Study areas

We obtained nuisance bear relocation and tag recovery data from the Ontario Ministry of Natural Resources (OMNR) for three jurisdictions in Ontario: Chapleau, Sudbury and Parry Sound (Fig. 1). Chapleau is a remote town located in north-central Ontario with a population of approximately 3,000 people (see Fig. 1). Immediately north of town lies the 7,000-km² Chapleau Crown Game Preserve. At the time of the study, hunting and trapping of wildlife was not permitted within the preserve. The Chapleau study area is located in the Boreal Forest (Rowe 1972), dominated by balsam fir Abies balsamea, black spruce Picea mariana and white birch Betula papyrifera. Jack pine Pinus banksiana is the dominant tree species on coarse mineral soils.

At the time of our study, the City of Sudbury (see Fig. 1) and surrounding communities had a population of about 162,000. Sudbury is on the northern edge of the Great Lakes-St. Lawrence Forest (Rowe 1972). Extensive disturbance from smelter operations in the early and mid 1900s greatly reduced the abundance of native flora near the city. Tree cover is dominated by hardy early successional species, such as trembling aspen Populus tremuloides, balsam poplar P. balsamifera, and white birch (Amiro & Courtin 1981). Jack pine occurs frequently on sand flats and other coarse textured soils. Farther from the city, forest cover is more representative of the Great Lakes-St. Lawrence Forest and includes northern red oak Quercus rubra, eastern white pine Pinus strobus, red pine Pinus resinosa, and red maple Acer rubrum.

Table 1.

Time periods and sample sizes for nuisance black bear relocation data sets from the Chapleau, Sudbury and Parry Sound study areas in Ontario, Canada. Data did not include cubs or dependent yearlings and each bear was only included once.

Parry Sound is located on the shore of Georgian Bay of Lake Huron (see Fig. 1). Including small rural communities, the Parry Sound area contains roughly 18,000 permanent and 66,000 seasonal residents. Parry Sound is in the Great Lakes-St. Lawrence Forest (Rowe 1972), characterized by a mixture of sugar maple Acer saccharum, American beech Fagus grandifolia, basswood Tilia americana, yellow birch Betula alleghaniensis, eastern hemlock Tsuga canadensis, eastern white pine, red maple, northern red oak, and white ash Fraxinus americana. White spruce Picea glauca is common on sand flats.

Data collection and analysis

Each of the three OMNR jurisdictions provided data for bears that were involved in human-bear conflicts and subsequently captured, transported and released at new locations (i.e. relocated). The OMNR captured nuisance black bears in traps or by darting free-ranging bears with immobilization agents. They immobilized captured bears using a mixture of ketamine hydrochloride and xylazine hydrochloride. Capture personnel extracted a premolar from immobilized bears for age determination (Johnston et al. 1987). Bears were ear-tagged with a unique identification number and tags were recovered on a voluntary basis from animals harvested by hunters, killed in vehicle collisions or destroyed as nuisance animals. The spring bear hunt was open from mid-April to mid-June, whereas the fall hunt started at the beginning of September and ended in mid-October.

The nuisance bear data sets included information on age, sex, capture location, release location, recovery location, and associated dates. The years of data collection varied among jurisdictions and ranged from 1982 to 1997 (Table 1). We excluded cubs and dependent yearlings from all analyses, as their movements were dependent upon those of their mothers. For age class comparisons, we categorized bears <4 years of age as juveniles and those ≥4 years as adults (Kolenosky 1990, Stiver 1991). For bears that were captured multiple times, only data from the first capture was used unless otherwise stated. Of the records from the Sudbury area, 14 did not include the exact capture, release or recovery locations. For these records, we used the center of the recorded township for movement analysis. Townships in this part of Ontario were approximately 100 km² (10 × 10 km). All distance calculations presented are straight-line estimates. We calculated the time in days from release to tag recovery, and then investigated the relationship between relocation distance and time to tag recovery for adults and juveniles using Spearman's rank correlation coefficient (r_s).

Recurrence of nuisance behaviour

We used nuisance recurrence rate to refer to the proportion of bears involved in human-bear conflicts subsequent to the initial relocation. This included all relocated bears recaptured or killed because of nuisance behaviour, irrespective of where it occurred. We excluded all bears relocated < 21 km from their capture site. We also excluded bears captured in the Sudbury District during 1997 because the database did not include recovery information after 1997. The resulting data set of 123 individual bears included 27 adult females, 23 adult males, 23 juvenile females and 50 juvenile males. We estimated minimum nuisance recurrence rates by age and sex categories, and then compared mean time to tag recovery for bears involved in nuisance recurrences to other recovery methods using an independent means t-test.

Post-relocation return

Extensive seasonal movements are frequently observed in both male and female black bears (Kolenosky & Strathearn 1987, Rogers 1987, Samson & Huot 1998), suggesting that the area bears are familiar with is substantially larger than their core home range. We identified a relocated bear as a returning animal if it was recovered ≤ 20 km from the capture site. We assumed that within a 20-km distance bears had returned to familiar territory, and were capable of returning to the capture site. In support of our chosen threshold distance, the average maximum distance between any two telemetry locations for 11 radio-collared bears in the Sudbury area was 20.70 km (SD = 9.13; L.J. Landriault, unpubl. data), and the average minimum convex polygon (MCP), including all locations for each animal, was 90.8 km² (SD = 51.2; Landriault 1998). Similar data were not available for the Chapleau and Parry Sound areas. However, based on the diversity of berry and hard mast crops, we would expect MCP home ranges to be larger in Chapleau and smaller in Parry Sound. Only bears that were recovered were included in our analysis of post-relocation return. We refer to the proportion of recovered bears that returned to within 20 km of the capture site as the return rate. We excluded from the analyses bears transferred < 21 km from their capture site. A bear was only identified as an animal that did not return to the capture area if it was recovered ≥ 20 days after release, to ensure the animal had sufficient time to return. We considered 20 days from relocation to recovery a sufficient period given the average relocation distances used in the three study areas (68–80 km). Radio-telemetry studies have shown that black bears home quickly and are capable of movement rates in excess of 10 km/day (Rogers 1986b, Landriault 1998).

We calculated the proportion of recovered bears that returned and then determined whether the return was affected by study area using logistic regression. We then developed logistic regression models to evaluate the relationships between post-relocation return and age, sex and relocation distance. We modeled all possible subsets of the three predictor variables, without interaction terms. We were interested in determining the shortest relocation distance that would be effective in removing bears from an area but wanted to include each individual bear only once. Therefore, for bears relocated multiple times, we included the record with the shortest distance from which they did not return. In cases where a bear returned after one relocation and not after the next (or vice versa), we used the record that was first chronologically. There were too few bears in the data set with multiple captures to consider prior nuisance behaviour or relocation experience as a factor in the models. A second data set was obtained from Sudbury OMNR and used for model validation only. The data included capture and recovery information for 26 individuals relocated from 2000 to 2002, comprised of five adult females, six adult males, seven juvenile females and eight juvenile males. The average relocation distance in the validation data set was 73 km (SD = 29.1, range: 21–150 km).

We used the information-theoretic approach described by Burnham & Anderson (2002) to guide model selection and inference based on the second-order Akaike's Information Criterion corrected for small sample sizes (AIC_c). We calculated the differences between the AICC of each model and that of the best approximating model (Δ_i) and AIC_c weights (w_i). AIC_c weights present the relative evidence that model i is the best model among those being compared (Burnham & Anderson 2002). We used the sum of wi to assess relative importance of variables as the number of models that included each variable were equal. A model is considered to be the best predictor when the AIC_c weight is > 0.90 (Burnham & Anderson 2002). When none of the models was clearly the best predictor, we generated model-averaged estimates and associated unconditional standard errors (Burnham & Anderson 2002). The effect of our independent variables on post-relocation return is discussed in terms of odds ratios. Odds ratios were calculated by raising the base of the natural logarithm to the power of the value of model coefficients. We subsequently produced graphs illustrating the probability of a bear returning to the capture site based on age, sex and relocation distance using the following equation:

We quantified predictive accuracy of the models using receiver-operating characteristic (ROC) curves. We then calculated the area under the curves (AUC) for the modeling and validation data sets. The ROC curve represents the relationship between sensitivity (i.e. proportion of animals expected to return that did return) and 1 - specificity (i.e. proportion of animals expected to return that did not). AUC values can range from 0 to 1.0. A value of 0.5 describes a model with discrimination ability no better than random, whereas a value of 1.0 indicates a model with perfect discrimination ability (Pearce & Ferrier 2000). AUC values can also be interpreted as the probability that a model will correctly distinguish between two observations, one positive and the other negative (Pearce & Ferrier 2000). For models where relocation distance odds ratios were significantly different from 1, we used the modeling data set ROC curve to estimate relocation distance thresholds (Guénette & Villard 2005). We chose the threshold that yielded the maximum accuracy, that is, where the sum of sensitivity and specificity was maximized. We used SPSS software (SPSS Incorporated 2005) for all statistical analyses.

Table 2.

Age and sex composition of tagged and recovered relocated nuisance black bears from the Chapleau, Sudbury and Parry Sound study areas in Ontario, Canada. Data did not include cubs or dependent yearlings. AF = adult females; AM = adult males; JF = juvenile females; JM = juvenile males.

Recurrence of nuisance behaviour

A minimum of 30% (37/123) of relocated bears were subsequently involved in at least one other nuisance event. Adult females were frequently involved in nuisance recurrences (48%), whereas this behaviour was rare in juvenile males (18%). A minimum of 39% of adult males and 26% of juvenile females were involved in nuisance recurrences. Of relocated bears, 57% were involved in subsequent nuisance behaviour in the same year, i.e. prior to hibernation, and 81% within one year (365 days). The mean number of days to recovery for bears recaptured or killed for subsequent nuisance behaviour (165 days, SD = 180.4) was significantly less than the mean number of days to recovery observed for bears recovered as a result of harvest or vehicle collisions (330 days, SD = 374.4; ln transformation, t = -2.516, df = 69, P = 0.012). Time to recovery was longest for bears captured in July, irrespective of whether they were recovered as a result of repeat nuisance behaviour or harvested (the two primary recovery methods).

Post-relocation return

Relocated adult black bears consistently had much higher return rates than juveniles (73 vs 29%; Table 3), suggesting that factors affecting return rates differed between age classes. Therefore, we developed separate models for the two age classes but still included the continuous age variable within the juvenile and adult models. Study area had no effect on whether a bear returned to the capture area (-2 log likelihood = 96.028, likelihood ratio χ² = 2.384, df=2, P = 0.304), so we pooled data from the three areas. The resulting data set included 17 adult females, 16 adult males, eight juvenile females and 30 juvenile males.

Table 3.

Proportion of recovered nuisance bears that returned to within 20 km of their capture area in the Chapleau, Sudbury and Parry Sound study areas in Ontario, Canada.

Table 4.

Differences in Akaike's Information Criterion between the i^th and top-ranked model (Δ_i) and Akaike's weights (w_i) for all models predicting the probability of return of relocated nuisance black bears in Ontario, Canada. All models include a constant.

Juveniles

Based on the sum of AICC weights of the individual predictor variables (Table 4), it appears that age was the most important variable in predicting return in juvenile bears (Σw_i = 0.8), followed by relocation distance (Σw_i = 0.6) and sex (Σw_i = 0.5). No model was clearly the best predictor of the probability of post-relocation return for juvenile bears, and the largest AIC_c weight (w_i) was only 0.239 (see Table 4). Therefore, we used a model-averaging approach. Based on model averaging, the odds of returning to the capture area were roughly four times greater for females than for males, and the odds of returning increased by approximately three times for each year of age (Table 5). Relocation distance had a negative effect on the probability of return for juvenile animals with the odds of return decreasing by 3% for every 1 km increase in relocation distance. The AUC value for the modeling data set was 0.822 (SE = 0.080) and the AUC value for the Sudbury validation data set was 0.750 (SE = 0.164), indicating that the averaged model correctly discriminated between returning and non-returning bears 75–82% of the time. The high standard error on the AUC value for the validation set is a reflection of the high variability around the coefficients of the independent variables in our averaged model (see Table 5). Our model achieved maximum accuracy when bears with a probability > 38% were predicted to have returned to the capture area. Using Figure 2, we obtained a minimum recommended relocation distance (i.e. the relocation distance corresponding to a 38% probability). For male 2- and 3-year-old bears, this resulted in relocation distances of 30 km and 64 km, respectively. The threshold values were 73 km and 106 km for females of corresponding ages (see Fig. 2). Caution should be used in interpreting return probabilities and relocation distance thresholds for juvenile females, as the sample size was small relative to juvenile males (N = 8 and N = 30, respectively).

Table 5.

Model-averaged coefficients (β), unconditional standard errors (SE) and odds ratios for models predicting the probability of return of relocated juvenile and adult nuisance black bears in Ontario, Canada. For Sex: female = 1, male = 0; the relocation distance is in km.

Figure 2.

Effects of sex and distance on the post-relocation probability of return for 1-year-old (A), 2-year-old (B) and 3-year-old (C) nuisance black bears in Ontario, Canada. Black symbols represent male bears and gray symbols represent females. Distance is presented in kilometers. Estimates of return were based on recovery of tagged bears. Bears were identified as returning animals if they were recovered within 20 km of their capture site.

Figure 3.

Effect of age on the post-relocation probability of return for relocated adult nuisance black bears in Ontario, Canada. Distance is presented in kilometers. Estimates of return were based on recovery of tagged bears. Bears were identified as returning animals if they were recovered within 20 km of the capture site.

Management implications

Relocation may be a useful management tool for the removal of juvenile animals involved in human-bear conflicts. Juvenile males generally make up the majority of nuisance bears (Harger 1970, Rutherglen & Herbison 1977, Rogers 1987, Stiver 1991, Schwartz & Franzmann 1992, Shull 1994) and are least likely to return to the capture area or be involved in subsequent nuisance events. Based on our analyses, we expect that 2- and 3-year-old male bears will not return to within 20 km of the capture area if relocated a minimum of 30 km and 64 km, respectively. We anticipate threshold distances to differ in areas with more extreme physiography or more significant anthropogenic barriers than those in our study areas. Although most juvenile male bears may not return to the capture area, it is worth noting that they are also unlikely to remain near the release area and often demonstrate extensive post-relocation movements (Harger 1970, Rogers 1984, Linnell et al. 1997, Landriault 1998).

Based on our analyses, relocation may be ineffective in managing adult animals involved in human-bear conflicts. A large proportion of adults returned to within 20 km of the capture area, and recurrence rates of nuisance behaviour were high. Based on our results, if 100 adult female bears were relocated, a minimum of 50 would be involved in a subsequent nuisance occurrence, and 40 would do so within one year. All of our study areas were adjacent to large forested areas, such that nuisance bears likely made up a small proportion of the black bears on the landscape. However, jurisdictions with a policy of destroying bears involved in more than one nuisance incident should take into consideration the number of adult females being destroyed and the potential effect on local bear populations. Given the low success rates for relocation of adult bears, we suggest that nuisance bear management strategies emphasize the management of anthropogenic food sources and take into consideration the efficacy and acceptability of alternative approaches (e.g. aversive conditioning) for managing persistent individuals.