46) 14 6 (42) 5 (159) 13 (91) 38 (306) 7 (300) 5 (289) 1 (103) 32 (2272) 7 (438)Medium 68 145 3686 (82) 46 26 (216) 16 (444) 120 (710) 208 (1416) 8 (331) 5 (283) 2 (127) 28 (143) 6 (427)High 76 205 1500 (36) 65 35 (294) 10 (259) 100 (553) 210 (1171) 8 (325) 6 (258) 2 (121) 16 (2590) 6 (359)

Total 221 091 7010 (164) 125 67 (552) 31 (862) 233 (1354) 456 (2893) 8 (322) 5 (268) 2 (117) 28 (1581) 6 (268)

Guanaco abundance distribution and monitoring Wildlife Research E

as Supplementary Material to this paper) The probability ofguanaco occurrence decreased with primary productivity andwith mean altitude whereas it increased with the distance tothe nearest urban centre and oil field (Fig 2)

The mean ( se) AUC of the best model ranged from080 (002) to 082 (002) for the evaluation dataset(Table S1) These results indicate that models have anacceptable discrimination ability and are useful for predictingguanaco distribution

The HSM of guanaco occurrence (Fig 3) corresponding tothe predictions of the best model (Table S1) showed that the

probability of guanaco occurrence was positively associatedwith a gradient of increasing aridity from west to east Areaswith low probability of guanaco occurrence predominatedthroughout the western third of Chubut within the sub-Andean and occidental physiographic districts (Soriano 1956)from the transition from forests to short grasslands steppesand scrublands to the centre of the Province (Ares et al1990) From there the probability of guanaco occurrencesteeply increased eastward up to the coastline and theValdeacutes Peninsula throughout the driest lands with the mostxeromorphic floristic composition (Fig 3)

0

ndash4 ndash2

ndash10ndash10

0505

20

20

ndash5

1

ndash8

60 120 180

Mean NDVI

Par

tial e

ffect

Altitude Distance urban Distance oil

200 000100 000120 00060 000014007000 0

Fig 2 Partial effects of the predictors included in the most-parsimonious model about the variables influencing guanaco occurrence (Table S1)Dashed lines represent 95 confidence intervals for the mean effect NDVI Normalised Difference Vegetation Index

60deg W

30deg S

50deg S

100 km

Low

Medium

High

Sightings

N

Fig 3 Habitat suitability map of guanaco constructed from the most parsimonious model (Table S1) in central Patagonia Argentina Valuesrepresent the probability of guanaco contact in a 1-km cell and are categorised into three classes (low lt033 medium 033ndash066 high gt066)Guanaco sightings are indicated by circles Areas in white correspond to regions without predictions sea lakes forested areas or outside themodelrsquos environmental space

F Wildlife Research J Pedrana et al

Guanaco density estimation

Predictions of guanaco occurrence were not made for waterbodies forests and other habitats unsuitable for guanaco sothe total area for which abundance was estimated included221 091 km2 which represents 98 of the extent of Chubut

After considering competing detection functions thesmallest AIC value indicated that the hazard-rate key withthree-parameter cosine adjustment term should be selected forabundance estimation for each guanaco social group and one-parameter cosine adjustment term for abundance estimationat each HSM category of guanaco occurrence We truncateddistance data at 200m in all models in order to attain a robustline transect analysis after the elimination of outliers (Bucklandet al 2001) Our truncation distance closely follows g(x) = 015where g is the detection function and x is the perpendiculardistance

For the entire Chubut we estimated a total guanacopopulation of 657 082 individuals (95 CI 457 437 to944 059) which represents a mean density of 297 individualsper km2 (Table 3) Mean density varied considerably betweenguanaco breeding groups and non-breeding groups (048 and021 individuals per km2 respectively Table 3) In areasclassified as having low habitat suitability (HS) for guanacothe estimated density was 049 individuals per km2 whereasin the medium and high HS categories densities were 274and 393 individuals per km2 respectively The CV variedbetween 41 and 20 (Table 3) As expected mean densitypositively increased in relation to the HS categories (Table 3)

Guanaco monitoring

Assuming annual monitoring surveys simulations showed thatat least thirty 30-km transects are needed to detect confidentlyany decrease (sustainable use scenario) or increase (alternativescenario) over the next 6 or 10 years in the low HS category(Fig 4a) Required power (080) was attained for most scenariosand HS categories based on a quite reasonable sampling effortHowever power over 090 was attained only with 10 30-kmtransects for any monitoring scenario for the medium andhigh HS categories (Fig 4b c)

Discussion

The sustainable use of any wildlife species requires a thoroughknowledgeof its distribution and abundanceAlso it is important

to develop a scientifically soundmonitoring protocol to evaluatepopulation trends under management Here we provide thesethree elements for the guanaco a valuable wildlife species thatinhabits one of the Neotropical ecosystems most severelyaffected by desertification

Guanaco distribution in Chubut showed a clear geographicalpattern with a gradient from low occurrence probability in thewest and to higher values in the east (Fig 3) The areas withhigher habitat suitability for the guanaco are coincident withthePatagonian semi-deserts (Paruelo et al 1998) and the ʻCentralDistrictrsquo delimited by Soriano (1956) and described by Leoacutenet al (1998) This is the less productive area in Chubut althoughit could be considered one of the most productive semi-deserts(450 kg handash1 yearndash1) in the world (Paruelo et al 1998) Theguanaco distribution pattern observed here has a certainsimilarity to those described for Tierra del Fuego (Raedeke1979 1982) and Santa Cruz (Pedrana et al 2010) provinceswhere higher habitat suitability for the species was found in theless productive and more arid areas in remarkable coincidencewith those areas where sheep husbandry is absent or sheepoccur at a very low density (Pedrana et al 2010) The westof Chubut where guanaco density is low is characterised bya mosaic of shrub-grass steppes with an annual net primaryproductivity higher than that of the eastern semi-deserts (650 kghandash1 yearndash1 Paruelo et al 1998) Habitat suitability modelshighlighted the western region as a low-quality habitat forguanaco Throughout this area the higher productivity hasfavoured more intense and sustained sheep ranching (Baldiet al 2001) Probably as a consequence of interference fromthis activity guanacos have become scarce although they seemto have remained in sheep-free less productive areas Guanacooccurrence in Chubut was positively associated with lowproductivity areas characterised by low NDVI values and lowelevation and far away from oil exploitation and urban areasOur results can be interpreted as a guanaco preference inducedby anthropogenic factors for this arid habitat however it alsoseems plausible that this pattern may not reflect a true habitatpreference but an indirect response to exogenous factors(competition with sheep and a response to direct persecutionby ranchers or poaching) We also found a positive associationof guanaco with remote areas far away from centres of humaninfluence such as cities and oil camps which is consistentwith a guanaco avoidance of areas that could have intensivepoaching These relationships could suggest the hypothesis that

Table 3 Estimates of guanaco density (D) and total population size (N) for each social group and in each class of Habitat Suitability Map (HSM)in central Patagonia Argentina

Upper and lower limits of 95 confidence of intervals for density estimates are given CV coefficient of variation of density estimates

Group kmndash2 D (ind kmndash2) Lower limit Upper limit CV N

Social Groups Breeding 008 048 023 098 3834 105 239Non Breeding 004 021 011 037 3071 45 324Solitary 015 085 046 157 3194 187 706Undetermined 025 144 088 236 2547 318 813Pooled 051 297 270 427 1861 657 082

HSM classes Low 009 049 022 110 4102 38 217Medium 047 274 165 456 2622 186 649High 068 393 265 582 2016 299 105Pooled 046 269 192 377 1736 523 971

Guanaco abundance distribution and monitoring Wildlife Research G

human activities have a direct and negative effect on guanacooccurrence suggesting either a behavioural avoidance of siteswhere mortality risk by humans is high and recurrent or atrue decline of guanaco local population due to overhuntingAccording to Pedrana et al (2010) the same habitat selectionpattern was found in Santa Cruz where high-suitability habitatfor guanacos corresponded with the less productive areas

The extent to which the strong inverse correlation betweenvegetation productivity and guanaco density depends onavoidance by guanacos of high sheep density areas theirillegal culling by ranchers or because they actively selectmore arid vegetation communities remains an unresolved

question (Pedrana et al 2010) One way to address thisquestion is estimating guanaco density and modelling habitatin protected areas with different vegetation communitieswhere sheep and poaching has been excluded (or reduced)Comparing guanaco population and density estimates betweenChubut and other published guanaco figures is challengingsince stratification of the study area survey effort andestimation methods are quite different It has been debatedthat road transects are not an adequate tool for estimatingguanaco population size and occurrence (Schroeder et al2018) because roads rarely cross a region in a random patternand because guanacos could be attracted or repelled by roadsIn Patagonia besides vehicle survey there are two alternativemethods that could be used to calculate guanaco abundancebut it is unclear that these methods would allow extensivedistance sampling over large regions In the case ofhorseback distance sampling has several disadvantagesrugged terrain could hardly be sampled surveying areas faraway from human habitation would be unfeasible (with theconsequent bias) and it would be difficult to cover largeregions during a short session Aircraft distance samplingcould solve some of these problems but has its own uniquedisadvantages such as unreliable counting of individuals ordetermination of social units Although sampling away fromroads would be desirable what we propose here is a cost-effective method of density estimation because aerial surveysare much more expensive than road surveys (Travaini et al2007) and hence rarely undertaken in practice Indeed nopublished guanaco population estimates exist for large areasbased on aerial surveys We covered three times more distancethan the total length of aerial transects reported by Amaya et al(2001) Testing empirically which of the two methods givesmore accurate and cost-effective estimates for guanaco densityand population size in large areas is an important topic to beaddressed in future studies

According to the International Union for Conservationof Nature (IUCN) most of the worldrsquos remaining guanacosoccur in Argentina Argentina is estimated to be home to12ndash19million guanacos (between 81ndash86 of the entirepopulation) most of them concentrated in the Patagonianregion (Baldi et al 2016) Even though its distribution rangeis extended in almost all of Argentine Patagonia guanacopopulations seem to be more fragmented in the northern andcentral provinces (Chubut Riacuteo Negro and Neuqueacuten) comparedwith the southern ones (Santa Cruz and Tierra del Fuego Baldiet al 2016) Guanaco local density estimations for Patagoniaranged from 040 individuals per km2 in some areas of Tierradel Fuego (Montes et al 2000) to more than 10 individuals perkm2 in Neuqueacuten (Rey et al 2009) Riacuteo Negro (Rey 2010) andChubut (Saba and Battro 1987) Our guanaco density estimatescan be comparedwith those in Santa Cruz which were producedusing the same methods (Pedrana et al 2010 Travaini et al2015) In the present study we found that guanaco densitiesin Chubut (2ndash4 individuals per km2) were lower than thoseestimated in Santa Cruz (3ndash7 individuals per km2 table 2 inTravaini et al 2015) While both provinces are similar in size(Chubut 221 866 km2 Santa Cruz 223 117 km2) we estimatethat the Chubut guanaco population is only half of that in SantaCruz (Travaini et al 2015)

100 (a)

(b)

(c)

095

090

085

080

075

070

Pow

er plusmn

sd

065

060

055

050

100

098

096

094

092

090

40 change over thenext 6 years

Annualsurveys Bi-annual

surveys Tri-annualsurveys

Annualsurveys

Bi-annualsurveys

Tri-annualsurveys

Annualsurveys

Annualsurveys

Annualsurveys

Annualsurveys

Bi-annualsurveys

Bi-annualsurveys

Bi-annualsurveys

Bi-annualsurveys

Tri-annualsurveys

Tri-annualsurveys

Tri-annualsurveys

Tri-annualsurveys

40 change over thenext 10 years

088

086

084

100

098

096

094

092

090

088

086

084Monitoring scenarios

Fig 4 Power (sd) for the detection of a decrease in the Chubut guanaco population (solid diamonds) or an increase (solid squares) under different scenarios 40 change over the next 6 or 10 years and survey frequency(a) low-density strata and thirty 30-km transects of road survey (b) medium-density and 10 30-km transects of road survey and (c) high-density strataand 10 30-km transects of road survey

H Wildlife Research J Pedrana et al

Finally if guanaco populations are to be sustained adaptivemanagement would require the monitoring of these populationsfor the early detection of any increase or unwanted decrease intheir numbers Traveling 600 km every 3 years is adequate todetect quite large changes within the two most dense guanacostrata in Chubut where any extractive enterprise might startBased on our experience 30-km transects should be distributedthroughout the strata area which we estimate would take sixcomplete working days by two or three technicians (Travainiet al 2015) However if the low HS category is included inthe follow-up the effort needed would be greater about ninefull-time working days per year and 15 fieldwork days every3 years Ideally an articulated association between Chubutand its neighbouring Santa Cruz could result in the bestbenefits both for the monitoring and conservation of guanacowild populations Moreover these two provinces should evendevelop common strategies for the sustainable use andregulation in the trade of guanaco products

Conservation and management implications

Patagonia holds some 700 000 km2 of grasslands and scrublandsthat still maintain healthy guanaco populations (Baldi et al2010) despite having been deeply modified by humanactivities Both ecosystem structure and functioning werealtered by the introduction of domestic herbivores at thebeginning of the 20th Century (Aguiar et al 1996 Bisigatoand Bertiller 1997) Grazing mining (Radovani et al 2014)and more recently global climate change (Pinto et al 2008)are human-related disturbances acting at different spatial andtemporal scales that have degraded Patagonian vegetationCurrent conditions are very different from the ones encounteredby the first European settlers who were unable to maintaina traditional sheep ranching system that was environmentallysustainable and economically profitable (Texeira and Paruelo2006) Traditional sheep ranching involved keeping sheepgrazing in extensive paddocks (requiring several thousandhectares) with limited attention a few occasions a year fordeworming or shearing (Garciacutea Brea et al 2010) Restoringthe Patagonian shrub-steppe as an ecologically functionalecosystem (Chardonnet et al 2002) would benefit fromefficient and modernised sheep husbandry complemented byactivities such as the sustainable use of wildlife and ecotourismIn this framework guanaco could play a role as an importanteconomic resource for landowners and local communities(Franklin et al 1997) In the absence of alternatives economicincentives could be the most effective driver for the conservationof the species (Fritz and Franklin 1994)

The greatest utility of our results is that they provide anexcellent overview of guanaco distribution and abundance in2006 and the big picture for the entire Chubut from whichthe government could start planning the sustainable use of thespecies Considering our HSM it is possible to identify theareas of the territory with higher densities where it will beeasier to select ranches with adequate guanaco abundance(Baldi et al 2010) to develop profitable herding and shearingexperiences Our HSM for Chubut together with otherenvironmental and human spatial data (eg land use roadcover and cadastral) should be used to generate a decision

tool that guarantees informed sustainable use of the speciesFinally our results need to be implemented in accordance withenvironmental education programs to change the perceptionof ranchers in Chubut province on native wildlife

Conflicts of interest

The authors declare no conflicts of interest

Acknowledgements

This work was primarily funded by the BBVA Foundation through a grantunder the Conservation Biology Program to A Rodriacuteguez and also bythe Agencia Nacional de Promocioacuten Cientiacutefica y Tecnoloacutegica (PICTON 30723) Additional support was provided by Universidad Nacional dela Patagonia Austral CONICET (PEI-6065) and CONAE

References

Aguiar M R Paruelo J M Sala O E and Laurenroth W K (1996)Ecosystem consequences of plant functional types changes in semiaridgrassland Journal of Vegetation Science 7 381ndash390 doi1023073236281

Amaya JN von Thuumlngen J and de Lamo DA (2001) Relevamiento ydistribucioacutendeguanacosen laPatagoniaComunicacioacutenTeacutecnicano109(Instituto Nacional de Tecnologiacutea Agropecuaria Bariloche Argentina)[In Spanish]

Ares J O BeeskowAM BertillerM B Rostagno CM IrisarriM PAnchorena J Defosseacute G E and Meroni C (1990) Structuraland dynamic characteristics of overgrazed grasslands of northernPatagonia Argentina In lsquoManaged Grasslands Regional Studiesrsquo(Ed A Breymeyer) pp 149ndash175 (Elsevier Amsterdam)

Baldi R Albon S D and Elston D A (2001) Guanacos and sheepevidence for continuing competition in arid Patagonia Oecologia 129561ndash570 doi101007s004420100770

Baldi R Novaro A Funes M Walker S Ferrando P Failla M andCarmanchahi P (2010) Guanacomanagement in Patagonian rangelandsa conservation opportunity on the brink of collapse In lsquoConservingWildlife while Maintaining Livestock in Semi-Arid Ecosystemsrsquo (EdsJ Du Toit R Kock and J Deutsch) pp 266ndash290 (Blackwell PublishingOxford UK)

Baldi R Acebes P Cueacutellar E Funes M Hoces D Puig S andFranklin W L (2016) Lama guanicoe The IUCN Red List ofThreatened Species 2016 Available at httpwwwiucnredlistorg[Accessed 6 April 2018]

BarrosVR ScianBV andMattioHF (1979)Camposdeprecipitacioacutende la provincia de Chubut (1931ndash1960) Geoacta 10 175ndash192

Bisigato A J andBertillerMB (1997) Grazing effects on patchy drylandvegetation in northern Patagonia Journal of Arid Environments 36639ndash653 doi101006jare19960247

Borrelli P and Cibils A (2005) Rural depopulation and grasslandmanagement in Patagonia In lsquoGrasslands Development OpportunitiesPerspectivesrsquo (Eds S G Reynolds and J Frame) pp 461ndash488 (FAOEnfield NH)

Buckland S T Anderson D R Burnham K P Laake J L BorchersD L and Thomas L (2001) lsquoIntroduction to Distance SamplingEstimating Abundance of Biological Populationsrsquo (Oxford UniversityPress Oxford)

Burnham K P and Anderson D R (2002) lsquoModel Selection andMultimodel Inference a Practical Information-theoretic Approachrsquo(Springer-Verlag New York)

Caro J Zapata S C Zanoacuten J I Rodriacuteguez A and Travaini A (2017)Ganaderiacutea ovina y usos alternativos del suelo en la Patagonia australArgentina Multequina (Mendoza) 26 33ndash50

Guanaco abundance distribution and monitoring Wildlife Research I

Sampling unit selection and field workTo account for the expected regional variation in guanacodensities associated with major environmental variability weused a stratified random design to select the road segments tobe surveyed For this reason we categorised the study area intosix geographical regions based on the combination of twohabitat characteristics that affect guanaco detection anddensity (1) Normalised Difference Vegetation Index (NDVI)and (2) mean slope (Fig 1 Travaini et al 2007) Indeedboth variables influenced guanaco distribution in Santa Cruz(Pedrana et al 2010)

Fieldworkwas conducted during one austral springndashsummerwhich overlaps with guanaco breeding season Using a vectordata of road coverage we randomly selected 110 survey tracks(road segments of 5 to 80 km) that summed to 4440 km oftransects during the first survey (9ndash27 November 2006Fig 1) During our second survey (7ndash17 December 2006) werandomly selected another 54 survey tracks that added up to2570 km (Fig 1) Almost 90 of the survey tracks were dirtand secondary roads with very low traffic density (ie fewerthan 10 vehicles per day)

Because of their body size colouration and behaviourguanacos are easily observable from vehicles driving on roadsand secondary trails (Puig 1995) in the open steppe or shrub-steppe These conditions are excellent for estimating presenceand even abundance by vehicle survey using distance samplingmethodology (Buckland et al 2001) Road surveys were carriedout from a vehicle driving mostly at a speed of 20 kmh1 witha driver and one observer during one breeding season (Fig 1)following recommendations by Pedrana et al (2010) Whena guanaco was detected we stopped the vehicle recorded groupsize and assigned the group to one of our observable socialgroups We measured the distance to the animal or the groupcentre with a laser range finder (Leica LRF 1200 RangemasterGermany) We also recorded the azimuth of the observerrsquostrajectory obtained from the inertial compass of a GPS unit(Garmin GPS mapR 76S USA) and the angle of the animalrelative to our bearing (using a protractor) Distance and azimuthallowed us to obtain the actual positions of guanacos andtheir perpendicular distance to the survey line as required bydistance-sampling methods (Buckland et al 2001) All datawere collected in a personal digital assistant (PDA TungstenT3 USA) synchronised with the GPS which was used to recordthe trajectory and location of the survey track

Habitat suitability maps for guanacoWe selected eight potential predictors that summarised themost relevant environmental and landscape gradients as wellas anthropogenic features needed to assess which factors mightinfluence the regional distribution of guanacos (Table 1 forfurther details see Travaini et al 2007 Pedrana et al 2010)Distribution modelling requires defining units in whichpresence or absence is recorded For this purpose we overlaidthe surveyed tracks on top of a 1-km grid following the spatialresolution of NDVI data Then the original 457 sightingsregistered during field surveys were also overlaid and gridcells with at least one guanaco sighting were considered as apresence cell and all remaining cells traversed by census tracks

were considered as absences As cells with guanaco presence(n= 310) were outnumbered by absence cells (n= 15 606)we used a resampling scheme to obtain a presenceabsencebalanced sample (Travaini et al 2015) We randomly selected310 out of the 15 606 cells with guanaco absences repeatingthis procedure 100 times Each time we generated two datasetsa random sample of 80 of the cells for calibrating theoccurrence models and the remaining 20 for evaluating themodels (hereafter referred as the construction and evaluationdatasets)All the cellswith presencewere used in each repetitionwhile cells with absence were sampled without replacement

To identify which predictors were most likely to affectguanaco occurrence model fitting was done on the constructiondataset with Generalised Additive Models (GAMs Hastieand Tibshirani 1990) using a binomial error and a logit linkPredictors for the models were selected by a backwardndashforwardstepwise procedure (Sakamoto et al 1986) The AkaikeInformation Criterion (AIC) was used to retain a term(Burnham and Anderson 2002) We considered as competingmodels those for which the differences between AIC and theAIC of the best candidate model (the one with the smallestAIC) was D 2 (Burnham and Anderson 2002)

Model evaluation was made using the validation dataset bycomparing predicted and observed values using a threshold-independent metric ie the area under the curve (AUC) ofthe Receiver Operating Characteristic (ROC) plot which wascomputed for each of the 100 fitted models (Murtaugh 1996)The AUC ranges from 0 (when model discrimination isnot better than random) to 1 (perfect discriminatory ability)Predictive models were considered usable if AUC 07

Finally predictions of the resulting distribution models wereused to create habitat suitability maps (HSM) Predictions forthe entire study area were calculated using R version 322 (RDevelopment Core Team 2016) and transformed to probabilitymaps using IDRISI Selva GIS software (Clark Labs WorcesterMA USA) We categorised and mapped the HSM of guanacooccurrence probability in three classes (low lt033 medium033ndash066 and high gt066)

Table 1 Description of the variables used in thepredictive distributionmodels developed for guanaco in central Patagonia Argentina

NVDI Normalised Difference Vegetation Index

Predictor Predictor description

MeanNDVI Mean Normalised Difference Vegetation Indexcalculated by using the VGT-S10 product thatis a 10-day maximum composite value from theVEGETATION sensor on board of Spot-4 satellitefrom April 1999 to March 2005

SeasonMAX Month at which the NDVI reaches its annualmaximum value

CVNDVI Coefficient of variation of NDVIDistancestream Distance (km) to the nearest stream or riverAltitude Mean elevation in meters above sea level of the

focal cell obtained from the Shuttle RadarTopographic Mission Elevation (SRTM)

Slope Terrain slope in degrees in a 1 km pixelfrom the SRTM

Distanceurban Distance (km) to the nearest urban areaDistanceoil Distance (km) to the nearest oil camp

D Wildlife Research J Pedrana et al

Guanaco density estimationLine transect data were analysed using the DISTANCEprogram (version 60 Buckland et al 2001 Thomas et al2010) Estimates of guanaco densities were done by fitting adetection function from the perpendicular distances of guanacosto the survey line (Buckland et al 2001) We pooled distancedata across all transects within each HSM class of predictedguanaco occurrence to estimate the detection functions g(x) foreach social unit in order to attain the required minimum numberof observations (Buckland et al 2001) Although detectabilityvaries among transects the property of lsquopooling robustnessrsquoensures the reliability of the abundance estimates (Bucklandet al 2001 Thomas et al 2010) Pooling also helps to even outminor fluctuations in object density across surveyed strips andto even out chance fluctuations in object distribution (Fewsteret al 2005) The DISTANCE program allows for carefuldata exploration facilitating the identification of any outlyingobservation at extreme distances or sighting clumping andsuggests appropriate levels of truncation and grouping fordetection function fitting We used the smallest value of AICto choose the best competing model (Key function + adjustmentterms) Density estimates were calculated for each social unitand for all guanaco contacts pooled together and for each HSMcategory and for all of Chubut Considering the extent of eachHSM category density estimates were converted into estimatesof population size

Monitoring programWe estimated and simulated the sampling effort in terms ofnumber of 30 km transects of road survey needed to monitorguanaco population in Chubut and also to detect declines frompresent status in guanaco population using MONITOR (version1100 Gibbs and Ene 2010) We modelled the encounter ratewhich was calculated as the number of guanacos detected persampled kilometre over time and to generate detection ratesderived from route-regression analyses (Gibbs and Melvin1997) This simulation procedure is useful for evaluating thetrade-offs between sampling effort logistical constraints andpower to detect trends (Field et al 2005) To estimate powerwe supplied the programwith initial estimates of encounter ratesand its variance derived empirically from our own results Wearbitrarily selected 30 km as the length of sampling units Basedon our experience a 30-km-long transect usually takes abouttwo hours to complete the interval we used to exchangeobservers during field work to guarantee observer attention

Additionally in the study area random 30-km transects canbe selected from the vector data of road coverage more easilythan longer segments To select suitable sampling efforts wefixed the minimum acceptable power at (1-b) = 080 (Di Stefano2001) ie the highest probability of failing to detect a realtrend was 020 Type I error was fixed at 005 To define theeffect size component we selected two alternative scenariosboth highly probable given the guanaco current situationpreviously described The first scenario is under the hypothesisof a sustainable use of guanaco populations to detect early a 40decrease in the next six to 10 years An alternative scenariothe second is if the guanaco is considered as a pest species inorder to early detect a 40 increase in the next six and 10 yearsAs we had sampled each one of the 164 survey tracks oncerandom resampling without replacement of 30-km transectswas used to estimate the mean and the standard deviation ofguanaco encounter rates as a surrogate of spatial variabilityAs an estimate of power variability we estimated the standarddeviation from 20 simulations for each monitoring scenarioand 2000 iterations for each simulation

Results

We surveyed 7010 km almost 76 of available national andprovincial roads in Chubut We recorded 456 guanaco socialgroups which added up to 2893 individuals (Table 2) Almost20 of individuals were registered as belonging to breedinggroups 30 as non-breeding groups and 4 as solitaryindividuals (Table 2) The remaining 47 of individuals wereclassified as undetermined (Table 2) The high percentageof undetermined individuals resulted from our interest inminimising the chances of social group misclassification Weopted for considering family groups without juveniles orwith undetected juveniles as undetermined rather thanmisclassifying them as non-breeding family groups (Pedranaet al 2009) Family groups (n) averaged 8 (sd 322) individualswhere the mean number of females and chulengos per groupwere 5 (sd 268) and 2 (sd 117) respectively (Table 2)Non-breeding and undetermined groups averaged 28 (sd1581) and 6 (sd 268) individuals per group respectively(Table 2)

Habitat suitability maps for guanaco

Themost parsimonious GAMof guanaco presence incorporatedas predictors themeanNDVI altitude the distance to the nearesturban centre and to the closest oil field (Table S1 available

Table 2 Number of guanaco groups and individuals (within brackets) and mean group size recorded during one breeding season(NovemberndashDecember 2006) in the three areas defined by Habitat Suitability Maps (HSM) and for the total area covered in central

Patagonia Argentinasd standard deviation Und Undetermined groups

HSM classes Area Effort (km) Guanaco social groups (individuals) Total Mean group size (sd)(km2) ( transects) Solitary Breeding Non

BreedingUnd Breeding Females

per groupChulengosper group

NonBreeding

Und

Low 76 741 1824 (46) 14 6 (42) 5 (159) 13 (91) 38 (306) 7 (300) 5 (289) 1 (103) 32 (2272) 7 (438)Medium 68 145 3686 (82) 46 26 (216) 16 (444) 120 (710) 208 (1416) 8 (331) 5 (283) 2 (127) 28 (143) 6 (427)High 76 205 1500 (36) 65 35 (294) 10 (259) 100 (553) 210 (1171) 8 (325) 6 (258) 2 (121) 16 (2590) 6 (359)

Total 221 091 7010 (164) 125 67 (552) 31 (862) 233 (1354) 456 (2893) 8 (322) 5 (268) 2 (117) 28 (1581) 6 (268)

Guanaco abundance distribution and monitoring Wildlife Research E

as Supplementary Material to this paper) The probability ofguanaco occurrence decreased with primary productivity andwith mean altitude whereas it increased with the distance tothe nearest urban centre and oil field (Fig 2)

The mean ( se) AUC of the best model ranged from080 (002) to 082 (002) for the evaluation dataset(Table S1) These results indicate that models have anacceptable discrimination ability and are useful for predictingguanaco distribution

The HSM of guanaco occurrence (Fig 3) corresponding tothe predictions of the best model (Table S1) showed that the

probability of guanaco occurrence was positively associatedwith a gradient of increasing aridity from west to east Areaswith low probability of guanaco occurrence predominatedthroughout the western third of Chubut within the sub-Andean and occidental physiographic districts (Soriano 1956)from the transition from forests to short grasslands steppesand scrublands to the centre of the Province (Ares et al1990) From there the probability of guanaco occurrencesteeply increased eastward up to the coastline and theValdeacutes Peninsula throughout the driest lands with the mostxeromorphic floristic composition (Fig 3)

0

ndash4 ndash2

ndash10ndash10

0505

20

20

ndash5

1

ndash8

60 120 180

Mean NDVI

Par

tial e

ffect

Altitude Distance urban Distance oil

200 000100 000120 00060 000014007000 0

Fig 2 Partial effects of the predictors included in the most-parsimonious model about the variables influencing guanaco occurrence (Table S1)Dashed lines represent 95 confidence intervals for the mean effect NDVI Normalised Difference Vegetation Index

60deg W

30deg S

50deg S

100 km

Low

Medium

High

Sightings

N

Fig 3 Habitat suitability map of guanaco constructed from the most parsimonious model (Table S1) in central Patagonia Argentina Valuesrepresent the probability of guanaco contact in a 1-km cell and are categorised into three classes (low lt033 medium 033ndash066 high gt066)Guanaco sightings are indicated by circles Areas in white correspond to regions without predictions sea lakes forested areas or outside themodelrsquos environmental space

F Wildlife Research J Pedrana et al

Guanaco density estimation

Predictions of guanaco occurrence were not made for waterbodies forests and other habitats unsuitable for guanaco sothe total area for which abundance was estimated included221 091 km2 which represents 98 of the extent of Chubut

After considering competing detection functions thesmallest AIC value indicated that the hazard-rate key withthree-parameter cosine adjustment term should be selected forabundance estimation for each guanaco social group and one-parameter cosine adjustment term for abundance estimationat each HSM category of guanaco occurrence We truncateddistance data at 200m in all models in order to attain a robustline transect analysis after the elimination of outliers (Bucklandet al 2001) Our truncation distance closely follows g(x) = 015where g is the detection function and x is the perpendiculardistance

For the entire Chubut we estimated a total guanacopopulation of 657 082 individuals (95 CI 457 437 to944 059) which represents a mean density of 297 individualsper km2 (Table 3) Mean density varied considerably betweenguanaco breeding groups and non-breeding groups (048 and021 individuals per km2 respectively Table 3) In areasclassified as having low habitat suitability (HS) for guanacothe estimated density was 049 individuals per km2 whereasin the medium and high HS categories densities were 274and 393 individuals per km2 respectively The CV variedbetween 41 and 20 (Table 3) As expected mean densitypositively increased in relation to the HS categories (Table 3)

Guanaco monitoring

Assuming annual monitoring surveys simulations showed thatat least thirty 30-km transects are needed to detect confidentlyany decrease (sustainable use scenario) or increase (alternativescenario) over the next 6 or 10 years in the low HS category(Fig 4a) Required power (080) was attained for most scenariosand HS categories based on a quite reasonable sampling effortHowever power over 090 was attained only with 10 30-kmtransects for any monitoring scenario for the medium andhigh HS categories (Fig 4b c)

Discussion

The sustainable use of any wildlife species requires a thoroughknowledgeof its distribution and abundanceAlso it is important

to develop a scientifically soundmonitoring protocol to evaluatepopulation trends under management Here we provide thesethree elements for the guanaco a valuable wildlife species thatinhabits one of the Neotropical ecosystems most severelyaffected by desertification

Guanaco distribution in Chubut showed a clear geographicalpattern with a gradient from low occurrence probability in thewest and to higher values in the east (Fig 3) The areas withhigher habitat suitability for the guanaco are coincident withthePatagonian semi-deserts (Paruelo et al 1998) and the ʻCentralDistrictrsquo delimited by Soriano (1956) and described by Leoacutenet al (1998) This is the less productive area in Chubut althoughit could be considered one of the most productive semi-deserts(450 kg handash1 yearndash1) in the world (Paruelo et al 1998) Theguanaco distribution pattern observed here has a certainsimilarity to those described for Tierra del Fuego (Raedeke1979 1982) and Santa Cruz (Pedrana et al 2010) provinceswhere higher habitat suitability for the species was found in theless productive and more arid areas in remarkable coincidencewith those areas where sheep husbandry is absent or sheepoccur at a very low density (Pedrana et al 2010) The westof Chubut where guanaco density is low is characterised bya mosaic of shrub-grass steppes with an annual net primaryproductivity higher than that of the eastern semi-deserts (650 kghandash1 yearndash1 Paruelo et al 1998) Habitat suitability modelshighlighted the western region as a low-quality habitat forguanaco Throughout this area the higher productivity hasfavoured more intense and sustained sheep ranching (Baldiet al 2001) Probably as a consequence of interference fromthis activity guanacos have become scarce although they seemto have remained in sheep-free less productive areas Guanacooccurrence in Chubut was positively associated with lowproductivity areas characterised by low NDVI values and lowelevation and far away from oil exploitation and urban areasOur results can be interpreted as a guanaco preference inducedby anthropogenic factors for this arid habitat however it alsoseems plausible that this pattern may not reflect a true habitatpreference but an indirect response to exogenous factors(competition with sheep and a response to direct persecutionby ranchers or poaching) We also found a positive associationof guanaco with remote areas far away from centres of humaninfluence such as cities and oil camps which is consistentwith a guanaco avoidance of areas that could have intensivepoaching These relationships could suggest the hypothesis that

Table 3 Estimates of guanaco density (D) and total population size (N) for each social group and in each class of Habitat Suitability Map (HSM)in central Patagonia Argentina

Upper and lower limits of 95 confidence of intervals for density estimates are given CV coefficient of variation of density estimates

Group kmndash2 D (ind kmndash2) Lower limit Upper limit CV N

Social Groups Breeding 008 048 023 098 3834 105 239Non Breeding 004 021 011 037 3071 45 324Solitary 015 085 046 157 3194 187 706Undetermined 025 144 088 236 2547 318 813Pooled 051 297 270 427 1861 657 082

HSM classes Low 009 049 022 110 4102 38 217Medium 047 274 165 456 2622 186 649High 068 393 265 582 2016 299 105Pooled 046 269 192 377 1736 523 971

Guanaco abundance distribution and monitoring Wildlife Research G

human activities have a direct and negative effect on guanacooccurrence suggesting either a behavioural avoidance of siteswhere mortality risk by humans is high and recurrent or atrue decline of guanaco local population due to overhuntingAccording to Pedrana et al (2010) the same habitat selectionpattern was found in Santa Cruz where high-suitability habitatfor guanacos corresponded with the less productive areas

The extent to which the strong inverse correlation betweenvegetation productivity and guanaco density depends onavoidance by guanacos of high sheep density areas theirillegal culling by ranchers or because they actively selectmore arid vegetation communities remains an unresolved

question (Pedrana et al 2010) One way to address thisquestion is estimating guanaco density and modelling habitatin protected areas with different vegetation communitieswhere sheep and poaching has been excluded (or reduced)Comparing guanaco population and density estimates betweenChubut and other published guanaco figures is challengingsince stratification of the study area survey effort andestimation methods are quite different It has been debatedthat road transects are not an adequate tool for estimatingguanaco population size and occurrence (Schroeder et al2018) because roads rarely cross a region in a random patternand because guanacos could be attracted or repelled by roadsIn Patagonia besides vehicle survey there are two alternativemethods that could be used to calculate guanaco abundancebut it is unclear that these methods would allow extensivedistance sampling over large regions In the case ofhorseback distance sampling has several disadvantagesrugged terrain could hardly be sampled surveying areas faraway from human habitation would be unfeasible (with theconsequent bias) and it would be difficult to cover largeregions during a short session Aircraft distance samplingcould solve some of these problems but has its own uniquedisadvantages such as unreliable counting of individuals ordetermination of social units Although sampling away fromroads would be desirable what we propose here is a cost-effective method of density estimation because aerial surveysare much more expensive than road surveys (Travaini et al2007) and hence rarely undertaken in practice Indeed nopublished guanaco population estimates exist for large areasbased on aerial surveys We covered three times more distancethan the total length of aerial transects reported by Amaya et al(2001) Testing empirically which of the two methods givesmore accurate and cost-effective estimates for guanaco densityand population size in large areas is an important topic to beaddressed in future studies

According to the International Union for Conservationof Nature (IUCN) most of the worldrsquos remaining guanacosoccur in Argentina Argentina is estimated to be home to12ndash19million guanacos (between 81ndash86 of the entirepopulation) most of them concentrated in the Patagonianregion (Baldi et al 2016) Even though its distribution rangeis extended in almost all of Argentine Patagonia guanacopopulations seem to be more fragmented in the northern andcentral provinces (Chubut Riacuteo Negro and Neuqueacuten) comparedwith the southern ones (Santa Cruz and Tierra del Fuego Baldiet al 2016) Guanaco local density estimations for Patagoniaranged from 040 individuals per km2 in some areas of Tierradel Fuego (Montes et al 2000) to more than 10 individuals perkm2 in Neuqueacuten (Rey et al 2009) Riacuteo Negro (Rey 2010) andChubut (Saba and Battro 1987) Our guanaco density estimatescan be comparedwith those in Santa Cruz which were producedusing the same methods (Pedrana et al 2010 Travaini et al2015) In the present study we found that guanaco densitiesin Chubut (2ndash4 individuals per km2) were lower than thoseestimated in Santa Cruz (3ndash7 individuals per km2 table 2 inTravaini et al 2015) While both provinces are similar in size(Chubut 221 866 km2 Santa Cruz 223 117 km2) we estimatethat the Chubut guanaco population is only half of that in SantaCruz (Travaini et al 2015)

100 (a)

(b)

(c)

095

090

085

080

075

070

Pow

er plusmn

sd

065

060

055

050

100

098

096

094

092

090

40 change over thenext 6 years

Annualsurveys Bi-annual

surveys Tri-annualsurveys

Annualsurveys

Bi-annualsurveys

Tri-annualsurveys

Annualsurveys

Annualsurveys

Annualsurveys

Annualsurveys

Bi-annualsurveys

Bi-annualsurveys

Bi-annualsurveys

Bi-annualsurveys

Tri-annualsurveys

Tri-annualsurveys

Tri-annualsurveys

Tri-annualsurveys

40 change over thenext 10 years

088

086

084

100

098

096

094

092

090

088

086

084Monitoring scenarios

Fig 4 Power (sd) for the detection of a decrease in the Chubut guanaco population (solid diamonds) or an increase (solid squares) under different scenarios 40 change over the next 6 or 10 years and survey frequency(a) low-density strata and thirty 30-km transects of road survey (b) medium-density and 10 30-km transects of road survey and (c) high-density strataand 10 30-km transects of road survey

H Wildlife Research J Pedrana et al

Finally if guanaco populations are to be sustained adaptivemanagement would require the monitoring of these populationsfor the early detection of any increase or unwanted decrease intheir numbers Traveling 600 km every 3 years is adequate todetect quite large changes within the two most dense guanacostrata in Chubut where any extractive enterprise might startBased on our experience 30-km transects should be distributedthroughout the strata area which we estimate would take sixcomplete working days by two or three technicians (Travainiet al 2015) However if the low HS category is included inthe follow-up the effort needed would be greater about ninefull-time working days per year and 15 fieldwork days every3 years Ideally an articulated association between Chubutand its neighbouring Santa Cruz could result in the bestbenefits both for the monitoring and conservation of guanacowild populations Moreover these two provinces should evendevelop common strategies for the sustainable use andregulation in the trade of guanaco products

Conservation and management implications

Patagonia holds some 700 000 km2 of grasslands and scrublandsthat still maintain healthy guanaco populations (Baldi et al2010) despite having been deeply modified by humanactivities Both ecosystem structure and functioning werealtered by the introduction of domestic herbivores at thebeginning of the 20th Century (Aguiar et al 1996 Bisigatoand Bertiller 1997) Grazing mining (Radovani et al 2014)and more recently global climate change (Pinto et al 2008)are human-related disturbances acting at different spatial andtemporal scales that have degraded Patagonian vegetationCurrent conditions are very different from the ones encounteredby the first European settlers who were unable to maintaina traditional sheep ranching system that was environmentallysustainable and economically profitable (Texeira and Paruelo2006) Traditional sheep ranching involved keeping sheepgrazing in extensive paddocks (requiring several thousandhectares) with limited attention a few occasions a year fordeworming or shearing (Garciacutea Brea et al 2010) Restoringthe Patagonian shrub-steppe as an ecologically functionalecosystem (Chardonnet et al 2002) would benefit fromefficient and modernised sheep husbandry complemented byactivities such as the sustainable use of wildlife and ecotourismIn this framework guanaco could play a role as an importanteconomic resource for landowners and local communities(Franklin et al 1997) In the absence of alternatives economicincentives could be the most effective driver for the conservationof the species (Fritz and Franklin 1994)

The greatest utility of our results is that they provide anexcellent overview of guanaco distribution and abundance in2006 and the big picture for the entire Chubut from whichthe government could start planning the sustainable use of thespecies Considering our HSM it is possible to identify theareas of the territory with higher densities where it will beeasier to select ranches with adequate guanaco abundance(Baldi et al 2010) to develop profitable herding and shearingexperiences Our HSM for Chubut together with otherenvironmental and human spatial data (eg land use roadcover and cadastral) should be used to generate a decision

tool that guarantees informed sustainable use of the speciesFinally our results need to be implemented in accordance withenvironmental education programs to change the perceptionof ranchers in Chubut province on native wildlife

Conflicts of interest

The authors declare no conflicts of interest

Acknowledgements

This work was primarily funded by the BBVA Foundation through a grantunder the Conservation Biology Program to A Rodriacuteguez and also bythe Agencia Nacional de Promocioacuten Cientiacutefica y Tecnoloacutegica (PICTON 30723) Additional support was provided by Universidad Nacional dela Patagonia Austral CONICET (PEI-6065) and CONAE

References

Aguiar M R Paruelo J M Sala O E and Laurenroth W K (1996)Ecosystem consequences of plant functional types changes in semiaridgrassland Journal of Vegetation Science 7 381ndash390 doi1023073236281

Amaya JN von Thuumlngen J and de Lamo DA (2001) Relevamiento ydistribucioacutendeguanacosen laPatagoniaComunicacioacutenTeacutecnicano109(Instituto Nacional de Tecnologiacutea Agropecuaria Bariloche Argentina)[In Spanish]

Ares J O BeeskowAM BertillerM B Rostagno CM IrisarriM PAnchorena J Defosseacute G E and Meroni C (1990) Structuraland dynamic characteristics of overgrazed grasslands of northernPatagonia Argentina In lsquoManaged Grasslands Regional Studiesrsquo(Ed A Breymeyer) pp 149ndash175 (Elsevier Amsterdam)

Baldi R Albon S D and Elston D A (2001) Guanacos and sheepevidence for continuing competition in arid Patagonia Oecologia 129561ndash570 doi101007s004420100770

Baldi R Novaro A Funes M Walker S Ferrando P Failla M andCarmanchahi P (2010) Guanacomanagement in Patagonian rangelandsa conservation opportunity on the brink of collapse In lsquoConservingWildlife while Maintaining Livestock in Semi-Arid Ecosystemsrsquo (EdsJ Du Toit R Kock and J Deutsch) pp 266ndash290 (Blackwell PublishingOxford UK)

Baldi R Acebes P Cueacutellar E Funes M Hoces D Puig S andFranklin W L (2016) Lama guanicoe The IUCN Red List ofThreatened Species 2016 Available at httpwwwiucnredlistorg[Accessed 6 April 2018]

BarrosVR ScianBV andMattioHF (1979)Camposdeprecipitacioacutende la provincia de Chubut (1931ndash1960) Geoacta 10 175ndash192

Bisigato A J andBertillerMB (1997) Grazing effects on patchy drylandvegetation in northern Patagonia Journal of Arid Environments 36639ndash653 doi101006jare19960247

Borrelli P and Cibils A (2005) Rural depopulation and grasslandmanagement in Patagonia In lsquoGrasslands Development OpportunitiesPerspectivesrsquo (Eds S G Reynolds and J Frame) pp 461ndash488 (FAOEnfield NH)

Buckland S T Anderson D R Burnham K P Laake J L BorchersD L and Thomas L (2001) lsquoIntroduction to Distance SamplingEstimating Abundance of Biological Populationsrsquo (Oxford UniversityPress Oxford)

Burnham K P and Anderson D R (2002) lsquoModel Selection andMultimodel Inference a Practical Information-theoretic Approachrsquo(Springer-Verlag New York)

Caro J Zapata S C Zanoacuten J I Rodriacuteguez A and Travaini A (2017)Ganaderiacutea ovina y usos alternativos del suelo en la Patagonia australArgentina Multequina (Mendoza) 26 33ndash50

Guanaco abundance distribution and monitoring Wildlife Research I

Guanaco density estimationLine transect data were analysed using the DISTANCEprogram (version 60 Buckland et al 2001 Thomas et al2010) Estimates of guanaco densities were done by fitting adetection function from the perpendicular distances of guanacosto the survey line (Buckland et al 2001) We pooled distancedata across all transects within each HSM class of predictedguanaco occurrence to estimate the detection functions g(x) foreach social unit in order to attain the required minimum numberof observations (Buckland et al 2001) Although detectabilityvaries among transects the property of lsquopooling robustnessrsquoensures the reliability of the abundance estimates (Bucklandet al 2001 Thomas et al 2010) Pooling also helps to even outminor fluctuations in object density across surveyed strips andto even out chance fluctuations in object distribution (Fewsteret al 2005) The DISTANCE program allows for carefuldata exploration facilitating the identification of any outlyingobservation at extreme distances or sighting clumping andsuggests appropriate levels of truncation and grouping fordetection function fitting We used the smallest value of AICto choose the best competing model (Key function + adjustmentterms) Density estimates were calculated for each social unitand for all guanaco contacts pooled together and for each HSMcategory and for all of Chubut Considering the extent of eachHSM category density estimates were converted into estimatesof population size

Monitoring programWe estimated and simulated the sampling effort in terms ofnumber of 30 km transects of road survey needed to monitorguanaco population in Chubut and also to detect declines frompresent status in guanaco population using MONITOR (version1100 Gibbs and Ene 2010) We modelled the encounter ratewhich was calculated as the number of guanacos detected persampled kilometre over time and to generate detection ratesderived from route-regression analyses (Gibbs and Melvin1997) This simulation procedure is useful for evaluating thetrade-offs between sampling effort logistical constraints andpower to detect trends (Field et al 2005) To estimate powerwe supplied the programwith initial estimates of encounter ratesand its variance derived empirically from our own results Wearbitrarily selected 30 km as the length of sampling units Basedon our experience a 30-km-long transect usually takes abouttwo hours to complete the interval we used to exchangeobservers during field work to guarantee observer attention

Additionally in the study area random 30-km transects canbe selected from the vector data of road coverage more easilythan longer segments To select suitable sampling efforts wefixed the minimum acceptable power at (1-b) = 080 (Di Stefano2001) ie the highest probability of failing to detect a realtrend was 020 Type I error was fixed at 005 To define theeffect size component we selected two alternative scenariosboth highly probable given the guanaco current situationpreviously described The first scenario is under the hypothesisof a sustainable use of guanaco populations to detect early a 40decrease in the next six to 10 years An alternative scenariothe second is if the guanaco is considered as a pest species inorder to early detect a 40 increase in the next six and 10 yearsAs we had sampled each one of the 164 survey tracks oncerandom resampling without replacement of 30-km transectswas used to estimate the mean and the standard deviation ofguanaco encounter rates as a surrogate of spatial variabilityAs an estimate of power variability we estimated the standarddeviation from 20 simulations for each monitoring scenarioand 2000 iterations for each simulation

Results

We surveyed 7010 km almost 76 of available national andprovincial roads in Chubut We recorded 456 guanaco socialgroups which added up to 2893 individuals (Table 2) Almost20 of individuals were registered as belonging to breedinggroups 30 as non-breeding groups and 4 as solitaryindividuals (Table 2) The remaining 47 of individuals wereclassified as undetermined (Table 2) The high percentageof undetermined individuals resulted from our interest inminimising the chances of social group misclassification Weopted for considering family groups without juveniles orwith undetected juveniles as undetermined rather thanmisclassifying them as non-breeding family groups (Pedranaet al 2009) Family groups (n) averaged 8 (sd 322) individualswhere the mean number of females and chulengos per groupwere 5 (sd 268) and 2 (sd 117) respectively (Table 2)Non-breeding and undetermined groups averaged 28 (sd1581) and 6 (sd 268) individuals per group respectively(Table 2)

Habitat suitability maps for guanaco

Themost parsimonious GAMof guanaco presence incorporatedas predictors themeanNDVI altitude the distance to the nearesturban centre and to the closest oil field (Table S1 available

Table 2 Number of guanaco groups and individuals (within brackets) and mean group size recorded during one breeding season(NovemberndashDecember 2006) in the three areas defined by Habitat Suitability Maps (HSM) and for the total area covered in central

Patagonia Argentinasd standard deviation Und Undetermined groups

HSM classes Area Effort (km) Guanaco social groups (individuals) Total Mean group size (sd)(km2) ( transects) Solitary Breeding Non

BreedingUnd Breeding Females

per groupChulengosper group

NonBreeding

Und

Low 76 741 1824 (46) 14 6 (42) 5 (159) 13 (91) 38 (306) 7 (300) 5 (289) 1 (103) 32 (2272) 7 (438)Medium 68 145 3686 (82) 46 26 (216) 16 (444) 120 (710) 208 (1416) 8 (331) 5 (283) 2 (127) 28 (143) 6 (427)High 76 205 1500 (36) 65 35 (294) 10 (259) 100 (553) 210 (1171) 8 (325) 6 (258) 2 (121) 16 (2590) 6 (359)

Total 221 091 7010 (164) 125 67 (552) 31 (862) 233 (1354) 456 (2893) 8 (322) 5 (268) 2 (117) 28 (1581) 6 (268)

Guanaco abundance distribution and monitoring Wildlife Research E

as Supplementary Material to this paper) The probability ofguanaco occurrence decreased with primary productivity andwith mean altitude whereas it increased with the distance tothe nearest urban centre and oil field (Fig 2)

The mean ( se) AUC of the best model ranged from080 (002) to 082 (002) for the evaluation dataset(Table S1) These results indicate that models have anacceptable discrimination ability and are useful for predictingguanaco distribution

The HSM of guanaco occurrence (Fig 3) corresponding tothe predictions of the best model (Table S1) showed that the

probability of guanaco occurrence was positively associatedwith a gradient of increasing aridity from west to east Areaswith low probability of guanaco occurrence predominatedthroughout the western third of Chubut within the sub-Andean and occidental physiographic districts (Soriano 1956)from the transition from forests to short grasslands steppesand scrublands to the centre of the Province (Ares et al1990) From there the probability of guanaco occurrencesteeply increased eastward up to the coastline and theValdeacutes Peninsula throughout the driest lands with the mostxeromorphic floristic composition (Fig 3)

0

ndash4 ndash2

ndash10ndash10

0505

20

20

ndash5

1

ndash8

60 120 180

Mean NDVI

Par

tial e

ffect

Altitude Distance urban Distance oil

200 000100 000120 00060 000014007000 0

Fig 2 Partial effects of the predictors included in the most-parsimonious model about the variables influencing guanaco occurrence (Table S1)Dashed lines represent 95 confidence intervals for the mean effect NDVI Normalised Difference Vegetation Index

60deg W

30deg S

50deg S

100 km

Low

Medium

High

Sightings

N

Fig 3 Habitat suitability map of guanaco constructed from the most parsimonious model (Table S1) in central Patagonia Argentina Valuesrepresent the probability of guanaco contact in a 1-km cell and are categorised into three classes (low lt033 medium 033ndash066 high gt066)Guanaco sightings are indicated by circles Areas in white correspond to regions without predictions sea lakes forested areas or outside themodelrsquos environmental space

F Wildlife Research J Pedrana et al

Guanaco density estimation

Predictions of guanaco occurrence were not made for waterbodies forests and other habitats unsuitable for guanaco sothe total area for which abundance was estimated included221 091 km2 which represents 98 of the extent of Chubut

After considering competing detection functions thesmallest AIC value indicated that the hazard-rate key withthree-parameter cosine adjustment term should be selected forabundance estimation for each guanaco social group and one-parameter cosine adjustment term for abundance estimationat each HSM category of guanaco occurrence We truncateddistance data at 200m in all models in order to attain a robustline transect analysis after the elimination of outliers (Bucklandet al 2001) Our truncation distance closely follows g(x) = 015where g is the detection function and x is the perpendiculardistance

For the entire Chubut we estimated a total guanacopopulation of 657 082 individuals (95 CI 457 437 to944 059) which represents a mean density of 297 individualsper km2 (Table 3) Mean density varied considerably betweenguanaco breeding groups and non-breeding groups (048 and021 individuals per km2 respectively Table 3) In areasclassified as having low habitat suitability (HS) for guanacothe estimated density was 049 individuals per km2 whereasin the medium and high HS categories densities were 274and 393 individuals per km2 respectively The CV variedbetween 41 and 20 (Table 3) As expected mean densitypositively increased in relation to the HS categories (Table 3)

Guanaco monitoring

Assuming annual monitoring surveys simulations showed thatat least thirty 30-km transects are needed to detect confidentlyany decrease (sustainable use scenario) or increase (alternativescenario) over the next 6 or 10 years in the low HS category(Fig 4a) Required power (080) was attained for most scenariosand HS categories based on a quite reasonable sampling effortHowever power over 090 was attained only with 10 30-kmtransects for any monitoring scenario for the medium andhigh HS categories (Fig 4b c)

Discussion

The sustainable use of any wildlife species requires a thoroughknowledgeof its distribution and abundanceAlso it is important

to develop a scientifically soundmonitoring protocol to evaluatepopulation trends under management Here we provide thesethree elements for the guanaco a valuable wildlife species thatinhabits one of the Neotropical ecosystems most severelyaffected by desertification

Guanaco distribution in Chubut showed a clear geographicalpattern with a gradient from low occurrence probability in thewest and to higher values in the east (Fig 3) The areas withhigher habitat suitability for the guanaco are coincident withthePatagonian semi-deserts (Paruelo et al 1998) and the ʻCentralDistrictrsquo delimited by Soriano (1956) and described by Leoacutenet al (1998) This is the less productive area in Chubut althoughit could be considered one of the most productive semi-deserts(450 kg handash1 yearndash1) in the world (Paruelo et al 1998) Theguanaco distribution pattern observed here has a certainsimilarity to those described for Tierra del Fuego (Raedeke1979 1982) and Santa Cruz (Pedrana et al 2010) provinceswhere higher habitat suitability for the species was found in theless productive and more arid areas in remarkable coincidencewith those areas where sheep husbandry is absent or sheepoccur at a very low density (Pedrana et al 2010) The westof Chubut where guanaco density is low is characterised bya mosaic of shrub-grass steppes with an annual net primaryproductivity higher than that of the eastern semi-deserts (650 kghandash1 yearndash1 Paruelo et al 1998) Habitat suitability modelshighlighted the western region as a low-quality habitat forguanaco Throughout this area the higher productivity hasfavoured more intense and sustained sheep ranching (Baldiet al 2001) Probably as a consequence of interference fromthis activity guanacos have become scarce although they seemto have remained in sheep-free less productive areas Guanacooccurrence in Chubut was positively associated with lowproductivity areas characterised by low NDVI values and lowelevation and far away from oil exploitation and urban areasOur results can be interpreted as a guanaco preference inducedby anthropogenic factors for this arid habitat however it alsoseems plausible that this pattern may not reflect a true habitatpreference but an indirect response to exogenous factors(competition with sheep and a response to direct persecutionby ranchers or poaching) We also found a positive associationof guanaco with remote areas far away from centres of humaninfluence such as cities and oil camps which is consistentwith a guanaco avoidance of areas that could have intensivepoaching These relationships could suggest the hypothesis that

Table 3 Estimates of guanaco density (D) and total population size (N) for each social group and in each class of Habitat Suitability Map (HSM)in central Patagonia Argentina

Upper and lower limits of 95 confidence of intervals for density estimates are given CV coefficient of variation of density estimates

Group kmndash2 D (ind kmndash2) Lower limit Upper limit CV N

Social Groups Breeding 008 048 023 098 3834 105 239Non Breeding 004 021 011 037 3071 45 324Solitary 015 085 046 157 3194 187 706Undetermined 025 144 088 236 2547 318 813Pooled 051 297 270 427 1861 657 082

HSM classes Low 009 049 022 110 4102 38 217Medium 047 274 165 456 2622 186 649High 068 393 265 582 2016 299 105Pooled 046 269 192 377 1736 523 971

Guanaco abundance distribution and monitoring Wildlife Research G

human activities have a direct and negative effect on guanacooccurrence suggesting either a behavioural avoidance of siteswhere mortality risk by humans is high and recurrent or atrue decline of guanaco local population due to overhuntingAccording to Pedrana et al (2010) the same habitat selectionpattern was found in Santa Cruz where high-suitability habitatfor guanacos corresponded with the less productive areas

The extent to which the strong inverse correlation betweenvegetation productivity and guanaco density depends onavoidance by guanacos of high sheep density areas theirillegal culling by ranchers or because they actively selectmore arid vegetation communities remains an unresolved

question (Pedrana et al 2010) One way to address thisquestion is estimating guanaco density and modelling habitatin protected areas with different vegetation communitieswhere sheep and poaching has been excluded (or reduced)Comparing guanaco population and density estimates betweenChubut and other published guanaco figures is challengingsince stratification of the study area survey effort andestimation methods are quite different It has been debatedthat road transects are not an adequate tool for estimatingguanaco population size and occurrence (Schroeder et al2018) because roads rarely cross a region in a random patternand because guanacos could be attracted or repelled by roadsIn Patagonia besides vehicle survey there are two alternativemethods that could be used to calculate guanaco abundancebut it is unclear that these methods would allow extensivedistance sampling over large regions In the case ofhorseback distance sampling has several disadvantagesrugged terrain could hardly be sampled surveying areas faraway from human habitation would be unfeasible (with theconsequent bias) and it would be difficult to cover largeregions during a short session Aircraft distance samplingcould solve some of these problems but has its own uniquedisadvantages such as unreliable counting of individuals ordetermination of social units Although sampling away fromroads would be desirable what we propose here is a cost-effective method of density estimation because aerial surveysare much more expensive than road surveys (Travaini et al2007) and hence rarely undertaken in practice Indeed nopublished guanaco population estimates exist for large areasbased on aerial surveys We covered three times more distancethan the total length of aerial transects reported by Amaya et al(2001) Testing empirically which of the two methods givesmore accurate and cost-effective estimates for guanaco densityand population size in large areas is an important topic to beaddressed in future studies

According to the International Union for Conservationof Nature (IUCN) most of the worldrsquos remaining guanacosoccur in Argentina Argentina is estimated to be home to12ndash19million guanacos (between 81ndash86 of the entirepopulation) most of them concentrated in the Patagonianregion (Baldi et al 2016) Even though its distribution rangeis extended in almost all of Argentine Patagonia guanacopopulations seem to be more fragmented in the northern andcentral provinces (Chubut Riacuteo Negro and Neuqueacuten) comparedwith the southern ones (Santa Cruz and Tierra del Fuego Baldiet al 2016) Guanaco local density estimations for Patagoniaranged from 040 individuals per km2 in some areas of Tierradel Fuego (Montes et al 2000) to more than 10 individuals perkm2 in Neuqueacuten (Rey et al 2009) Riacuteo Negro (Rey 2010) andChubut (Saba and Battro 1987) Our guanaco density estimatescan be comparedwith those in Santa Cruz which were producedusing the same methods (Pedrana et al 2010 Travaini et al2015) In the present study we found that guanaco densitiesin Chubut (2ndash4 individuals per km2) were lower than thoseestimated in Santa Cruz (3ndash7 individuals per km2 table 2 inTravaini et al 2015) While both provinces are similar in size(Chubut 221 866 km2 Santa Cruz 223 117 km2) we estimatethat the Chubut guanaco population is only half of that in SantaCruz (Travaini et al 2015)

100 (a)

(b)

(c)

095

090

085

080

075

070

Pow

er plusmn

sd

065

060

055

050

100

098

096

094

092

090

40 change over thenext 6 years

Annualsurveys Bi-annual

surveys Tri-annualsurveys

Annualsurveys

Bi-annualsurveys

Tri-annualsurveys

Annualsurveys

Annualsurveys

Annualsurveys

Annualsurveys

Bi-annualsurveys

Bi-annualsurveys

Bi-annualsurveys

Bi-annualsurveys

Tri-annualsurveys

Tri-annualsurveys

Tri-annualsurveys

Tri-annualsurveys

40 change over thenext 10 years

088

086

084

100

098

096

094

092

090

088

086

084Monitoring scenarios

Fig 4 Power (sd) for the detection of a decrease in the Chubut guanaco population (solid diamonds) or an increase (solid squares) under different scenarios 40 change over the next 6 or 10 years and survey frequency(a) low-density strata and thirty 30-km transects of road survey (b) medium-density and 10 30-km transects of road survey and (c) high-density strataand 10 30-km transects of road survey

H Wildlife Research J Pedrana et al

Finally if guanaco populations are to be sustained adaptivemanagement would require the monitoring of these populationsfor the early detection of any increase or unwanted decrease intheir numbers Traveling 600 km every 3 years is adequate todetect quite large changes within the two most dense guanacostrata in Chubut where any extractive enterprise might startBased on our experience 30-km transects should be distributedthroughout the strata area which we estimate would take sixcomplete working days by two or three technicians (Travainiet al 2015) However if the low HS category is included inthe follow-up the effort needed would be greater about ninefull-time working days per year and 15 fieldwork days every3 years Ideally an articulated association between Chubutand its neighbouring Santa Cruz could result in the bestbenefits both for the monitoring and conservation of guanacowild populations Moreover these two provinces should evendevelop common strategies for the sustainable use andregulation in the trade of guanaco products

Conservation and management implications

Patagonia holds some 700 000 km2 of grasslands and scrublandsthat still maintain healthy guanaco populations (Baldi et al2010) despite having been deeply modified by humanactivities Both ecosystem structure and functioning werealtered by the introduction of domestic herbivores at thebeginning of the 20th Century (Aguiar et al 1996 Bisigatoand Bertiller 1997) Grazing mining (Radovani et al 2014)and more recently global climate change (Pinto et al 2008)are human-related disturbances acting at different spatial andtemporal scales that have degraded Patagonian vegetationCurrent conditions are very different from the ones encounteredby the first European settlers who were unable to maintaina traditional sheep ranching system that was environmentallysustainable and economically profitable (Texeira and Paruelo2006) Traditional sheep ranching involved keeping sheepgrazing in extensive paddocks (requiring several thousandhectares) with limited attention a few occasions a year fordeworming or shearing (Garciacutea Brea et al 2010) Restoringthe Patagonian shrub-steppe as an ecologically functionalecosystem (Chardonnet et al 2002) would benefit fromefficient and modernised sheep husbandry complemented byactivities such as the sustainable use of wildlife and ecotourismIn this framework guanaco could play a role as an importanteconomic resource for landowners and local communities(Franklin et al 1997) In the absence of alternatives economicincentives could be the most effective driver for the conservationof the species (Fritz and Franklin 1994)

The greatest utility of our results is that they provide anexcellent overview of guanaco distribution and abundance in2006 and the big picture for the entire Chubut from whichthe government could start planning the sustainable use of thespecies Considering our HSM it is possible to identify theareas of the territory with higher densities where it will beeasier to select ranches with adequate guanaco abundance(Baldi et al 2010) to develop profitable herding and shearingexperiences Our HSM for Chubut together with otherenvironmental and human spatial data (eg land use roadcover and cadastral) should be used to generate a decision

tool that guarantees informed sustainable use of the speciesFinally our results need to be implemented in accordance withenvironmental education programs to change the perceptionof ranchers in Chubut province on native wildlife

Conflicts of interest

The authors declare no conflicts of interest

Acknowledgements

This work was primarily funded by the BBVA Foundation through a grantunder the Conservation Biology Program to A Rodriacuteguez and also bythe Agencia Nacional de Promocioacuten Cientiacutefica y Tecnoloacutegica (PICTON 30723) Additional support was provided by Universidad Nacional dela Patagonia Austral CONICET (PEI-6065) and CONAE

References

Aguiar M R Paruelo J M Sala O E and Laurenroth W K (1996)Ecosystem consequences of plant functional types changes in semiaridgrassland Journal of Vegetation Science 7 381ndash390 doi1023073236281

Amaya JN von Thuumlngen J and de Lamo DA (2001) Relevamiento ydistribucioacutendeguanacosen laPatagoniaComunicacioacutenTeacutecnicano109(Instituto Nacional de Tecnologiacutea Agropecuaria Bariloche Argentina)[In Spanish]

Ares J O BeeskowAM BertillerM B Rostagno CM IrisarriM PAnchorena J Defosseacute G E and Meroni C (1990) Structuraland dynamic characteristics of overgrazed grasslands of northernPatagonia Argentina In lsquoManaged Grasslands Regional Studiesrsquo(Ed A Breymeyer) pp 149ndash175 (Elsevier Amsterdam)

Baldi R Albon S D and Elston D A (2001) Guanacos and sheepevidence for continuing competition in arid Patagonia Oecologia 129561ndash570 doi101007s004420100770

Baldi R Novaro A Funes M Walker S Ferrando P Failla M andCarmanchahi P (2010) Guanacomanagement in Patagonian rangelandsa conservation opportunity on the brink of collapse In lsquoConservingWildlife while Maintaining Livestock in Semi-Arid Ecosystemsrsquo (EdsJ Du Toit R Kock and J Deutsch) pp 266ndash290 (Blackwell PublishingOxford UK)

Baldi R Acebes P Cueacutellar E Funes M Hoces D Puig S andFranklin W L (2016) Lama guanicoe The IUCN Red List ofThreatened Species 2016 Available at httpwwwiucnredlistorg[Accessed 6 April 2018]

BarrosVR ScianBV andMattioHF (1979)Camposdeprecipitacioacutende la provincia de Chubut (1931ndash1960) Geoacta 10 175ndash192

Bisigato A J andBertillerMB (1997) Grazing effects on patchy drylandvegetation in northern Patagonia Journal of Arid Environments 36639ndash653 doi101006jare19960247

Borrelli P and Cibils A (2005) Rural depopulation and grasslandmanagement in Patagonia In lsquoGrasslands Development OpportunitiesPerspectivesrsquo (Eds S G Reynolds and J Frame) pp 461ndash488 (FAOEnfield NH)

Buckland S T Anderson D R Burnham K P Laake J L BorchersD L and Thomas L (2001) lsquoIntroduction to Distance SamplingEstimating Abundance of Biological Populationsrsquo (Oxford UniversityPress Oxford)

Burnham K P and Anderson D R (2002) lsquoModel Selection andMultimodel Inference a Practical Information-theoretic Approachrsquo(Springer-Verlag New York)

Caro J Zapata S C Zanoacuten J I Rodriacuteguez A and Travaini A (2017)Ganaderiacutea ovina y usos alternativos del suelo en la Patagonia australArgentina Multequina (Mendoza) 26 33ndash50

Guanaco abundance distribution and monitoring Wildlife Research I

as Supplementary Material to this paper) The probability ofguanaco occurrence decreased with primary productivity andwith mean altitude whereas it increased with the distance tothe nearest urban centre and oil field (Fig 2)

The mean ( se) AUC of the best model ranged from080 (002) to 082 (002) for the evaluation dataset(Table S1) These results indicate that models have anacceptable discrimination ability and are useful for predictingguanaco distribution

The HSM of guanaco occurrence (Fig 3) corresponding tothe predictions of the best model (Table S1) showed that the

probability of guanaco occurrence was positively associatedwith a gradient of increasing aridity from west to east Areaswith low probability of guanaco occurrence predominatedthroughout the western third of Chubut within the sub-Andean and occidental physiographic districts (Soriano 1956)from the transition from forests to short grasslands steppesand scrublands to the centre of the Province (Ares et al1990) From there the probability of guanaco occurrencesteeply increased eastward up to the coastline and theValdeacutes Peninsula throughout the driest lands with the mostxeromorphic floristic composition (Fig 3)

0

ndash4 ndash2

ndash10ndash10

0505

20

20

ndash5

1

ndash8

60 120 180

Mean NDVI

Par

tial e

ffect

Altitude Distance urban Distance oil

200 000100 000120 00060 000014007000 0

Fig 2 Partial effects of the predictors included in the most-parsimonious model about the variables influencing guanaco occurrence (Table S1)Dashed lines represent 95 confidence intervals for the mean effect NDVI Normalised Difference Vegetation Index

60deg W

30deg S

50deg S

100 km

Low

Medium

High

Sightings

N

Fig 3 Habitat suitability map of guanaco constructed from the most parsimonious model (Table S1) in central Patagonia Argentina Valuesrepresent the probability of guanaco contact in a 1-km cell and are categorised into three classes (low lt033 medium 033ndash066 high gt066)Guanaco sightings are indicated by circles Areas in white correspond to regions without predictions sea lakes forested areas or outside themodelrsquos environmental space

F Wildlife Research J Pedrana et al

Guanaco density estimation

Predictions of guanaco occurrence were not made for waterbodies forests and other habitats unsuitable for guanaco sothe total area for which abundance was estimated included221 091 km2 which represents 98 of the extent of Chubut

After considering competing detection functions thesmallest AIC value indicated that the hazard-rate key withthree-parameter cosine adjustment term should be selected forabundance estimation for each guanaco social group and one-parameter cosine adjustment term for abundance estimationat each HSM category of guanaco occurrence We truncateddistance data at 200m in all models in order to attain a robustline transect analysis after the elimination of outliers (Bucklandet al 2001) Our truncation distance closely follows g(x) = 015where g is the detection function and x is the perpendiculardistance

For the entire Chubut we estimated a total guanacopopulation of 657 082 individuals (95 CI 457 437 to944 059) which represents a mean density of 297 individualsper km2 (Table 3) Mean density varied considerably betweenguanaco breeding groups and non-breeding groups (048 and021 individuals per km2 respectively Table 3) In areasclassified as having low habitat suitability (HS) for guanacothe estimated density was 049 individuals per km2 whereasin the medium and high HS categories densities were 274and 393 individuals per km2 respectively The CV variedbetween 41 and 20 (Table 3) As expected mean densitypositively increased in relation to the HS categories (Table 3)

Guanaco monitoring

Assuming annual monitoring surveys simulations showed thatat least thirty 30-km transects are needed to detect confidentlyany decrease (sustainable use scenario) or increase (alternativescenario) over the next 6 or 10 years in the low HS category(Fig 4a) Required power (080) was attained for most scenariosand HS categories based on a quite reasonable sampling effortHowever power over 090 was attained only with 10 30-kmtransects for any monitoring scenario for the medium andhigh HS categories (Fig 4b c)

Discussion

The sustainable use of any wildlife species requires a thoroughknowledgeof its distribution and abundanceAlso it is important

to develop a scientifically soundmonitoring protocol to evaluatepopulation trends under management Here we provide thesethree elements for the guanaco a valuable wildlife species thatinhabits one of the Neotropical ecosystems most severelyaffected by desertification

Guanaco distribution in Chubut showed a clear geographicalpattern with a gradient from low occurrence probability in thewest and to higher values in the east (Fig 3) The areas withhigher habitat suitability for the guanaco are coincident withthePatagonian semi-deserts (Paruelo et al 1998) and the ʻCentralDistrictrsquo delimited by Soriano (1956) and described by Leoacutenet al (1998) This is the less productive area in Chubut althoughit could be considered one of the most productive semi-deserts(450 kg handash1 yearndash1) in the world (Paruelo et al 1998) Theguanaco distribution pattern observed here has a certainsimilarity to those described for Tierra del Fuego (Raedeke1979 1982) and Santa Cruz (Pedrana et al 2010) provinceswhere higher habitat suitability for the species was found in theless productive and more arid areas in remarkable coincidencewith those areas where sheep husbandry is absent or sheepoccur at a very low density (Pedrana et al 2010) The westof Chubut where guanaco density is low is characterised bya mosaic of shrub-grass steppes with an annual net primaryproductivity higher than that of the eastern semi-deserts (650 kghandash1 yearndash1 Paruelo et al 1998) Habitat suitability modelshighlighted the western region as a low-quality habitat forguanaco Throughout this area the higher productivity hasfavoured more intense and sustained sheep ranching (Baldiet al 2001) Probably as a consequence of interference fromthis activity guanacos have become scarce although they seemto have remained in sheep-free less productive areas Guanacooccurrence in Chubut was positively associated with lowproductivity areas characterised by low NDVI values and lowelevation and far away from oil exploitation and urban areasOur results can be interpreted as a guanaco preference inducedby anthropogenic factors for this arid habitat however it alsoseems plausible that this pattern may not reflect a true habitatpreference but an indirect response to exogenous factors(competition with sheep and a response to direct persecutionby ranchers or poaching) We also found a positive associationof guanaco with remote areas far away from centres of humaninfluence such as cities and oil camps which is consistentwith a guanaco avoidance of areas that could have intensivepoaching These relationships could suggest the hypothesis that

Table 3 Estimates of guanaco density (D) and total population size (N) for each social group and in each class of Habitat Suitability Map (HSM)in central Patagonia Argentina

Upper and lower limits of 95 confidence of intervals for density estimates are given CV coefficient of variation of density estimates

Group kmndash2 D (ind kmndash2) Lower limit Upper limit CV N

Social Groups Breeding 008 048 023 098 3834 105 239Non Breeding 004 021 011 037 3071 45 324Solitary 015 085 046 157 3194 187 706Undetermined 025 144 088 236 2547 318 813Pooled 051 297 270 427 1861 657 082

HSM classes Low 009 049 022 110 4102 38 217Medium 047 274 165 456 2622 186 649High 068 393 265 582 2016 299 105Pooled 046 269 192 377 1736 523 971

Guanaco abundance distribution and monitoring Wildlife Research G

human activities have a direct and negative effect on guanacooccurrence suggesting either a behavioural avoidance of siteswhere mortality risk by humans is high and recurrent or atrue decline of guanaco local population due to overhuntingAccording to Pedrana et al (2010) the same habitat selectionpattern was found in Santa Cruz where high-suitability habitatfor guanacos corresponded with the less productive areas

The extent to which the strong inverse correlation betweenvegetation productivity and guanaco density depends onavoidance by guanacos of high sheep density areas theirillegal culling by ranchers or because they actively selectmore arid vegetation communities remains an unresolved

question (Pedrana et al 2010) One way to address thisquestion is estimating guanaco density and modelling habitatin protected areas with different vegetation communitieswhere sheep and poaching has been excluded (or reduced)Comparing guanaco population and density estimates betweenChubut and other published guanaco figures is challengingsince stratification of the study area survey effort andestimation methods are quite different It has been debatedthat road transects are not an adequate tool for estimatingguanaco population size and occurrence (Schroeder et al2018) because roads rarely cross a region in a random patternand because guanacos could be attracted or repelled by roadsIn Patagonia besides vehicle survey there are two alternativemethods that could be used to calculate guanaco abundancebut it is unclear that these methods would allow extensivedistance sampling over large regions In the case ofhorseback distance sampling has several disadvantagesrugged terrain could hardly be sampled surveying areas faraway from human habitation would be unfeasible (with theconsequent bias) and it would be difficult to cover largeregions during a short session Aircraft distance samplingcould solve some of these problems but has its own uniquedisadvantages such as unreliable counting of individuals ordetermination of social units Although sampling away fromroads would be desirable what we propose here is a cost-effective method of density estimation because aerial surveysare much more expensive than road surveys (Travaini et al2007) and hence rarely undertaken in practice Indeed nopublished guanaco population estimates exist for large areasbased on aerial surveys We covered three times more distancethan the total length of aerial transects reported by Amaya et al(2001) Testing empirically which of the two methods givesmore accurate and cost-effective estimates for guanaco densityand population size in large areas is an important topic to beaddressed in future studies

According to the International Union for Conservationof Nature (IUCN) most of the worldrsquos remaining guanacosoccur in Argentina Argentina is estimated to be home to12ndash19million guanacos (between 81ndash86 of the entirepopulation) most of them concentrated in the Patagonianregion (Baldi et al 2016) Even though its distribution rangeis extended in almost all of Argentine Patagonia guanacopopulations seem to be more fragmented in the northern andcentral provinces (Chubut Riacuteo Negro and Neuqueacuten) comparedwith the southern ones (Santa Cruz and Tierra del Fuego Baldiet al 2016) Guanaco local density estimations for Patagoniaranged from 040 individuals per km2 in some areas of Tierradel Fuego (Montes et al 2000) to more than 10 individuals perkm2 in Neuqueacuten (Rey et al 2009) Riacuteo Negro (Rey 2010) andChubut (Saba and Battro 1987) Our guanaco density estimatescan be comparedwith those in Santa Cruz which were producedusing the same methods (Pedrana et al 2010 Travaini et al2015) In the present study we found that guanaco densitiesin Chubut (2ndash4 individuals per km2) were lower than thoseestimated in Santa Cruz (3ndash7 individuals per km2 table 2 inTravaini et al 2015) While both provinces are similar in size(Chubut 221 866 km2 Santa Cruz 223 117 km2) we estimatethat the Chubut guanaco population is only half of that in SantaCruz (Travaini et al 2015)

100 (a)

(b)

(c)

095

090

085

080

075

070

Pow

er plusmn

sd

065

060

055

050

100

098

096

094

092

090

40 change over thenext 6 years

Annualsurveys Bi-annual

surveys Tri-annualsurveys

Annualsurveys

Bi-annualsurveys

Tri-annualsurveys

Annualsurveys

Annualsurveys

Annualsurveys

Annualsurveys

Bi-annualsurveys

Bi-annualsurveys

Bi-annualsurveys

Bi-annualsurveys

Tri-annualsurveys

Tri-annualsurveys

Tri-annualsurveys

Tri-annualsurveys

40 change over thenext 10 years

088

086

084

100

098

096

094

092

090

088

086

084Monitoring scenarios

Fig 4 Power (sd) for the detection of a decrease in the Chubut guanaco population (solid diamonds) or an increase (solid squares) under different scenarios 40 change over the next 6 or 10 years and survey frequency(a) low-density strata and thirty 30-km transects of road survey (b) medium-density and 10 30-km transects of road survey and (c) high-density strataand 10 30-km transects of road survey

H Wildlife Research J Pedrana et al

Finally if guanaco populations are to be sustained adaptivemanagement would require the monitoring of these populationsfor the early detection of any increase or unwanted decrease intheir numbers Traveling 600 km every 3 years is adequate todetect quite large changes within the two most dense guanacostrata in Chubut where any extractive enterprise might startBased on our experience 30-km transects should be distributedthroughout the strata area which we estimate would take sixcomplete working days by two or three technicians (Travainiet al 2015) However if the low HS category is included inthe follow-up the effort needed would be greater about ninefull-time working days per year and 15 fieldwork days every3 years Ideally an articulated association between Chubutand its neighbouring Santa Cruz could result in the bestbenefits both for the monitoring and conservation of guanacowild populations Moreover these two provinces should evendevelop common strategies for the sustainable use andregulation in the trade of guanaco products

Conservation and management implications

Patagonia holds some 700 000 km2 of grasslands and scrublandsthat still maintain healthy guanaco populations (Baldi et al2010) despite having been deeply modified by humanactivities Both ecosystem structure and functioning werealtered by the introduction of domestic herbivores at thebeginning of the 20th Century (Aguiar et al 1996 Bisigatoand Bertiller 1997) Grazing mining (Radovani et al 2014)and more recently global climate change (Pinto et al 2008)are human-related disturbances acting at different spatial andtemporal scales that have degraded Patagonian vegetationCurrent conditions are very different from the ones encounteredby the first European settlers who were unable to maintaina traditional sheep ranching system that was environmentallysustainable and economically profitable (Texeira and Paruelo2006) Traditional sheep ranching involved keeping sheepgrazing in extensive paddocks (requiring several thousandhectares) with limited attention a few occasions a year fordeworming or shearing (Garciacutea Brea et al 2010) Restoringthe Patagonian shrub-steppe as an ecologically functionalecosystem (Chardonnet et al 2002) would benefit fromefficient and modernised sheep husbandry complemented byactivities such as the sustainable use of wildlife and ecotourismIn this framework guanaco could play a role as an importanteconomic resource for landowners and local communities(Franklin et al 1997) In the absence of alternatives economicincentives could be the most effective driver for the conservationof the species (Fritz and Franklin 1994)

The greatest utility of our results is that they provide anexcellent overview of guanaco distribution and abundance in2006 and the big picture for the entire Chubut from whichthe government could start planning the sustainable use of thespecies Considering our HSM it is possible to identify theareas of the territory with higher densities where it will beeasier to select ranches with adequate guanaco abundance(Baldi et al 2010) to develop profitable herding and shearingexperiences Our HSM for Chubut together with otherenvironmental and human spatial data (eg land use roadcover and cadastral) should be used to generate a decision

tool that guarantees informed sustainable use of the speciesFinally our results need to be implemented in accordance withenvironmental education programs to change the perceptionof ranchers in Chubut province on native wildlife

Conflicts of interest

The authors declare no conflicts of interest

Acknowledgements

This work was primarily funded by the BBVA Foundation through a grantunder the Conservation Biology Program to A Rodriacuteguez and also bythe Agencia Nacional de Promocioacuten Cientiacutefica y Tecnoloacutegica (PICTON 30723) Additional support was provided by Universidad Nacional dela Patagonia Austral CONICET (PEI-6065) and CONAE

References

Aguiar M R Paruelo J M Sala O E and Laurenroth W K (1996)Ecosystem consequences of plant functional types changes in semiaridgrassland Journal of Vegetation Science 7 381ndash390 doi1023073236281

Amaya JN von Thuumlngen J and de Lamo DA (2001) Relevamiento ydistribucioacutendeguanacosen laPatagoniaComunicacioacutenTeacutecnicano109(Instituto Nacional de Tecnologiacutea Agropecuaria Bariloche Argentina)[In Spanish]

Ares J O BeeskowAM BertillerM B Rostagno CM IrisarriM PAnchorena J Defosseacute G E and Meroni C (1990) Structuraland dynamic characteristics of overgrazed grasslands of northernPatagonia Argentina In lsquoManaged Grasslands Regional Studiesrsquo(Ed A Breymeyer) pp 149ndash175 (Elsevier Amsterdam)

Baldi R Albon S D and Elston D A (2001) Guanacos and sheepevidence for continuing competition in arid Patagonia Oecologia 129561ndash570 doi101007s004420100770

Baldi R Novaro A Funes M Walker S Ferrando P Failla M andCarmanchahi P (2010) Guanacomanagement in Patagonian rangelandsa conservation opportunity on the brink of collapse In lsquoConservingWildlife while Maintaining Livestock in Semi-Arid Ecosystemsrsquo (EdsJ Du Toit R Kock and J Deutsch) pp 266ndash290 (Blackwell PublishingOxford UK)

Baldi R Acebes P Cueacutellar E Funes M Hoces D Puig S andFranklin W L (2016) Lama guanicoe The IUCN Red List ofThreatened Species 2016 Available at httpwwwiucnredlistorg[Accessed 6 April 2018]

BarrosVR ScianBV andMattioHF (1979)Camposdeprecipitacioacutende la provincia de Chubut (1931ndash1960) Geoacta 10 175ndash192

Bisigato A J andBertillerMB (1997) Grazing effects on patchy drylandvegetation in northern Patagonia Journal of Arid Environments 36639ndash653 doi101006jare19960247

Borrelli P and Cibils A (2005) Rural depopulation and grasslandmanagement in Patagonia In lsquoGrasslands Development OpportunitiesPerspectivesrsquo (Eds S G Reynolds and J Frame) pp 461ndash488 (FAOEnfield NH)

Buckland S T Anderson D R Burnham K P Laake J L BorchersD L and Thomas L (2001) lsquoIntroduction to Distance SamplingEstimating Abundance of Biological Populationsrsquo (Oxford UniversityPress Oxford)

Burnham K P and Anderson D R (2002) lsquoModel Selection andMultimodel Inference a Practical Information-theoretic Approachrsquo(Springer-Verlag New York)

Caro J Zapata S C Zanoacuten J I Rodriacuteguez A and Travaini A (2017)Ganaderiacutea ovina y usos alternativos del suelo en la Patagonia australArgentina Multequina (Mendoza) 26 33ndash50

Guanaco abundance distribution and monitoring Wildlife Research I

Guanaco density estimation

Predictions of guanaco occurrence were not made for waterbodies forests and other habitats unsuitable for guanaco sothe total area for which abundance was estimated included221 091 km2 which represents 98 of the extent of Chubut

After considering competing detection functions thesmallest AIC value indicated that the hazard-rate key withthree-parameter cosine adjustment term should be selected forabundance estimation for each guanaco social group and one-parameter cosine adjustment term for abundance estimationat each HSM category of guanaco occurrence We truncateddistance data at 200m in all models in order to attain a robustline transect analysis after the elimination of outliers (Bucklandet al 2001) Our truncation distance closely follows g(x) = 015where g is the detection function and x is the perpendiculardistance

For the entire Chubut we estimated a total guanacopopulation of 657 082 individuals (95 CI 457 437 to944 059) which represents a mean density of 297 individualsper km2 (Table 3) Mean density varied considerably betweenguanaco breeding groups and non-breeding groups (048 and021 individuals per km2 respectively Table 3) In areasclassified as having low habitat suitability (HS) for guanacothe estimated density was 049 individuals per km2 whereasin the medium and high HS categories densities were 274and 393 individuals per km2 respectively The CV variedbetween 41 and 20 (Table 3) As expected mean densitypositively increased in relation to the HS categories (Table 3)

Guanaco monitoring

Assuming annual monitoring surveys simulations showed thatat least thirty 30-km transects are needed to detect confidentlyany decrease (sustainable use scenario) or increase (alternativescenario) over the next 6 or 10 years in the low HS category(Fig 4a) Required power (080) was attained for most scenariosand HS categories based on a quite reasonable sampling effortHowever power over 090 was attained only with 10 30-kmtransects for any monitoring scenario for the medium andhigh HS categories (Fig 4b c)

Discussion

The sustainable use of any wildlife species requires a thoroughknowledgeof its distribution and abundanceAlso it is important

to develop a scientifically soundmonitoring protocol to evaluatepopulation trends under management Here we provide thesethree elements for the guanaco a valuable wildlife species thatinhabits one of the Neotropical ecosystems most severelyaffected by desertification

Guanaco distribution in Chubut showed a clear geographicalpattern with a gradient from low occurrence probability in thewest and to higher values in the east (Fig 3) The areas withhigher habitat suitability for the guanaco are coincident withthePatagonian semi-deserts (Paruelo et al 1998) and the ʻCentralDistrictrsquo delimited by Soriano (1956) and described by Leoacutenet al (1998) This is the less productive area in Chubut althoughit could be considered one of the most productive semi-deserts(450 kg handash1 yearndash1) in the world (Paruelo et al 1998) Theguanaco distribution pattern observed here has a certainsimilarity to those described for Tierra del Fuego (Raedeke1979 1982) and Santa Cruz (Pedrana et al 2010) provinceswhere higher habitat suitability for the species was found in theless productive and more arid areas in remarkable coincidencewith those areas where sheep husbandry is absent or sheepoccur at a very low density (Pedrana et al 2010) The westof Chubut where guanaco density is low is characterised bya mosaic of shrub-grass steppes with an annual net primaryproductivity higher than that of the eastern semi-deserts (650 kghandash1 yearndash1 Paruelo et al 1998) Habitat suitability modelshighlighted the western region as a low-quality habitat forguanaco Throughout this area the higher productivity hasfavoured more intense and sustained sheep ranching (Baldiet al 2001) Probably as a consequence of interference fromthis activity guanacos have become scarce although they seemto have remained in sheep-free less productive areas Guanacooccurrence in Chubut was positively associated with lowproductivity areas characterised by low NDVI values and lowelevation and far away from oil exploitation and urban areasOur results can be interpreted as a guanaco preference inducedby anthropogenic factors for this arid habitat however it alsoseems plausible that this pattern may not reflect a true habitatpreference but an indirect response to exogenous factors(competition with sheep and a response to direct persecutionby ranchers or poaching) We also found a positive associationof guanaco with remote areas far away from centres of humaninfluence such as cities and oil camps which is consistentwith a guanaco avoidance of areas that could have intensivepoaching These relationships could suggest the hypothesis that

Table 3 Estimates of guanaco density (D) and total population size (N) for each social group and in each class of Habitat Suitability Map (HSM)in central Patagonia Argentina

Upper and lower limits of 95 confidence of intervals for density estimates are given CV coefficient of variation of density estimates

Group kmndash2 D (ind kmndash2) Lower limit Upper limit CV N

Social Groups Breeding 008 048 023 098 3834 105 239Non Breeding 004 021 011 037 3071 45 324Solitary 015 085 046 157 3194 187 706Undetermined 025 144 088 236 2547 318 813Pooled 051 297 270 427 1861 657 082

HSM classes Low 009 049 022 110 4102 38 217Medium 047 274 165 456 2622 186 649High 068 393 265 582 2016 299 105Pooled 046 269 192 377 1736 523 971

Guanaco abundance distribution and monitoring Wildlife Research G

human activities have a direct and negative effect on guanacooccurrence suggesting either a behavioural avoidance of siteswhere mortality risk by humans is high and recurrent or atrue decline of guanaco local population due to overhuntingAccording to Pedrana et al (2010) the same habitat selectionpattern was found in Santa Cruz where high-suitability habitatfor guanacos corresponded with the less productive areas

The extent to which the strong inverse correlation betweenvegetation productivity and guanaco density depends onavoidance by guanacos of high sheep density areas theirillegal culling by ranchers or because they actively selectmore arid vegetation communities remains an unresolved

question (Pedrana et al 2010) One way to address thisquestion is estimating guanaco density and modelling habitatin protected areas with different vegetation communitieswhere sheep and poaching has been excluded (or reduced)Comparing guanaco population and density estimates betweenChubut and other published guanaco figures is challengingsince stratification of the study area survey effort andestimation methods are quite different It has been debatedthat road transects are not an adequate tool for estimatingguanaco population size and occurrence (Schroeder et al2018) because roads rarely cross a region in a random patternand because guanacos could be attracted or repelled by roadsIn Patagonia besides vehicle survey there are two alternativemethods that could be used to calculate guanaco abundancebut it is unclear that these methods would allow extensivedistance sampling over large regions In the case ofhorseback distance sampling has several disadvantagesrugged terrain could hardly be sampled surveying areas faraway from human habitation would be unfeasible (with theconsequent bias) and it would be difficult to cover largeregions during a short session Aircraft distance samplingcould solve some of these problems but has its own uniquedisadvantages such as unreliable counting of individuals ordetermination of social units Although sampling away fromroads would be desirable what we propose here is a cost-effective method of density estimation because aerial surveysare much more expensive than road surveys (Travaini et al2007) and hence rarely undertaken in practice Indeed nopublished guanaco population estimates exist for large areasbased on aerial surveys We covered three times more distancethan the total length of aerial transects reported by Amaya et al(2001) Testing empirically which of the two methods givesmore accurate and cost-effective estimates for guanaco densityand population size in large areas is an important topic to beaddressed in future studies

According to the International Union for Conservationof Nature (IUCN) most of the worldrsquos remaining guanacosoccur in Argentina Argentina is estimated to be home to12ndash19million guanacos (between 81ndash86 of the entirepopulation) most of them concentrated in the Patagonianregion (Baldi et al 2016) Even though its distribution rangeis extended in almost all of Argentine Patagonia guanacopopulations seem to be more fragmented in the northern andcentral provinces (Chubut Riacuteo Negro and Neuqueacuten) comparedwith the southern ones (Santa Cruz and Tierra del Fuego Baldiet al 2016) Guanaco local density estimations for Patagoniaranged from 040 individuals per km2 in some areas of Tierradel Fuego (Montes et al 2000) to more than 10 individuals perkm2 in Neuqueacuten (Rey et al 2009) Riacuteo Negro (Rey 2010) andChubut (Saba and Battro 1987) Our guanaco density estimatescan be comparedwith those in Santa Cruz which were producedusing the same methods (Pedrana et al 2010 Travaini et al2015) In the present study we found that guanaco densitiesin Chubut (2ndash4 individuals per km2) were lower than thoseestimated in Santa Cruz (3ndash7 individuals per km2 table 2 inTravaini et al 2015) While both provinces are similar in size(Chubut 221 866 km2 Santa Cruz 223 117 km2) we estimatethat the Chubut guanaco population is only half of that in SantaCruz (Travaini et al 2015)

100 (a)

(b)

(c)

095

090

085

080

075

070

Pow

er plusmn

sd

065

060

055

050

100

098

096

094

092

090

40 change over thenext 6 years

Annualsurveys Bi-annual

surveys Tri-annualsurveys

Annualsurveys

Bi-annualsurveys

Tri-annualsurveys

Annualsurveys

Annualsurveys

Annualsurveys

Annualsurveys

Bi-annualsurveys

Bi-annualsurveys

Bi-annualsurveys

Bi-annualsurveys

Tri-annualsurveys

Tri-annualsurveys

Tri-annualsurveys

Tri-annualsurveys

40 change over thenext 10 years

088

086

084

100

098

096

094

092

090

088

086

084Monitoring scenarios

Fig 4 Power (sd) for the detection of a decrease in the Chubut guanaco population (solid diamonds) or an increase (solid squares) under different scenarios 40 change over the next 6 or 10 years and survey frequency(a) low-density strata and thirty 30-km transects of road survey (b) medium-density and 10 30-km transects of road survey and (c) high-density strataand 10 30-km transects of road survey

H Wildlife Research J Pedrana et al

Finally if guanaco populations are to be sustained adaptivemanagement would require the monitoring of these populationsfor the early detection of any increase or unwanted decrease intheir numbers Traveling 600 km every 3 years is adequate todetect quite large changes within the two most dense guanacostrata in Chubut where any extractive enterprise might startBased on our experience 30-km transects should be distributedthroughout the strata area which we estimate would take sixcomplete working days by two or three technicians (Travainiet al 2015) However if the low HS category is included inthe follow-up the effort needed would be greater about ninefull-time working days per year and 15 fieldwork days every3 years Ideally an articulated association between Chubutand its neighbouring Santa Cruz could result in the bestbenefits both for the monitoring and conservation of guanacowild populations Moreover these two provinces should evendevelop common strategies for the sustainable use andregulation in the trade of guanaco products

Conservation and management implications

Patagonia holds some 700 000 km2 of grasslands and scrublandsthat still maintain healthy guanaco populations (Baldi et al2010) despite having been deeply modified by humanactivities Both ecosystem structure and functioning werealtered by the introduction of domestic herbivores at thebeginning of the 20th Century (Aguiar et al 1996 Bisigatoand Bertiller 1997) Grazing mining (Radovani et al 2014)and more recently global climate change (Pinto et al 2008)are human-related disturbances acting at different spatial andtemporal scales that have degraded Patagonian vegetationCurrent conditions are very different from the ones encounteredby the first European settlers who were unable to maintaina traditional sheep ranching system that was environmentallysustainable and economically profitable (Texeira and Paruelo2006) Traditional sheep ranching involved keeping sheepgrazing in extensive paddocks (requiring several thousandhectares) with limited attention a few occasions a year fordeworming or shearing (Garciacutea Brea et al 2010) Restoringthe Patagonian shrub-steppe as an ecologically functionalecosystem (Chardonnet et al 2002) would benefit fromefficient and modernised sheep husbandry complemented byactivities such as the sustainable use of wildlife and ecotourismIn this framework guanaco could play a role as an importanteconomic resource for landowners and local communities(Franklin et al 1997) In the absence of alternatives economicincentives could be the most effective driver for the conservationof the species (Fritz and Franklin 1994)

The greatest utility of our results is that they provide anexcellent overview of guanaco distribution and abundance in2006 and the big picture for the entire Chubut from whichthe government could start planning the sustainable use of thespecies Considering our HSM it is possible to identify theareas of the territory with higher densities where it will beeasier to select ranches with adequate guanaco abundance(Baldi et al 2010) to develop profitable herding and shearingexperiences Our HSM for Chubut together with otherenvironmental and human spatial data (eg land use roadcover and cadastral) should be used to generate a decision

tool that guarantees informed sustainable use of the speciesFinally our results need to be implemented in accordance withenvironmental education programs to change the perceptionof ranchers in Chubut province on native wildlife

Conflicts of interest

The authors declare no conflicts of interest

Acknowledgements

This work was primarily funded by the BBVA Foundation through a grantunder the Conservation Biology Program to A Rodriacuteguez and also bythe Agencia Nacional de Promocioacuten Cientiacutefica y Tecnoloacutegica (PICTON 30723) Additional support was provided by Universidad Nacional dela Patagonia Austral CONICET (PEI-6065) and CONAE

References

Aguiar M R Paruelo J M Sala O E and Laurenroth W K (1996)Ecosystem consequences of plant functional types changes in semiaridgrassland Journal of Vegetation Science 7 381ndash390 doi1023073236281

Amaya JN von Thuumlngen J and de Lamo DA (2001) Relevamiento ydistribucioacutendeguanacosen laPatagoniaComunicacioacutenTeacutecnicano109(Instituto Nacional de Tecnologiacutea Agropecuaria Bariloche Argentina)[In Spanish]

Ares J O BeeskowAM BertillerM B Rostagno CM IrisarriM PAnchorena J Defosseacute G E and Meroni C (1990) Structuraland dynamic characteristics of overgrazed grasslands of northernPatagonia Argentina In lsquoManaged Grasslands Regional Studiesrsquo(Ed A Breymeyer) pp 149ndash175 (Elsevier Amsterdam)

Baldi R Albon S D and Elston D A (2001) Guanacos and sheepevidence for continuing competition in arid Patagonia Oecologia 129561ndash570 doi101007s004420100770

Baldi R Novaro A Funes M Walker S Ferrando P Failla M andCarmanchahi P (2010) Guanacomanagement in Patagonian rangelandsa conservation opportunity on the brink of collapse In lsquoConservingWildlife while Maintaining Livestock in Semi-Arid Ecosystemsrsquo (EdsJ Du Toit R Kock and J Deutsch) pp 266ndash290 (Blackwell PublishingOxford UK)

Baldi R Acebes P Cueacutellar E Funes M Hoces D Puig S andFranklin W L (2016) Lama guanicoe The IUCN Red List ofThreatened Species 2016 Available at httpwwwiucnredlistorg[Accessed 6 April 2018]

BarrosVR ScianBV andMattioHF (1979)Camposdeprecipitacioacutende la provincia de Chubut (1931ndash1960) Geoacta 10 175ndash192

Bisigato A J andBertillerMB (1997) Grazing effects on patchy drylandvegetation in northern Patagonia Journal of Arid Environments 36639ndash653 doi101006jare19960247

Borrelli P and Cibils A (2005) Rural depopulation and grasslandmanagement in Patagonia In lsquoGrasslands Development OpportunitiesPerspectivesrsquo (Eds S G Reynolds and J Frame) pp 461ndash488 (FAOEnfield NH)

Buckland S T Anderson D R Burnham K P Laake J L BorchersD L and Thomas L (2001) lsquoIntroduction to Distance SamplingEstimating Abundance of Biological Populationsrsquo (Oxford UniversityPress Oxford)

Burnham K P and Anderson D R (2002) lsquoModel Selection andMultimodel Inference a Practical Information-theoretic Approachrsquo(Springer-Verlag New York)

Caro J Zapata S C Zanoacuten J I Rodriacuteguez A and Travaini A (2017)Ganaderiacutea ovina y usos alternativos del suelo en la Patagonia australArgentina Multequina (Mendoza) 26 33ndash50

Guanaco abundance distribution and monitoring Wildlife Research I

human activities have a direct and negative effect on guanacooccurrence suggesting either a behavioural avoidance of siteswhere mortality risk by humans is high and recurrent or atrue decline of guanaco local population due to overhuntingAccording to Pedrana et al (2010) the same habitat selectionpattern was found in Santa Cruz where high-suitability habitatfor guanacos corresponded with the less productive areas

The extent to which the strong inverse correlation betweenvegetation productivity and guanaco density depends onavoidance by guanacos of high sheep density areas theirillegal culling by ranchers or because they actively selectmore arid vegetation communities remains an unresolved

question (Pedrana et al 2010) One way to address thisquestion is estimating guanaco density and modelling habitatin protected areas with different vegetation communitieswhere sheep and poaching has been excluded (or reduced)Comparing guanaco population and density estimates betweenChubut and other published guanaco figures is challengingsince stratification of the study area survey effort andestimation methods are quite different It has been debatedthat road transects are not an adequate tool for estimatingguanaco population size and occurrence (Schroeder et al2018) because roads rarely cross a region in a random patternand because guanacos could be attracted or repelled by roadsIn Patagonia besides vehicle survey there are two alternativemethods that could be used to calculate guanaco abundancebut it is unclear that these methods would allow extensivedistance sampling over large regions In the case ofhorseback distance sampling has several disadvantagesrugged terrain could hardly be sampled surveying areas faraway from human habitation would be unfeasible (with theconsequent bias) and it would be difficult to cover largeregions during a short session Aircraft distance samplingcould solve some of these problems but has its own uniquedisadvantages such as unreliable counting of individuals ordetermination of social units Although sampling away fromroads would be desirable what we propose here is a cost-effective method of density estimation because aerial surveysare much more expensive than road surveys (Travaini et al2007) and hence rarely undertaken in practice Indeed nopublished guanaco population estimates exist for large areasbased on aerial surveys We covered three times more distancethan the total length of aerial transects reported by Amaya et al(2001) Testing empirically which of the two methods givesmore accurate and cost-effective estimates for guanaco densityand population size in large areas is an important topic to beaddressed in future studies

According to the International Union for Conservationof Nature (IUCN) most of the worldrsquos remaining guanacosoccur in Argentina Argentina is estimated to be home to12ndash19million guanacos (between 81ndash86 of the entirepopulation) most of them concentrated in the Patagonianregion (Baldi et al 2016) Even though its distribution rangeis extended in almost all of Argentine Patagonia guanacopopulations seem to be more fragmented in the northern andcentral provinces (Chubut Riacuteo Negro and Neuqueacuten) comparedwith the southern ones (Santa Cruz and Tierra del Fuego Baldiet al 2016) Guanaco local density estimations for Patagoniaranged from 040 individuals per km2 in some areas of Tierradel Fuego (Montes et al 2000) to more than 10 individuals perkm2 in Neuqueacuten (Rey et al 2009) Riacuteo Negro (Rey 2010) andChubut (Saba and Battro 1987) Our guanaco density estimatescan be comparedwith those in Santa Cruz which were producedusing the same methods (Pedrana et al 2010 Travaini et al2015) In the present study we found that guanaco densitiesin Chubut (2ndash4 individuals per km2) were lower than thoseestimated in Santa Cruz (3ndash7 individuals per km2 table 2 inTravaini et al 2015) While both provinces are similar in size(Chubut 221 866 km2 Santa Cruz 223 117 km2) we estimatethat the Chubut guanaco population is only half of that in SantaCruz (Travaini et al 2015)

100 (a)

(b)

(c)

095

090

085

080

075

070

Pow

er plusmn

sd

065

060

055

050

100

098

096

094

092

090

40 change over thenext 6 years

Annualsurveys Bi-annual

surveys Tri-annualsurveys

Annualsurveys

Bi-annualsurveys

Tri-annualsurveys

Annualsurveys

Annualsurveys

Annualsurveys

Annualsurveys

Bi-annualsurveys

Bi-annualsurveys

Bi-annualsurveys

Bi-annualsurveys

Tri-annualsurveys

Tri-annualsurveys

Tri-annualsurveys

Tri-annualsurveys

40 change over thenext 10 years

088

086

084

100

098

096

094

092

090

088

086

084Monitoring scenarios

Fig 4 Power (sd) for the detection of a decrease in the Chubut guanaco population (solid diamonds) or an increase (solid squares) under different scenarios 40 change over the next 6 or 10 years and survey frequency(a) low-density strata and thirty 30-km transects of road survey (b) medium-density and 10 30-km transects of road survey and (c) high-density strataand 10 30-km transects of road survey

H Wildlife Research J Pedrana et al

Finally if guanaco populations are to be sustained adaptivemanagement would require the monitoring of these populationsfor the early detection of any increase or unwanted decrease intheir numbers Traveling 600 km every 3 years is adequate todetect quite large changes within the two most dense guanacostrata in Chubut where any extractive enterprise might startBased on our experience 30-km transects should be distributedthroughout the strata area which we estimate would take sixcomplete working days by two or three technicians (Travainiet al 2015) However if the low HS category is included inthe follow-up the effort needed would be greater about ninefull-time working days per year and 15 fieldwork days every3 years Ideally an articulated association between Chubutand its neighbouring Santa Cruz could result in the bestbenefits both for the monitoring and conservation of guanacowild populations Moreover these two provinces should evendevelop common strategies for the sustainable use andregulation in the trade of guanaco products

Conservation and management implications

Patagonia holds some 700 000 km2 of grasslands and scrublandsthat still maintain healthy guanaco populations (Baldi et al2010) despite having been deeply modified by humanactivities Both ecosystem structure and functioning werealtered by the introduction of domestic herbivores at thebeginning of the 20th Century (Aguiar et al 1996 Bisigatoand Bertiller 1997) Grazing mining (Radovani et al 2014)and more recently global climate change (Pinto et al 2008)are human-related disturbances acting at different spatial andtemporal scales that have degraded Patagonian vegetationCurrent conditions are very different from the ones encounteredby the first European settlers who were unable to maintaina traditional sheep ranching system that was environmentallysustainable and economically profitable (Texeira and Paruelo2006) Traditional sheep ranching involved keeping sheepgrazing in extensive paddocks (requiring several thousandhectares) with limited attention a few occasions a year fordeworming or shearing (Garciacutea Brea et al 2010) Restoringthe Patagonian shrub-steppe as an ecologically functionalecosystem (Chardonnet et al 2002) would benefit fromefficient and modernised sheep husbandry complemented byactivities such as the sustainable use of wildlife and ecotourismIn this framework guanaco could play a role as an importanteconomic resource for landowners and local communities(Franklin et al 1997) In the absence of alternatives economicincentives could be the most effective driver for the conservationof the species (Fritz and Franklin 1994)

The greatest utility of our results is that they provide anexcellent overview of guanaco distribution and abundance in2006 and the big picture for the entire Chubut from whichthe government could start planning the sustainable use of thespecies Considering our HSM it is possible to identify theareas of the territory with higher densities where it will beeasier to select ranches with adequate guanaco abundance(Baldi et al 2010) to develop profitable herding and shearingexperiences Our HSM for Chubut together with otherenvironmental and human spatial data (eg land use roadcover and cadastral) should be used to generate a decision

tool that guarantees informed sustainable use of the speciesFinally our results need to be implemented in accordance withenvironmental education programs to change the perceptionof ranchers in Chubut province on native wildlife

Conflicts of interest

The authors declare no conflicts of interest

Acknowledgements

This work was primarily funded by the BBVA Foundation through a grantunder the Conservation Biology Program to A Rodriacuteguez and also bythe Agencia Nacional de Promocioacuten Cientiacutefica y Tecnoloacutegica (PICTON 30723) Additional support was provided by Universidad Nacional dela Patagonia Austral CONICET (PEI-6065) and CONAE

References

Aguiar M R Paruelo J M Sala O E and Laurenroth W K (1996)Ecosystem consequences of plant functional types changes in semiaridgrassland Journal of Vegetation Science 7 381ndash390 doi1023073236281

Amaya JN von Thuumlngen J and de Lamo DA (2001) Relevamiento ydistribucioacutendeguanacosen laPatagoniaComunicacioacutenTeacutecnicano109(Instituto Nacional de Tecnologiacutea Agropecuaria Bariloche Argentina)[In Spanish]

Ares J O BeeskowAM BertillerM B Rostagno CM IrisarriM PAnchorena J Defosseacute G E and Meroni C (1990) Structuraland dynamic characteristics of overgrazed grasslands of northernPatagonia Argentina In lsquoManaged Grasslands Regional Studiesrsquo(Ed A Breymeyer) pp 149ndash175 (Elsevier Amsterdam)

Baldi R Albon S D and Elston D A (2001) Guanacos and sheepevidence for continuing competition in arid Patagonia Oecologia 129561ndash570 doi101007s004420100770

Baldi R Novaro A Funes M Walker S Ferrando P Failla M andCarmanchahi P (2010) Guanacomanagement in Patagonian rangelandsa conservation opportunity on the brink of collapse In lsquoConservingWildlife while Maintaining Livestock in Semi-Arid Ecosystemsrsquo (EdsJ Du Toit R Kock and J Deutsch) pp 266ndash290 (Blackwell PublishingOxford UK)

Baldi R Acebes P Cueacutellar E Funes M Hoces D Puig S andFranklin W L (2016) Lama guanicoe The IUCN Red List ofThreatened Species 2016 Available at httpwwwiucnredlistorg[Accessed 6 April 2018]

BarrosVR ScianBV andMattioHF (1979)Camposdeprecipitacioacutende la provincia de Chubut (1931ndash1960) Geoacta 10 175ndash192

Bisigato A J andBertillerMB (1997) Grazing effects on patchy drylandvegetation in northern Patagonia Journal of Arid Environments 36639ndash653 doi101006jare19960247

Borrelli P and Cibils A (2005) Rural depopulation and grasslandmanagement in Patagonia In lsquoGrasslands Development OpportunitiesPerspectivesrsquo (Eds S G Reynolds and J Frame) pp 461ndash488 (FAOEnfield NH)

Buckland S T Anderson D R Burnham K P Laake J L BorchersD L and Thomas L (2001) lsquoIntroduction to Distance SamplingEstimating Abundance of Biological Populationsrsquo (Oxford UniversityPress Oxford)

Burnham K P and Anderson D R (2002) lsquoModel Selection andMultimodel Inference a Practical Information-theoretic Approachrsquo(Springer-Verlag New York)

Caro J Zapata S C Zanoacuten J I Rodriacuteguez A and Travaini A (2017)Ganaderiacutea ovina y usos alternativos del suelo en la Patagonia australArgentina Multequina (Mendoza) 26 33ndash50

Guanaco abundance distribution and monitoring Wildlife Research I

Finally if guanaco populations are to be sustained adaptivemanagement would require the monitoring of these populationsfor the early detection of any increase or unwanted decrease intheir numbers Traveling 600 km every 3 years is adequate todetect quite large changes within the two most dense guanacostrata in Chubut where any extractive enterprise might startBased on our experience 30-km transects should be distributedthroughout the strata area which we estimate would take sixcomplete working days by two or three technicians (Travainiet al 2015) However if the low HS category is included inthe follow-up the effort needed would be greater about ninefull-time working days per year and 15 fieldwork days every3 years Ideally an articulated association between Chubutand its neighbouring Santa Cruz could result in the bestbenefits both for the monitoring and conservation of guanacowild populations Moreover these two provinces should evendevelop common strategies for the sustainable use andregulation in the trade of guanaco products

Conservation and management implications

Patagonia holds some 700 000 km2 of grasslands and scrublandsthat still maintain healthy guanaco populations (Baldi et al2010) despite having been deeply modified by humanactivities Both ecosystem structure and functioning werealtered by the introduction of domestic herbivores at thebeginning of the 20th Century (Aguiar et al 1996 Bisigatoand Bertiller 1997) Grazing mining (Radovani et al 2014)and more recently global climate change (Pinto et al 2008)are human-related disturbances acting at different spatial andtemporal scales that have degraded Patagonian vegetationCurrent conditions are very different from the ones encounteredby the first European settlers who were unable to maintaina traditional sheep ranching system that was environmentallysustainable and economically profitable (Texeira and Paruelo2006) Traditional sheep ranching involved keeping sheepgrazing in extensive paddocks (requiring several thousandhectares) with limited attention a few occasions a year fordeworming or shearing (Garciacutea Brea et al 2010) Restoringthe Patagonian shrub-steppe as an ecologically functionalecosystem (Chardonnet et al 2002) would benefit fromefficient and modernised sheep husbandry complemented byactivities such as the sustainable use of wildlife and ecotourismIn this framework guanaco could play a role as an importanteconomic resource for landowners and local communities(Franklin et al 1997) In the absence of alternatives economicincentives could be the most effective driver for the conservationof the species (Fritz and Franklin 1994)

The greatest utility of our results is that they provide anexcellent overview of guanaco distribution and abundance in2006 and the big picture for the entire Chubut from whichthe government could start planning the sustainable use of thespecies Considering our HSM it is possible to identify theareas of the territory with higher densities where it will beeasier to select ranches with adequate guanaco abundance(Baldi et al 2010) to develop profitable herding and shearingexperiences Our HSM for Chubut together with otherenvironmental and human spatial data (eg land use roadcover and cadastral) should be used to generate a decision

tool that guarantees informed sustainable use of the speciesFinally our results need to be implemented in accordance withenvironmental education programs to change the perceptionof ranchers in Chubut province on native wildlife

Conflicts of interest

The authors declare no conflicts of interest

Acknowledgements

This work was primarily funded by the BBVA Foundation through a grantunder the Conservation Biology Program to A Rodriacuteguez and also bythe Agencia Nacional de Promocioacuten Cientiacutefica y Tecnoloacutegica (PICTON 30723) Additional support was provided by Universidad Nacional dela Patagonia Austral CONICET (PEI-6065) and CONAE

References

Aguiar M R Paruelo J M Sala O E and Laurenroth W K (1996)Ecosystem consequences of plant functional types changes in semiaridgrassland Journal of Vegetation Science 7 381ndash390 doi1023073236281

Amaya JN von Thuumlngen J and de Lamo DA (2001) Relevamiento ydistribucioacutendeguanacosen laPatagoniaComunicacioacutenTeacutecnicano109(Instituto Nacional de Tecnologiacutea Agropecuaria Bariloche Argentina)[In Spanish]

Ares J O BeeskowAM BertillerM B Rostagno CM IrisarriM PAnchorena J Defosseacute G E and Meroni C (1990) Structuraland dynamic characteristics of overgrazed grasslands of northernPatagonia Argentina In lsquoManaged Grasslands Regional Studiesrsquo(Ed A Breymeyer) pp 149ndash175 (Elsevier Amsterdam)

Baldi R Albon S D and Elston D A (2001) Guanacos and sheepevidence for continuing competition in arid Patagonia Oecologia 129561ndash570 doi101007s004420100770

Baldi R Novaro A Funes M Walker S Ferrando P Failla M andCarmanchahi P (2010) Guanacomanagement in Patagonian rangelandsa conservation opportunity on the brink of collapse In lsquoConservingWildlife while Maintaining Livestock in Semi-Arid Ecosystemsrsquo (EdsJ Du Toit R Kock and J Deutsch) pp 266ndash290 (Blackwell PublishingOxford UK)

Baldi R Acebes P Cueacutellar E Funes M Hoces D Puig S andFranklin W L (2016) Lama guanicoe The IUCN Red List ofThreatened Species 2016 Available at httpwwwiucnredlistorg[Accessed 6 April 2018]

BarrosVR ScianBV andMattioHF (1979)Camposdeprecipitacioacutende la provincia de Chubut (1931ndash1960) Geoacta 10 175ndash192

Bisigato A J andBertillerMB (1997) Grazing effects on patchy drylandvegetation in northern Patagonia Journal of Arid Environments 36639ndash653 doi101006jare19960247

Borrelli P and Cibils A (2005) Rural depopulation and grasslandmanagement in Patagonia In lsquoGrasslands Development OpportunitiesPerspectivesrsquo (Eds S G Reynolds and J Frame) pp 461ndash488 (FAOEnfield NH)

Buckland S T Anderson D R Burnham K P Laake J L BorchersD L and Thomas L (2001) lsquoIntroduction to Distance SamplingEstimating Abundance of Biological Populationsrsquo (Oxford UniversityPress Oxford)

Burnham K P and Anderson D R (2002) lsquoModel Selection andMultimodel Inference a Practical Information-theoretic Approachrsquo(Springer-Verlag New York)

Caro J Zapata S C Zanoacuten J I Rodriacuteguez A and Travaini A (2017)Ganaderiacutea ovina y usos alternativos del suelo en la Patagonia australArgentina Multequina (Mendoza) 26 33ndash50

Guanaco abundance distribution and monitoring Wildlife Research I

Chardonnet P desClers B Fischer J GerholdR Jori F andLamarqueF (2002) The value of wildlife Revue scientifique et technique(International Office of Epizootics) 21 15ndash51

Child B A (1988) The economic potential and utilization of wildlifein Zimbabwe Revue scientifique et technique (International Office ofEpizootics) 7 773ndash782

Child B A Musengezi J Parent G D and Child G F T (2012) Theeconomics and institutional economics of wildlife on private land inAfrica Pastoralism 2 18 doi1011862041-7136-2-18

Di Stefano J (2001) Power analysis and sustainable forest managementForest Ecology and Management 154 141ndash153 doi101016S0378-1127(00)00627-7

Donadio E and Burskik S W (2006) Flight behavior in guanacos andvicuntildeas in areas with and without poaching in western ArgentinaBiological Conservation 127 139ndash145 doi101016jbiocon200508004

Fewster R M Laake J L and Buckland S T (2005) Line transectsampling in small and large regionsBiometrics 61 856ndash861 doi101111j1541-0420200500413_1x

Field S A Tyre A J and PossinghamH P (2005) Optimizing allocationof monitoring effort under economic and observational constraints TheJournal of Wildlife Management 69 473ndash482 doi1021930022-541X(2005)069[0473OAOMEU]20CO2

Franklin W L (1983) Contrasting socioecologies of South Americarsquoscamelids the vicugna and Guanaco In lsquoAdvances in the Study ofMammalian Behaviorrsquo (Eds J F Eisenberg and D Kleiman)pp 573ndash629 (Special Publications No 7 American Society ofMammalogists Shippensburg PA)

Franklin W L Bas F Bonacic C F Cunazza C and Soto N (1997)Striving to manage Patagonia guanacos for sustained use in thegrazing agroecosystem of southern Chile Wildlife Society Bulletin 2565ndash73

Fritz M A and Franklin W L (1994) First estimates of guanaco malegroup harvestability in the Patagonia of South America Vida SilvestreNeotropical 3 84ndash90

Garciacutea Brea A Zapata S C Procopio D E Martiacutenez Peck R andTravaini A (2010) Evaluacioacuten del intereacutes de productores ganaderosen el control selectivo y eficiente de predadores en la Patagonia AustralActa ZooloacutegicaMexicana 26 303ndash321 doi1021829azm2010262703

Gerken M (2010) Relationships between integumental characteristics andthermoregulation in South American camelids Animal 4 1451ndash1459doi101017S1751731109991443

Gibbs J P and Ene E (2010) Program monitor estimating the statisticalpower of ecological monitoring programs Version 1100 Availableat httpwwwesfeduefbgibbsmonitor [Accessed 9 April 2018]

Gibbs J P and Melvin S M (1997) Power to detect trends in waterbirdabundance with callndashresponse surveys The Journal of WildlifeManagement 61 1262ndash1267 doi1023073802125

Golluscio R A Deregibus V A and Paruelo J M (1998) Sustainabilityand range management in the Patagonian steppes Ecologiacutea Austral 8265ndash284

Hastie T J and Tibshirani R J (1990) lsquoGeneralized Additive Modelsrsquo(Chapman amp Hall London)

Leoacuten R J C BranD CollantesM Paruelo JM andSorianoA (1998)Grandes unidades de vegetacioacuten en la Patagonia Ecologiacutea Austral 8123ndash141

Lichtenstein G and Carmanchahi P D (2012) Guanaco managementby pastoralists in the Southern Andes Pastoralism 2 1ndash16 doi1011862041-7136-2-16

Montes C De Lamo D A and Zavatti J (2000) Distribucioacuten deabundancias de Guanacos (Lama guanicoe) en los distintos ambientesde Tierra del Fuego Argentina Mastozoologiacutea Neotropical 7 23ndash31

Montes M C Carmanchahi P D Rey A and Funes M (2006) Liveshearing free-ranging guanacos (Lama guanicoe) in Patagonia for

sustainable use Journal of Arid Environments 64 616ndash625 doi101016jjaridenv200505008

Murtaugh P A (1996) The statistical evaluation of ecological indicatorsEcological Applications 6 132ndash139 doi1023072269559

Nabte M J Marino A I Rodriacuteguez M V Monjeau A and Saba S L(2013) Range management affects native ungulate populations inPeninsula Valdeacutes a World Natural Heritage PLoS One 8 e55655doi101371journalpone0055655

Ortega I M and Franklin W I (1995) Social organization distributionand movements of a migratory Guanaco population in the ChileanPatagonia Revista Chilena de Historia Natural 68 489ndash500

Paruelo JM Jobbaacutegy EG and SalaO E (1998) Biozones of PatagoniaEcologiacutea Austral 8 145ndash153

Pedrana J RodriacuteguezABustamante J TravainiA andZanoacutenMartiacutenezJ I (2009) Failure to estimate reliable sex ratios of guanaco from road-survey data Canadian Journal of Zoology 87 886ndash894 doi101139Z09-079

Pedrana J Bustamante J Travaini A and Rodriacuteguez A (2010) Factorsinfluencing guanaco distribution in southern Argentine Patagonia andimplications for its sustainable use Biodiversity and Conservation 193499ndash3512 doi101007s10531-010-9910-1

Pinto J Bonacic CHamilton-West C Romero J andLubroth J (2008)Climate change and animal diseases in South America InternationalOffice of Epizootics 27 599ndash613

Puig S (1995) Abundancia y distribucioacuten de las poblaciones de GuanacosIn lsquoTeacutecnicas para el manejo del Guanacorsquo (Ed S Puig) pp 57ndash68(IUCN Gland Switzerland) [In Spanish]

R Development Core Team (2016) lsquoR A Language and Environment forStatistical Computingrsquo Available at wwwR-projectorg [Verified 11November 2018]

Radovani N I Funes M C Walker R S Gader R and Novaro A J(2014) Guanaco Lama guanicoe numbers plummet in an area subjectto poaching from oil-exploration trails in Patagonia Oryx 49 1ndash9

Raedeke K H (1979) Population dynamics and socioecology of theguanaco (Lama guanicoe) of Magallanes Chile PhD ThesisUniversity of Washington Seattle WA

Raedeke K H (1982) Habitat use by guanacos (Lama guanicoe) and sheepon common range Tierra del Fuego Chile Turrialba 32 309ndash314

Redford K H and Eisenberg J F (1992) lsquoMammals of theNeotropics the Southern Conersquo (The University of Chicago PressChicago IL)

Rey A (2010) Efectos del manejo sobre la dinaacutemica de poblacionesde guanacos silvestres (Lama guanicoe) y mortalidad por enganchesen alambrados en campos ganaderos de Patagonia Norte ArgentinaPhD Thesis University of Buenos Aires Argentina [In Spanish]

Rey A Carmanchahi P D Puig S and Guichoacuten M L (2009)Densidad estructura social actividad y manejo de guanacos silvestres(Lama guanicoe) en el sur del Neuqueacuten Argentina MastozoologiacuteaNeotropical 16 389ndash401

Roth H H and Merz G (1996) lsquoWildlife Resources a Global Account ofEconomic Usersquo (Springer Berlin)

Saba S L and Battro P (1987) Estimacioacuten de la densidad poblacional deGuanacos (Lama guanicoe Muller) Turrialba 37 113ndash118

Sacchero D Maurino M J and Lanari M R (2006) Diferencias decalidad y proporcioacuten de down en muestras individuales de vellonesde Guanaco (Lama guanicoe) en distintas ecorregiones de ArgentinaRevista Argentina de Produccioacuten Animal 26 211ndash216

Sakamoto Y Ishiguro M and Kitagawa G (1986) lsquoAkaike informationcriterion statisticsrsquo (KTK Scientific Publishers Tokyo)

Schroeder N Matteucci S D Moreno P G Gregorio P Ovejero RTaraborelli P and Carmanchahi P D (2014) Spatial and seasonaldynamic of abundance and distribution of guanaco and livestockinsights from using density surface and null models PLoS One 9e85960 doi101371journalpone0085960

J Wildlife Research J Pedrana et al

Schroeder N M Gonzaacutelez A WisdomM Nielson R Rowland MMand Novaro A J (2018) Roads have no effect on guanaco habitatselection at a Patagonian site with limited poaching Global Ecologyand Conservation 14 e00394 doi101016jgecco2018e00394

Soriano A (1956) Los distritos floriacutesticos de la Provincia PatagoacutenicaRevista Investigacioacuten Agropecuaria 10 323ndash347

Soriano A and Paruelo J M (1990) El manejo de campos de pastoreo enPatagonia aplicacioacuten de principios ecoloacutegicos Ciencia Hoy 2 44ndash53

Texeira M and Paruelo J M (2006) Demography population dynamicsand sustainability of the Patagonian sheep flocks Agricultural Systems87 123ndash146 doi101016jagsy200411005

Thomas L Buckland S T Rexstad E A Laake J L Strindberg SHedley S L Bishop J R B Marques T A and Burnham K P(2010) Distance software design and analysis of distance samplingsurveys for estimating population size Journal of Applied Ecology 475ndash14 doi101111j1365-2664200901737x

Travaini A Bustamante J Rodriacuteguez A Zapata S Procopio DPedrana J and Martiacutenez Peck R (2007) An integrated framework

to map animal distributions in large and remote regions Diversity ampDistributions 13 289ndash298 doi101111j1472-4642200700338x

Travaini A Zapata S C Bustamante J Pedrana J Zanoacuten J I andRodriacuteguez A (2015) Guanaco abundance and monitoring in SouthernPatagonia distance sampling reveals substantially greater numbersthan previously reported Zoological Studies (Taipei Taiwan) 54 23

Von Thuumlngen J and Lanari M R (2010) Profitability of sheep farmingand wildlife management in Patagonia Pastoralism 1 274ndash290

Wheeler J C (2006) Guanaco 1 Working to save Perursquos endangeredguanacos Camelid Quiarterly 5 39ndash42

Young J K and Franklin W L (2004) Territorial fidelity of maleguanacos in the Patagonia of Southern Chile Journal of Mammalogy85 72ndash78 doi1016441545-1542(2004)085lt0072TFOMGIgt20CO2

ZanoacutenMartiacutenez J I Travaini A Zapata S C Procopio D and SantillaacutenM A (2012) The ecological role of native and introduced species inthe diet of the puma Puma concolor in southern Patagonia Oryx 46106ndash111 doi101017S0030605310001821

Guanaco abundance distribution and monitoring Wildlife Research K

wwwpublishcsiroaujournalswr