habita abundance for bobcats

Upload: rkalsi

Post on 03-Apr-2018

223 views

Category:

Documents


0 download

TRANSCRIPT

  • 7/28/2019 Habita Abundance for Bobcats

    1/10

    Habitat-Relative Abundance Relationship for Bobcats in Southern Illinois

    Clayton K. Nielsen; Alan Woolf

    Wildlife Society Bulletin, Vol. 30, No. 1. (Spring, 2002), pp. 222-230.

    Stable URL:

    http://links.jstor.org/sici?sici=0091-7648%28200221%2930%3A1%3C222%3AHARFBI%3E2.0.CO%3B2-C

    Wildlife Society Bulletin is currently published by Alliance Communications Group.

    Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available athttp://www.jstor.org/about/terms.html. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtainedprior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content inthe JSTOR archive only for your personal, non-commercial use.

    Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained athttp://www.jstor.org/journals/acg.html.

    Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printedpage of such transmission.

    The JSTOR Archive is a trusted digital repository providing for long-term preservation and access to leading academicjournals and scholarly literature from around the world. The Archive is supported by libraries, scholarly societies, publishers,and foundations. It is an initiative of JSTOR, a not-for-profit organization with a mission to help the scholarly community takeadvantage of advances in technology. For more information regarding JSTOR, please contact [email protected].

    http://www.jstor.orgSun Aug 19 03:00:27 2007

    http://links.jstor.org/sici?sici=0091-7648%28200221%2930%3A1%3C222%3AHARFBI%3E2.0.CO%3B2-Chttp://www.jstor.org/about/terms.htmlhttp://www.jstor.org/journals/acg.htmlhttp://www.jstor.org/journals/acg.htmlhttp://www.jstor.org/about/terms.htmlhttp://links.jstor.org/sici?sici=0091-7648%28200221%2930%3A1%3C222%3AHARFBI%3E2.0.CO%3B2-C
  • 7/28/2019 Habita Abundance for Bobcats

    2/10

    Carnivore Research and Management 222 H AB IT AT -R EL AT IVE A B U N D A N C E R EL A T IO N SH IP F O R BOBCATS

    Habitat-relative abundance relationship for bobcats in southern Illinois CZaj^t~iiK. .\i'el.seti aitcl Alan Wtoolf

    Abstract Large-scale multivariate models of habitat suitability have been developed for severalcarnivore species. However, rarely have habitat models been combined wi th demo-graphic information to create habitat-relative abundance relationships. We usedremotely sensed land-cover information, bobcat habitat use data, multivariate distancestatistics, and a geographic information system to provide wildl ife managers a spatiallyexplicit depiction of bobcat (Lynx rufus) relative abundance in a 12,512-kmL portion ofsouthern I llinois. We used the Penrose distance statistic to model regional habitat sim-ilarity to areas within core areas of 52 radiocollared bobcats captured during1995-1 999. Bobcat core areas were comprised primari ly of forest cover (61%). Con-versely, the southern Illinois region consisted of a more even mix of agricultural (36"/0),forest (29%), and grass cover (22%). Mean patch size of forest cover and proportion offorest cover were most correlated ( ~ 0 . 3 9 )o Penrose distance. The Penrose distancemodel was validated using an independent data set of bobcat sighting locations ( n =248). Thirty-one percent and 81OO of independent bobcat sightings occurred in the top10% and 25Y0 of distributions of Penrose distances, respectively. We then modeled rel-ative abundance for the region based on Penrose distance, abundance information fromareas occupied by radiocollared bobcats, and bobcat sighting locations. Given the col-lect ion of radiotelemetry data, this modeling technique can provide a better alternativethan more traditional methods (e.g., scent station surveys) to assess large-scale distribu-tion and abundance of solitary carnivores.

    Key words bobcat, habitat-relative abundance relationship, habitat modeling, Lynx rufus, multivari-ate distance statistics, southern Illinois

    Large-scale multivariate models of habitat suit- In a few instances, habitat models have beenability have been developed for several carnivore combined with demographic information to createspecies (e.g.. black bears [Ursus anzericnnus] habitat-population density relationships (Roseber-Clark et al. 1993;Florida panthers [Putnu concolor ry and Woolf 1998). Roloff and Haufler (1997)co r y i ] , Maehr and Cox 1995; and gray wolves demonstrated this approach, combining habitat[Canis Ez~pus] , ladenoff et al. 1995). Two primary suitability maps and species home range illforma-components of these models are habitat use data tion using published data from Canada lynx (Lynxfrom several animals and remotely sensed land- canadensis) studies. Although such approachescover information. These models are then often could pro\-ide managers with desirable methods toused to describe or predict habitat needs for con- estimate regional abundance or as part of a popula-servation purposes. tion viability analysis (Roloff and Haufler 199'),

    Authors' address: Cooperative LVildlife Research Laboratory and Department o i Zoolog)i mailc code 6504, Southern I l l inois Un i-versitv, Carb ondale, IL 62901, USA; e-mail for Nielsen: kez0920siu.edu .

    http:///reader/full/kez0920siu.eduhttp:///reader/full/kez0920siu.edu
  • 7/28/2019 Habita Abundance for Bobcats

    3/10

    Habitat-relative abundance relationship for bobcats Nielsen and Woolf 223they are rare for solitary carnivores.

    Wildlife managers in Illinois needed such infor-mation to focus management of bobcats (L. rufus).Based on increased sightings and a statewide assess-ment of relative abundance and habitat suitability(Woolf et al. 2002), bobcats were taken off thestate's threatened species list in 1999 (Woolf et al.2000). We used remotely sensed land-cover infor-mation, bobcat habitat use data, multivariate dis-tance statistics, and a geographic information sys-tem to provide wildlife managers a spatially explicitdepiction of bobcat relative abundance in southernIllinois.

    Study areaWe trapped and radiocollared bobcats on 2 studyareas (eastern study area, 1,000 km2;western study

    area, 791 km2) in the 13 southernmost counties ofIllinois (12,512 km2;Woolf and Nielsen 1999). Thisregion included the Shawnee Hills, Ozark, LowerMississippi River Bottomlands, and Coastal Plain nat-ural divisions (Neely and Heister 1987) and wascomprised of 39% cropland, 25% hardwood forest,and 24% rural grassland (Luman et al. 1996).Streams and roads were abundant on the landscape(stream density = 1.1 km/km2, road density = 1.4km/km2). Elevation ranged from 92 to 316 m, withan average slope of l. dO .Human population densi-ty was 21.5 persons/km2. Land cover of the easternstudy area consisted primarily of closed-canopymixed hardwood forests (55%) dominated by whiteoak (Quercus alba), black oak (Q. rubra) and hick-ory (Cay a spp.), rural grasslands (26%),and crop-land (11%) characterized by corn and soybeans(Luman et al. 1996). Land cover of the westernstudy area consisted of a lower proportion of forests(46%) and rural grassland (8%)and a higher propor-tion of cropland (28%). Bobcats have been protect-ed from legal harvest in Illinois since 1971 and wereconsidered threatened in the state until 1999.

    MethodsApproach

    Our goal was to create a habitat-relative abun-dance relationship for adult bobcats in the south-ern Illinois region. We calculated core areas ofradiocollared bobcats and used core area size todetermine an appropriate scale for habitat model-ing. We then determined a set of regional habitatvariables based on remotely sensed land-cover data.

    Next, we modeled habitat similarity between bob-cat core areas and the rest of the region. Finally,wemodeled relative abundance for the region basedon habitat similarity, abundance information fromareas occupied by radiocollared bobcats, and bob-cat sighting locations.Trapping and handling

    We captured bobcats during November- march1995-1999 with either cage-type traps constructedof galvanized wire mesh (38 x 38 x 90 cm) orpadded number 3 Soft-catchB (Woodstream Co..Lititz, Pa., USA) foothold traps. All traps were bait-ed with whole or partial carcasses and chunks ofmeat from a variety of animals (e.g., roadkilledwhite-tailed deer [Odocoileus z~irgin ianus]).Visualattractants (e.g., bird wings) and commercial luresalso were used at most trap sites. Captured bobcatswere chemically immobilized for handling with acombination of ketamine hydrochloride (HCl) andxylazine HCl (both in 100 mg/mL concentrationsolution). We sexed,weighed, measured, and classi-fied bobcats as adults Q 2 yr) or juveniles based onmass (bobcats

  • 7/28/2019 Habita Abundance for Bobcats

    4/10

    224 Wildlife Society Bulletin 2002,30(1):222-230than 20 minutes elapsed between first and lastbearings for 94% of all locations. We used the pro-gram LOCATE I1 (Nams 1990) to estimate locationsaccording to the maximum likelihood estimator(Lenth 1981) and to calculate bearing error (n=200, .t=4.2f0.2) and error polygons (72 =200,9=1.6k0.1 ha, Springer 1979). Homing, capture,al, and aerial locations comprised 9% of total loca-tions and were plotted on USGS topographic maps.We radiotracked most (87%) bobcats for less than 1j7ear (2radiodaysbobcat =477i2G). We collectedan average of 8 3 f 12 annual locations/bobcat, >50locations were obtained for 81% of bobcats, and>29 locations were collected for all bobcats. Weobtained most (97%) locations of individuals 220hours apart.Core area estimation

    We used the program RANGES V (Kenward andHodder 1996) to estimate 50% core area bound-aries of addt bobcats using the minimum convexpolygon (MCP) estimator (Mohr 1947). We estimat-ed core areas based on all locations obtained foreach individual. Although other methods exist todepict spatial use (White and Garrott l990), weused the ,MCP estimator because it provided 1 areaof use/individual, which was suitable for our analy-sis. We used core areas rather than home rangesbecause core area overlap was minor (Nielsen andWoolf 2001) and core areas are estimated more reli-ably than home ranges (Seatnan et al. 1999). Weestimated core areas for 52 individuals (22 M, 30 F;Nielsen and Woolf 2001).Land-cover data

    We performed all geographic information sys-tems operations using ARCANFO and ARCVIEW(Environmental Systems Research Institute, Red-lands, Calif,. USA) and all statistical analyses (a=0.05) using STATISTIX c44nalytical Software 1996).We derived land-cover information from the IllinoisCritical Trends Assessment Project Land CoverDatabase (Luman et al. 1996) and reclassified fromthe original 23 cover classes into the follom7ing 8aggregations: urban. transportation (i.e., roads andrailroads), agriculture, grass. forest, open water,stream, and marsh. Because some roads andstreams were not included in the original land-cover classification, we converted road and streamdata (Illinois Department of Natural Resources1996) from vector to raster format and mergedwith the land-cover classification.

    Figure 1. Hexagons used to cl ip la nd co\,er for habitat-relativeabundance m ode l ing for bobcats in southern I l l inois,1995-1 999.

    Habitat variable selection%re calculated habitat variables based on land-

    scape metrics for the 13 southertltnost counties ofIllinois to represent the entire region. We created acontinuous coverage of 2,'04 non-overlappinghexagons of 4.5 km2 (i .e .,2size of MCP core areas:Nielsen 2000) that covered southern Illinois but didnot overlap the regional boundary (Figure 1). Weused FRAGSTATS (McGarigal and Marks 1995) tocalculate landscape metrics associated with landcover within each hexagon. We log-transformedcover-type proportions and used a value of 0.001for nd l proportions (Aebischer et al. 1993).

    We chose representative variables from each of 8major metric groups, resulting in 153 potential vari-ables for analysis (Table 1). We retained a smallergroup of variables for habitat modeling based onpresumed importance to bobcats and univariatestatistics. First, for each cover type, we retained itslog-transformed proportion. Second, within eachco~ierype, we calculated nonparametric Spearmanrank correlations for variables within each metricgroup and determined the number of nonsignifi-cant correlations per variable. Third, we eliminated1 of all pairs of correlated (1'50.0 5) variables with-in each metric group depending on number of non-significant correlations with other variables in thegroup. Values of significant correlations varied (r=0.088-0.920) because occasionally variables werenot present in hexagons, resulting in different sam-ple sizes. We retained variables most correlated toothers within each group (i.e., most representativeof the group). In cases of ties, we chose the variablewith suspected greater biological importance tobobcats (e.g., patch density of woods was selected

  • 7/28/2019 Habita Abundance for Bobcats

    5/10

    Habitat-relative abundance relationship for bobcats Nielsen and Woolf 225Table 1. Potential habitat variables used to mo del habitat use by bobcats in southern I lli- reduced this data set to 15 vari-nois, 1995-1 999 . Variables were calcu lated using FRAGSTATS software (McGa rigal andMarks 199 5: 14-1 5). Class metrics were ca lculated for all 8 cover classes except wh ennoted otherwise. Cover classes inclu de d urban (URB), transportation (TRAN), agricu lture(AG), grass (GRS), forest (FOR), open water (WATI, stream (STR), and marsh (MA R).- -CalculationArea metrics

    ClassClass-landscapeb

    Patch metric sCClass-landscapeClass-landscapeClass-landscape

    Edge metricsCClass-landscape

    Shape metricsCClass-landscapeClass-landscapeClass-landscapeClass-landscape

    Core area metricsciClassClass-landscapeClass-landscapeClass-landscapeClass-landscapeClass-landscapeClass-landscapeClass-landscapeDiversity metricsCLandscapeLandscapeLandscape

    Acronym Metric (unit)PCTa Percentage of landscape (ha)LPI Largest patch ind ex ('10)PD Patch density (n o. il0 0 ha)M P S ~ Mean patch size (ha)PSCV Patch size co eif ici en t of var iatio n (10)EDe Edge density (mlh a)MSI Mea n shape indexAW MSl Area-weighted mean shape indexLSI Landscape shape indexDLFD Do ubl e log fractal dimensionCO/oLANDCADTCAlgMCAlM C A lMCA2CACVlCACV2

    Core area percentage of landscapeCore area density (no.1100 ha)Total core area ind ex ('10)Mean core area (Oh)Mean core areafpatch (ha)Mean areafdisjunct corePatch core area coe ificient o f variation (%)Disjunc t core area coefiicient of variation (%)

    SHD l Shannon's diversity indexMS lDl Mo difie d Simpson's diversity indexPRD Patch richness density (no ./10 0 ha)

    ables by correlating physiog-nomic variables (e.g.,cover-typediversity) within each coverclass and retaining the variablemost correlated to others.Because urban land cover exist-ed within only 2 of 52 (4%)bobcat core areas, we excludedthe 2 variables associated withurban land cover and excluded2 additional variables (edge den-sity of water and marsh covertypes) because they existed inonly 25 and 35 of 52 (148%)bobcat core areas, respectively.This resulted in 11 final vari-ables for habitat-relative abun-dance modeling. Further analy-sis required data normality;therefore, we transformed 2variables (mean patch size offorest and grass cover) to nor-mal distributions (Wilk- Shapirostatistic = 0.989-0.992) usingsquare-root transformations.The other variables were dis-tributed normally and not trans-formed (Wilk-Shapiro statistic=0.898-0.989).

    Landscape RPR Relative patch richness (% iLand scape SHE1 Shannon's evenness index Habitat- relativeLa nd sca ~e MSlEl Mod ified Sim~ son'sevenness index abundance relationship

    Nearest neighbor metricsbc Modeling habitat sirizilarityClass-landscape M N K Mea n nearest neighbor distance (m) to areas occupied bll radiocol-Class-landscape NN CV Nearest neighbor coefficient of variation (% i lared bobcats. We used theClass-landscape MPI Mea n pro xim ity index Penrose distance statistic to

    Contagion metricsC measure similarity between theClass-landscape 111 Interspersion and juxtaposition index (%) mean habitat vector calculatedLandscape CO NTAG Contagion index (%I from core areas of radiocollareda Retained for analysis for all cover classes. adult bobcats and the rest of

    Landscape refers to each c ell used to c lip land cover. southern Illinois. We calculatedNot calculated for TRAN or STR cover classes. Penrose distance as

    C' Retained for analysis for the GRS and FOR cover classes onlyRetained for analysis for the landscape only.No t calculated for the WAT cover class.

    6 Retained for analysis for the AG cover class only. where populations i andj repre-sented bobcat core areas andstudy area hexagons, respective-

    over patch density of water). This approach result- 1y;pwas the number of habitat variables evaluated,ed in 50 potential habitat variables. We further y was the variable value, k was each observation,

  • 7/28/2019 Habita Abundance for Bobcats

    6/10

    226 WlEdltfeSociety BuUetin 2002,30(1):222-230Table 2. Mean values (+ SE) of 11 habitat variables used to model bobcat habitat in southernIll ino is and correlations between each variable and Penrose distance (PDI, 1995-1999.

    CorrelationbetweenMean Study study areaVariable" vectorb area and PDCPercentage of transportation coverPercentage of agricultural coverPercentage of grass coverPercentage o i orest coverPercentage of water coverPercentage of stream coverPercentage of marsh coverEdge density of the landscapeTotal core area index of agricultural coverMean patch size of grass coverMean patch slze of forest cover

    "over type values are shown as raw percentages and not their log-ratio transformedequivalents.Calculated from 50% minimum convex polygon core areas of 52 bobcats.

    represented only 1 bobcatcore area. We assumedthat areas occupied byradiocollared bobcats rep-resented areas of greatestrelative abundance (1.0bobcat/hexagon) and de-termined mean Penrosedistance for those hexa-gons containing bobcatcore areas (Figure I).Mean Penrose distancewas 0.91; hence. weassumed that hexagonswith Penrose distances-0.91 to contain 1.0 bob-cat.

    We estimated zero andintermediate relative abun-dance based on the afore-mentioned sighting loca-tion data . We overlaid

    Significant ( P < 0.05) correlat~ons re denoted as i s ) .

    and Vwas variance (Manly 1986). This technique issimilar to the LMahalanobis distance statistic (Manly1986) used previously by several researchers forhabitat modeling (Clark et al. 1993, Knick and Dyer1997, Corsi et al. 1999). We calculated the meanhabitat vector as the mean values of the 11 habitatvariables within bobcat core areas. We evaluatedeach hexagon on the study area relative to the meanhabitat vector, where a Penrose distance of 0 wasclosest to the mean vector. We made all calculationsin a spreadsheet and appended the output databaseto the hexagon coverage to create a GIS map ofregional Penrose distance. We then correlated Pen-rose distance to each habitat variable to determinerelative importance of each variable in calculatingPenrose distance over the entire study area.Model validation. We validated the Penrose dis-tance model using an independent data set of bob-cat sighting locations collected during 1995-1999( n=248;Woolf and Nielsen 1999). Following Knickand Dyer (1997) and Corsi et al. (1999), we overlaidsighting locations onto the habitat map and deter-mined the cumulative frequency of bobcat sight-i n g ~elative to Penrose distance.

    Modeling regional relative abundance. Wemodeled bobcat relative abundance for the regionbased on Penrose distance associated with eachhexagon, abundance information from areas occu-pied by radiocollared bobcats, and bobcat sightingdata. For simplicity,we assumed that each hexagon

    sighting locations on the Penrose distance map andcalculated frequency distributions. Because nobobcats were sighted in hexagons with Penrosedistance >6.7, we considered hexagons havingthese values to contain no bobcats. We created amathematical relationship between Penrose dis-tance versus frequency of bobcat sightings to esti-mate intermediate abundance (i.e., between 0 and1.0 bobcat/hexagon). A logarithmic curve yieldedthe greatest R2 (0.84); hence, the relationship J J =-0.51 ln[x] +0.95 ,where .Y is Penrose distance and-2' is bobcat abundance.

    ResultsBobcat core area hexagons were comprised pri-marily of forest cover (61%) with agricultural and

    grass cover combining for an additional 30% (Table2). Conversely, he region consisted of a more evenmix of agricultural (36%). forest (29%), and grasscover (22%). Although represented in relativelylow proportions, percentage water cover differedmost between core areas and the entire region; per-centage agricultural and forest cover also differedmarkedly. Mean patch size of grass cover and per-centage stream cover were nearly the samebetween core areas and the entire region. Meanpatch size of forest cover and proportion of forestcover were most correlated to Penrose distance(Table 2).

  • 7/28/2019 Habita Abundance for Bobcats

    7/10

    Habitat-relative abundance relationship for bobcats Nielsen and Woolf 227

    Figure 2. Penrose distance map depicting habitat similaritybetween bobcat core areas and the southern Illinois region,1995-1 999. Lesser Penrose distances indica te greater habitatsimilarity to bobcat core areas.

    Mean Penrose distances for bobcat core areasand the entire region were 0.91f 0.80, range=0.10-5.32) and 1.41+0.92 (range =0.14-9.13,respectively. Regionally 1,018 of 2,704 (38%) hexa-gons contained Penrose distances 10 .9 (Figure 2).Least average Penrose distance (i.e., with most sim-ilarity to bobcat core areas) occurred in east-south-east and west-central portions of southern Illinois.Penrose distance was greatest in the extreme north-east and Mississippi River bottomland areas. Theseareas consisted of proportionately more agricultur-al cover than the rest of the region. Thirty-one per-cent and 81% of independent bobcat sightingsoccurred in the top 10%and 25% of distributions ofPenrose distances, respectively. Predicted relativeabundance was relatively uniform across the studyarea, except for the western portions comprisingMississippi River bottomland and the extremenortheast (Figure 3).

    DiscussionRoloff and Haufler (1997) used existing literature

    from Canada lynx studies to illustrate the utility of anovel approach to assess population viability. Theirgoal was not to actually conduct a population viabil-ity assessment, but to illustrate the use of their tech-nique (Roloff and Haufler 1997). We created a habi-tat-relative abundance relationship for bobcats insouthern Illinois using a similar approach; however,ours was based on empirical data and was intendeddirectly for use by wildlife managers in Illinois.Biologists frequently use habitat suitability infor-mation to focus wildlife management. These efforts

    Figure 3. Bobcat relative abundance in southern Illino is basedon a habitat-relative abundance relationship, 1995-1 999.

    usually rely on habitat use data from relatively fewanimals, where inferences within a small study areaare extrapolated to regions animals are known toinhabit but that were not monitored explicitly(Maehr and Cox 1995, Mladenoff et al. 1995). Weused a multivariate distance statistic (i.e., Penrosedistance) to achieve this goal for bobcats in south-ern Illinois. Other studies have used distance sta-tistics to model habitat suitability (Clark et al. 1993,Knick and Dyer 1997, Corsi et al. 1999). The pri-mary advantage this has over other techniques(e.g., logistic regression) is that there is no need toassume used and unused habitats are classifiedwithout error. Other studies used the Mahalanobisdistance statistic, which accounts for correlationsamong variables (Ttlanly 1986). Although Penrosedistance accounts for variance of each variable, itdoes not account for correlations; in our model,high correlations of habitat variables were removedduring variable reduction procedures. Otherresearchers have used primarily compositional vari-ables and a set of vector grids coded 0 or 1 for non-use and use, respectively, whereas we performedcalculations for each core area hexagon in a spread-sheet using raw values. Because the Mahalanobisdistance equation does not contain squared terms(Manly 1986), the negative values provided byphysiognomic variables disallowed use of theMahalanobis distance statistic. Regardless, Manly(1986) demonstrated that Mahalanobis distanceand Penrose distance statistics provide similarresults.

    We determined hexagon size for modeling basedon core areas of radiocollared bobcats and modeledPenrose distance relative to areas occupied by

  • 7/28/2019 Habita Abundance for Bobcats

    8/10

    these bobcats. Bobcat core areas represented ageneral measure of areas used by bobcats and rep-resented a useful scale for modeling. Most coreareas were estimated based on individuals with >50locations. However, we obtained fewer locationsfor some individuals, which may have somewhatbiased estimates of core area size. Number of loca-tions was not correlated to core area size (r=0.22,P< 0.05 ); therefore, varying location sample sizeshad little effect on core area size for bobcats in ourstudy. There also was concern that location samplesizes were insufficient to delineate core areas; suchis especially true given that we obtained relativelyfew locations for some bobcats. However, averagelocation sample sizes for bobcat spatial use analysisrarely exceed those obtained during our studyespecially for those involving >50 individuals.

    Increasing the number of locations/bobcat likelywould have resulted in larger MCP core area esti-mates Uennrich and Turner 1969). However, webelieve this would not have changed modelingresults, because habitat use did not differ betweenbobcat home ranges and core areas in southern Illi-nois (C.K. Nielsen, unpublished data). Thus, habitatwithin slightly larger core areas would not have dif-fered from habitat within core areas as reported inthis study.

    The Penrose distance model was very accuratewhen validated with an independent data set ofbobcat sighting locations. Although validity andlocation accuracy of sighting data can obscurewildlife-habitat associations (Agee et al. 1989,Stoms et al. 1993, Palma et al. 1999), sighting datawas the only set of independent data over theentire region that was useful for model validation.Further discussion of limitations of our sightingdata are described in Woolf et al. (2000) and Woolfet al. (2002). Of additional concern were potentialcompounding of errors from multiple models andstatistical techniques, whose aggregate effect onthe final relative abundance model was unknown.However, we had no other means by which toassess this error o r validate the model (e.g., viascent-station indices).

    Mean patch size of forest and percentage forestcover were the 2 variables most correlated to Pen-rose distance, which demonstrated their impor-tance in determining bobcat habitat suitability.These variables were correlated negatively to Pen-rose distance, such that when mean patch size offorest cover was large and forest cover plentiful,Penrose distances were low, indicating suitable

    habitat. These results agree with our previous analy-ses (Woolf et al. 2002), which found that forest com-positional and physiognomic variables were primarypredictors of bobcat abundance and distribution andhabitat suitability statewide in Illinois. Forest coveris important to bobcats over large scales (Conner etal. 2001, Lovallo et al. 2001) for numerous reasons,including providing denning cover and preyresources (Anderson 1987, Rolley 1987). Positivecorrelations of percentage transportation, water,marsh, and agricultural cover and total core areaindex of agriculture to Penrose distance indicatedpoorer habitat suitability These cover types likelyprovided relatively little habitat value for bobcatswhen compared to forest cover types.

    Wildlife managers often require information onpopulation abundance and distribution, which isespecially challenging to determine for carnivores.Census techniques such as scent-station surveys areof limited value over larger scales due to poor pre-cision of population estimates and low visitationrates (Diefenbach et al. 1994, Woolf and Nielsen1999). Therefore, we used habitat suitability infor-mation in conjunction with calculated abundancein trapped areas and extrapolated abundance esti-mates to other areas using a bobcat sighting-habitatrelationship. Van Horne (1983) warned about usinghabitat suitability information to assess abundancewithout knowledge of survival, reproduction, andphysical condition of the species in question. How-ever, because bobcats in southern Illinois have rela-tively high natural survival and recruitment rates(Nielsen 2000), a habitat-relative abundance rela-tionship was appropriate.

    Management implicationsBecause of the rarity and secretive behavior of

    medium-sized and large carnivores, quantifyingtheir relative abundance over large scales is logisti-cally difficult. We used data commonly collected inradiotelemetry studies (e.g., spatial use estimates)that provided a more desirable alternative thanother methods (e.g.,scent-station surveys) to assessdistribution and abundance of bobcats in southernIllinois. These techniques can be used by wildlifemanagers to estimate relative abundance and delin-eate management unit boundaries of other species,given that similar data is collected from severalradiocollared animals.

    Managers also can use these techniques to esti-mate regional population size for management,

    http:///reader/full/(r=0.22http:///reader/full/(r=0.22
  • 7/28/2019 Habita Abundance for Bobcats

    9/10

    Habitat-relativeabundance relationship for bobcats Nielsen and Woolf 229especially to initiate a harvest season. It is impossi-ble to determine precision of population estimatesbased on this technique, and the technique is notuseful to track short-term changes in populationsize. However, a habitat-relative abundance rela-tionship can serve as a foundation for a populationmodel because appropriate data is frequently rarefor bobcats over large scales. Given the quantifica-tion of demographic (e.g ., survival, reproduction)and spatial characteristics (e.g.,juvenile dispersal).such a model could be used to predict populationresponse to harvest and potential gene flow amongsubpopulations.

    Acknou~ l e d gmen t s . This study was fundedthrough the Federal Aid in Wildlife Restoration pro-gram (Project W-126-R, Status of the Bobcat in Illi-nois), and the Cooperative Wildlife Research Labo-ratory at Southern Illinois University at Carbondaleand the Division ofwildlife Resources of the IllinoisDepartment of Natural Resources cooperated. Wethank B. Bluett of the Illinois Department of Natur-al Resources for support and encouragement. C .Greene, D. Kennedy,J. Kolowski,M.Krecja,C. Schiel-er, C.Vincent,J.Waddell, and M.Woodruff providedfield assistance with bobcat trapping andradiotelemetry. We thank all cooperating trappersand landowners for their assistance and access toproperty. M. Chamberlain, D. Diefenbach, and ananonymous reviewer provided helpful commentson an earlier draft of this manuscript.

    Literature citedAEHIS(.HER,. J., l? A. ROBERTSO[\.ZI I R. E. KE~W AR D.993. Com-

    positional analysis of habitat use from animal radio-trackingdata. Ecology 74: 1313-1325.

    AGEE.. K.. S. C. I: 51'1'1-1;M. NYQL-IST,SD R. KC)OT,1989. A geo-graphic analysis of historical grizzly bear sightings in theNorth Cascades. Photogrammetric Engineering and RemoteSensing 55: 1637-1642,

    A~.ILYTI(.AIO ~ A R E .996. STMISTIX for windows user's man-ual. Version 1. 0. Analytical Software. Tallahassee. Florida.L'Sh.

    ASDERSO~.. Itl. 198'. A critical review and annotated bibliog-raphy of literature on the bobcat. Colorado Division ofWildlife. Special Report Number 62. Denver,USA.

    CIARK.. D.. J. E. DINN, ANT)K. G. SMITH.1993. A multivariatemodel of female black bear habitat use for a GIS. Journal ofWildlife Mat~agement5: 519-526.

    COYNE:R.. M..B. D. LEOPOII).4h1) M. J. CKL\IHFRIAIU.001. Mul-tivariate habitat models for bobcats in southern forestedlandscapes. Pages il-55 in Current bobcat research andimplications for management-symposium proceedings. TheWildlife Society 12-16 September 2000. Nashville, Ten-nessee. USA.

    C ~ R S I .. E. DYPRE,NI) L. BCIITANI.999. A large-scale model ofwolf distribution in Ital>-for conservation planning. Conser-vation Biology 13:150-159.

    DIEFENBACH,. R.. M. J. COYROY,. J. WARREN.. E. JA%II;S.. A.BAKER.ND T. HON. 1994. A test of the scent-station surveytechnique for bobcats. Journal of Wildlife Management 58:10-17.ILLINOISEI~ .~R TM EN TF N.Y~L-R~LESOLTRCES.996. Digital data setof Illinois.CD-ROM. Volume 1. Illinois Geographic Informa-tion System.Springfield.L'SA.

    JENNRICH.. I. , .LYDE B. TLRSER.1969. Measurement of non-circular home range. Journal of Theoretical Biology 22:22'-23'.

    KEWARD,. E. .LYD N, A. HODDER.996. Software for analyzinganimal location data (RANGESV). Institute ofTerrestria1Ecol-og): Wareham, IJnited Kingdom.

    KUICK,. T.. AYD D. L. D Y ~R .99-. Distribution of black-tailedjackrabbit habitat determined by GIS in southwestern Idaho.Journal of Wildlife Management 61:'i-85 .LENTH,R. V 1981. On finding the source of a signal. Techno-n~et rics23: 149-154.LOWLO, M. J.. ti. L. STORM,. S. KLTE. AND W. hf. TZILKOU~SKI.

    2001. Multivariate models of bobcat habitat selection forPennsylvania landscapes. Pages 4-17 in Current bobcatresearch and implications for management-symposium pro-ceedings. The Wildlife Socieh, 12-16 September 2000.Nashville.Tennessee. IJSA.

    L u h m . D.. M. JOSELYN,ND L. SL-LOWA~.996. Critical trendsassessment project: landcover database. Illinois Natural His-tory Surve); Champaign, USA.

    IMAEHR.. S.. 4 h ~ J .A. COX. 19 9i . Landscape features and pan-thers in Florida. Conservation Biology 9: 1008-1019..MANL\~.B. E J. 1986. Multivariate statistics: a primer. Chapmanand Hall. New York. New York, USA.MCGARI(~.%L... ~ Y DB. J. MARKS. 1995. FRAGSTATS: spatial pat-

    tern analysis program for quantlfimg landscape structure.Version 2.0. United States Forest Service General TechnicalReport PNW-GTR-351. Portland. Oregon. USA.

    ML~DENOFI;. J.. T. A. SILKIEY. G. HAIGHT,AN D A. l? WYDEVEU.1995. A regional landscape analysis prediction of fdvordbfegray wolf habitat in the no rthern Great Lakes region. Con-sewation Biology 9: 2'9-294.

    MOHR,C. 0. 194T. Table of equivalent populations of NorthAmerican small mammals. American Midland Naturalist 3:233-249.

    NAMS.I. 0. 1990. LOCATE11 user's guide. Pacer.Truro.Nova Sco-tia, CanadaNEELE;. D.. AN D C. G. HEISTER.987. The natural resources of

    Illinois:in troduction and guide. Illinois Natural Histon- Sur-vey>Special Publication 6. Champaign, rSA.

    NIELSEN,. K. 2000. Habitat use and populat ion dynamics ofbobcats in southern Illinois. Dissertation. Southern IllinoisUniversity at Carbondale, Carbondale. USA.

    NIEI.SEN,. K . , AND A. WOOLF.2001. Spatial organization of bob-cats (Lynx rz l f i~s)n southern Illinois. American Midland Nat-uralist 146:43-52.

    PALMA,.. l? BEK~.IUD M. RODRI(;T-ES,999. The use of sightingdata to analyse Iberian lynx habitat and distribution. Journalof Applied Ecology 36:812-824.ROLLEY,. E. 1987. Bobcat. Pages 672-681 in M. Novak, J. A.Baker. M. E. Obbarci. and B. Malloch, editors. Wild fi~rhearerconservation and management in North America. OntarioMinistry of Natural Resources.Toronto. Canada.

  • 7/28/2019 Habita Abundance for Bobcats

    10/10

    230 WfIdlijeSociety Bulletin 2002,30(1):222-230KOI.OFI... J ., hY1) J. B. H.~LFI.ER.1997. Establishing population

    viability planning objectives based o n habitat potentials.Wildlife Societ)- Bulletin L7:X95-9O/I.

    R o s e s i ; ~ ~ ~ , A. VCTOOLF. dcnsi-. L. . A \ I ) 1998. 1I;lbitat-pop~~lationty relationships for white-tailed deer in Illinois. Wildlife Soci-e h Bulletin Lh:2 5 2 - 2 5 8 .

    S t ~ \ l i \ .D. E.. J. J. hlll.l.~P.i~