Ursus arctos, Linnaeus, 1758, Linnaeus, 1758

Parsons, Bethany, Wilson, Abbey E., Graham, Karen & Stenhouse, Gordon B., 2023, Grizzly bear (Ursus arctos) movements and habitat use predict human-caused mortality across temporal scales, Canadian Journal of Zoology 101 (2), pp. 81-94 : 85-89

GPS location data from a total of 23 grizzly bears

residing in Alberta, Canada, from 2005 to 2021, were used to determine whether movement and habitat use differed between bears that survived and those that died over various temporal scales. Not all bears had data from both the year of death and the preceding years; therefore, some of the bears used in the 2–4-year time block were different from the 1 year and 1 week time blocks ( Tables 2 and 3 ; Figs. 1 and 2 ). However, all three time block analyses and the countdown to death analysis used a paired sample of nine bears (nine bears that died paired with nine bears that survived). We found significant differences at all three time scales in the time block analyses ( Figs. 5 and 6 ), and the countdown to death analysis demonstrated that differences in behaviour between bears that died and bears that survived increased as the date of death approached ( Fig. 7 ).

Movement

Time block analyses

At the largest time scale (2–4 years prior to death), grizzly bears that died were overall more diurnal than bears that survived, although they were also more likely to have low diurnality than more moderate values, resulting in a J-shaped curve ( Table 4 ; Fig. 5A ). There were no significant differences in displacement. Similarly, 1 year prior to death, diurnality was significantly higher in bears that died than those that survived ( Table 4 ; Fig. 5B ). During this time, no significant difference in displacement between bears that survived and bears that died was observed. Conversely, 1 week prior to death, there was no significant difference in diurnality between bears that died and those that survived; however, displacement was significantly greater in bears that died compared with those that survived ( Table 4 ; Fig. 5C ).

Countdown to death

When modelling changes in movement over time as death approached, we found that although bears that died were overall more diurnal than bears that survived over the 60 days prior to death, there was no change in diurnality as death approached ( Table 5 ; Fig. 7A ). However, displacement increased exponentially in grizzly bears that died as death approached, but not in bears that survived ( Table 5 ; Fig. 7B ).

Habitat use

Time block analyses

Grizzly bears that died were found to have significantly different patterns of habitat use than bears that survived 2–4 years and 1 year prior to death ( Table 6 ; Figs. 6A–6D ). Bears that died used habitats with higher risk values and low-to-moderate RSF values. In addition, these variables interacted; the highest likelihood of anthropogenic mortality occurred in bears that used areas that were simultaneously low RSF and high risk ( Table 6 ; Figs. 6A–6D ). Similar patterns in risk and RSF were observed 1 week prior to death; however, there was no significant interaction between risk and RSF ( Table 6 ; Figs. 6E and 6F ).

Countdown to death analysis

We found that the use of risky habitat was significantly higher in bears that died compared with those that survived, indicating that bears that died used overall higher risk habitats, and this difference increased as death approached ( Table 7 ; Fig. 7C ). Similarly, bears that died used lower RSF habitats than bears that survived throughout the time frame, and this difference increased as death approached ( Table 7 ; Fig. 7D ).

Discussion

As bears are wide ranging and can travel many kilometres and through many different habitat types in a short time, the location of death tells us little about a grizzly bear’s habits and activities beforehand. Therefore, we examined the movement and habitat use of grizzly bears that died to understand temporal scales of behaviour and how it differed from bears that survived. We found evidence of significant differences in each behaviour variable between bears that died and bears that survived, rejecting the null hypothesis that grizzly bear deaths are random and unrelated to movement and habitat use. Three out of four behaviour variables differed 2–4 years and 1 year prior to death, supporting hypothesis 1 that grizzly bears that died exhibited long-term patterns of risky behaviour. In addition to displaying overall riskier behaviour than bears that survived, we found evidence that some behavioural differences were only observed 1 week prior to death and (or) increased as death approached, supporting hypothesis 2 that bears died during a period where they were exhibiting risky behaviour compared with their individual normal patterns of movement and habitat use. Our data suggest that although the likelihood of mortality was related to some long-term differences in behaviour between bears that died and bears that survived, these differences increased prior to death due to short-term changes in individual behaviour.

During the 2–4 years prior to death and 1 year prior to death, we found that bears that died were more diurnal than bears that survived. This difference was not significant 1 week prior to death and there was no evidence of an increasing trend in diurnality as death approached. We expected this behaviour to be evident at finer time scales, but the small number of GPS locations for the last week and countdown to death analyses may have precluded significance. As most human activity in grizzly bear habitat occurs during the daylight, diurnal behaviour increases the likelihood of human– grizzly bear interactions. In fact, grizzly bears have previously been found to be more nocturnal in areas of high human activity, suggesting a temporal avoidance strategy ( Gibeau et al. 2002; Hertel et al. 2017). This association between habitat use and diurnality may explain the J-shaped pattern found in the 2–4-year time block analysis, where both low and high diurnality were associated with increased likelihood of death. Bears that died that had lower than average diurnality may have been compensating for using high-risk human use areas. Alternatively, human mortality due to vehicle collisions may be more likely in bears at night due to limited visibility.

There was no significant long-term difference in displacement between bears that survived and bears that died 2–4 years and 1 year prior to death. However, bears that died were found to displace more 1 week prior to death. This result was supported with the countdown to death analysis showing an exponential increase in displacement shortly before death. Increased displacement is typically associated with traveling and foraging behaviour ( Hertel et al. 2019; Stenhouse et al. 2022) and can increase the chance of being seen or encountering people. Furthermore, a previous study on this population of grizzly bears showed that they select open areas that are visible to roads when traveling ( Parsons et al. 2020), which may put them at further risk for a mortality event. The results from these two movement variables highlight that both of our hypotheses contribute to variability in mortality risk due to behaviour; grizzly bears that died displayed increased diurnality over the long term and increased displacement on a short-term basis. This suggests that diurnality may be more central to a bear’s individual personality because it was seen in the long-term analysis, whereas bears change their displacement as a short-term response to different situations.

While there is well-documented data on how grizzly bears use habitat and that grizzly bear deaths occur in high-risk areas ( Nielsen et al. 2004 b, 2010), we aimed to determine how the use of these layers differed between grizzly bears that died and bears that survived and at what temporal scales. We found that bears that died used higher risk areas than bears that survived at all time scales, and this difference became more pronounced as death approached. Modelled high-risk habitats are often close to areas of high human use, which can lead to legal and illegal harvest being destroyed because of encounters with human habitation, property, or in response to self defense, and vehicular and (or) railway collisions ( McLellan et al. 1999; Nielsen et al. 2004 b). As the risk layer used has high densities of known mortality sources, it is not surprising that bears that died were found to have overall higher use of these areas than bears that survived, as well as particularly high use as death approached. While body condition was noted at the time of capture, it was not known at the time of a mortality event and may explain why some bears were willing to accept more risk to find food resources ( Boulanger et al. 2013).

Food resources and human-caused mortality are major factors in regulating grizzly bear populations, with food resources determining reproductive rates and human-caused mortality regulating survival ( Naves et al. 2003; Nielsen et al. 2010). Because many areas with high human activity also have good resources, grizzly bear habitat selection requires trade-offs between these two factors. The RSF layer incorporates measures relating to both resources and human activity, providing insight into this trade-off. Previous studies using this layer have shown that in our study area, the highest RSF habitats have both good resources and security (low risk), and very low RSF values largely represent remote alpine areas, which are simultaneously low resource and very low risk ( Nielsen et al. 2006). Our results show that grizzly bears that died were more likely to use low-to-moderate RSF habitats, which often represent areas that have good resources but low security, such as roadside vegetation and valley bottoms ( Nielsen et al. 2006). In addition, the interaction between risk and RSF 2–4 years and 1 year prior to death shows that grizzly bears that died were particularly likely to use habitats that were simultaneously low RSF and high risk. Whereas bears that survived were more likely to only use risky areas when the food resources made the risk worthwhile, bears that died used areas offering little habitat value while also being high risk. Grizzly bears may use habitat suboptimally in terms of anthropogenic risk and resources to avoid more dominant bears ( Stonorov and Stokes 1972; Rode et al. 2006). Female and subadult grizzly bears have been found to use areas of both higher anthropogenic activity and lower resources than adult males, which may be due to competitive exclusion as well as to reduce the risk of infanticide ( Rode et al. 2006). Small grizzly bears may also be displaced from the best habitats by larger grizzly bears ( Stonorov and Stokes 1972). More research on the role of social structure and dominance in grizzly bears is required to understand why some bears use suboptimal habitats.

Figures

Fig. 1. Home ranges of paired grizzly bears (Ursus arctos) used in the 2–4-year time block analysis that either died (A) or survived (B) in Alberta, Canada, from 2005 to 2021. Each pair has the same pattern across maps as shown in the legend. Shapefile for communities and protected areas from the Government of Alberta (https://geodiscover.alberta.ca/; Projection: NAD 1983 UTM Zone 11N). Shapefile for Canada boundary and Alberta boundaries from Natural Earth (https://www.naturalearthdata.com; Projection: WGS 1984).

Fig. 2. Home ranges of paired grizzly bears (Ursus arctos) used in the 1 year and 1 week time block analyses and countdown to death analysis that either died (A) or survived (B) in Alberta, Canada, from 2005 to 2021. Each pair has the same pattern across maps as shown in the legend. Shapefile for communities and protected areas from the Government of Alberta (https: //geodiscover.alberta.ca/; Projection: NAD 1983 UTM Zone 11N). Shapefile for Canada boundary and Alberta boundaries from Natural Earth (https://www.naturalearthdata.com; Projection: WGS 1984).

Fig. 5. Predicted likelihood of anthropogenic mortality in grizzly bears (Ursus arctos) for movement metrics during three time blocks prior to death. 2–4 years and 1 year prior to death, displacement was not found to influence likelihood of mortality and was excluded in the best model, and 1 week prior to death, diurnality was excluded from the best model. The y-axis is the predicted mean “risk score” with 95% confidence interval assigned by the conditional logistic regression model; this metric is on an arbitrary scale and represents relative risk of anthropogenic mortality.

Fig. 6. Predicted likelihood of anthropogenic mortality in grizzly bears (Ursus arctos) for habitat use metrics during three time blocks prior to death. 2–4 years and 1 year prior to death, Risk and RSF had significant interactions; therefore, the plots include how each metric affects the other. The y-axis is the predicted mean “risk score” with 95% confidence interval assigned by the conditional logistic regression model; this metric is on an arbitrary scale and represents relative risk of anthropogenic mortality.

Fig. 7. Results of the countdown to death analysis show how movement and habitat use change in the days leading up to anthropogenic death for grizzly bears (Ursus arctos) that died compared with matched bears that survived. “Days prior to death” was not included in the best model for diurnality. Error bars and envelopes represent the 95% confidence interval of the population mean.

Tables

Table 2. GPS location data from grizzly bears (Ursus arctos) that died of human causes and those that survived were used to determine differences in movements and habitat use 2–4 years prior to death.

Bear ID	Sex	Age	Pair	Years of data	Mortality date	Cause of death
G120	M	3–6	1	2012–2013, 2015	6 Nov. 2015	Illegal
G128	M	2–3	2	2013–2014	9 July 2014	Defense
G160	M	4–5	3	2015–2016	17 Sep. 2018	Defense
G204	F	2–3	4	2005–2006	12 Nov. 2008	Illegal
G223	F	20–22	5	2007–2009	28 May 2009	Vehicle
G260	F	5–11	6	2007–2013	23 Sep. 2013	Illegal
G285	M	2–3	7	2012–2013	17 May 2013	Illegal
G301	M	3–4	8	2014–2015	22 Sep. 2016	Vehicle
G365 a	F	14–15	9	2019–2020	1 Sep. 2021	Illegal
G129	M	4–5	1	2013–2014	–	–
G152	M	4–5	2	2013–2014	–	–
G142	M	5–6	3	2016–2017	–	–
G238	F	3–6	4	2007–2010	–	–
G023	F	20–22	5	2009–2011	–	–
G111	F	14–20	6	2008–2014	–	–
G284	M	3–4	7	2012–2013	–	–
G304	M	2–3	8	2016–2017	–	–
G355	F	9–10	9	2018–2019	–	–

Note: Bears that were also used in the 1 year and 1 week prior to death and in the countdown to death analyses are indicated

in bold text. Age is the range of ages for a bear that correspond to the years of data.

a G365’s collar was found cut off Sep. 2021. The bear could have been killed between Oct. 2020 and Sep. 2021.