Use of Non-Invasive Surveys to Validate Predicted Bobcat (Lynx rufus) Habitat Distribution in Wisconsin from Landscape-Scale GIS Information

Abstract

There has been mounting interest in Wisconsin to expand the bobcat harvest zone and a concomitant interest in increasing quota levels. However, bobcats are sensitive to over harvest and in Wisconsin current harvest boundaries are relatively arbitrary, are not defined by ecological boundaries, and do not adequately reflect the bobcat’s statewide distribution. Management decisions are hampered by a lack of information concerning bobcat-habitat relationships, bobcat distribution, and bobcat abundance due to the lack of reliable survey methods. I developed a model to predict bobcat habitat distribution throughout Wisconsin and compared the effectiveness of non-invasive survey methods (winter track-counts, hair-snares, and a scat-sniffing detector dog). Prior to the hair-snare survey, I conducted scent trials with captive bobcats to determine which scent-lure elicited the most interest and a rubbing response. For the first part of my study, I developed a predictive spatial distribution model using a multivariate distance statistic (Penrose distance) with adaptive kernel core areas estimated from over 1,000 locations from 10 radio collared female bobcats collected between 1991 - 1999 from three study sites in northern Wisconsin. The model identified 3 habitat variables that were highly correlated with bobcat habitat use: percent upland forest cover, percent forested wetland cover, and density of edge habitat. The model was applied statewide to hexagons of 4.5 km2 (mean size of a bobcat core area) using more recent land-cover information (2001) to identify high, moderate, and low levels of bobcat habitat suitability. The model classified 20% of Wisconsin as highly suitable, 17% moderate, and 63% low. The predictive accuracy of the model was evaluated with 5 independent data sets including bobcat scat obtained by a detector dog, noninvasive hair snares, sighting data, harvest data and track surveys. Sample sizes were small, and lacked the power to detect any significant relationships between predicted habitat quality and any of the survey techniques. The only exception were sightings of bobcats south of the harvest zone (N= 64) where 45% of the sightings occurred in highly and moderately suitable habitats which comprised only 27% of the area. (p < 0.001). For the second part of my study, I compared detection rates, cost, and field time required to complete non-invasive techniques using winter track counts, hair-snares, and a detection dog trained to locate bobcat scat. I also tested for correlation of results between the 3 techniques. Each of 10 track count routes (16 km long) were surveyed with 2 hair snare transects (each comprised of 10 stations along a 4.5 km route) and one 2-km long detector dog transect. In addition, 1-2 additional hair snare transects were established between 5-16 km apart from the winter track route. Compared to winter track counts (only 5 could be run in 2007), the detector dog found 4 times the number of bobcats (12 vs 3) while only surveying 12.5% of the length of a track count survey. The hair snare method produced no detections of bobcats. The 37 hair samples that were collected were primarily from rodents, flying squirrels, and dogs. Including materials, labor, and travel, the detection dog was the most expensive method per transect ($397) while the winter track-counts were the least expensive ($180-$260). However, the cost per detection of the detector dog ($330) was comparable to the winter track counts ($300-433) due to its high detection rate. All three survey methods required comparable field times. For the scent trials, I compared six scent lures: 1) Russ Carmen’s Canine Call, 2) Lenon’s Nature Call, 3) Lenon’s Super All Call, 4) O’Gorman’s Powder River Cat Call, 5) beaver castor, imitation catnip oil, and catnip, and 6) bobcat urine (BU) on 17 captive bobcats housed in 5 different facilities in Wisconsin and Florida. I recorded behavioral responses to the lures including sniff, rub, roll, and reach. A higher proportion of bobcats responded to scent lures than the control (P<0.05) however there was no difference (P > .05) in the proportion of bobcats that reacted when the 6 scent types were compared. Bobcats did sniff, rub, and roll more frequently on Lenon’s Super All Call (P<.05) than the other lures. However, of the 17 bobcats, only 74% sniffed, 49% rubbed, and 41% rolled in response to the lure. Predictive distribution models, combined with reliable non-invasive survey methods have the potential to aid wildlife managers in better understanding and managing bobcat populations that are difficult to survey on a landscape scale.