Discussion

We predicted that years with low Canada lynx abundance in the boreal forest were associated with a reduction in the extent of the southern lynx range. We found support for this idea, as it appeared that the southern range dynamics were mostly driven by boreal lynx population dynamics. The probability of harvesting a lynx at the southern periphery was predicted by lynx harvest the year earlier in the core boreal range. We also predicted that areas with high human disturbance, shallow snow, presence of competitors, and with low connectivity to boreal lynx populations were less likely to be occupied as part of the southern range. However, we found that the anthropogenic effect of roads and competition from bobcats and coyotes did not seem to have influenced the southern range of the Canada lynx in Ontario. We did find that areas that had deep winter snow were often found within the southern lynx range. However, this relationship did not vary temporally with the area of the southern lynx range (Table 1). We used this snow depth predictor as an index of climate change, since we thought that the highest impact of climate warming on lynx would be related to the timing of molt of its main prey the snowshoe hare. We also thought that competition would arise in areas with less snow over time and would become more hospitable to coyote and bobcat. However, we found that the average annual snow depth was not driving the temporal dynamics of the lynx southern range in Ontario but is a habitat condition that determines whether lynx will expand into an area.
There was a weak signal for the temporal dynamics of the coyote, but we did not have enough power to detect a significant relationship given number of tests we performed (Table 1). It is quite reasonable to think that the coyote is a competitor because they are generally found across the southern range apart from a few areas within the boreal forest in Ontario (Figure S3). Bobcat on the other hand, occupied a very small area and generally occurred in the southeastern corner of the west and northeast zones (Figure S4). This spatial relationship itself indicates that the bobcat is not responsible for the range contraction in the southeastern zone, since it is rarely found there. Recent finer scale studies suggested segregation between lynx and bobcat (Marrotte et al. 2020b, Morin et al. 2020).
We observed that the southern range of the lynx has recovered from a dramatic decline in the late 1940s (Figure 4). It has never returned however, to the short-lived maxima we observed during the 1960s. More recently, in the period 2012-2017, the range has reached its maximum extent in the west and central zones. After the mid-1980s, the southern range varied less and from 2010 to 2017, seemed to be increasing. Consequently, we did not find the substantial range loss in more recent times that has been observed in parts of the contiguous United States (Ruediger et al. 2000), and as in a previous analysis in Ontario (Koen et al. 2014). The stable and somewhat increasing range in Ontario is not unique across the lynx range, since lynx are increasing in numbers in Maine, USA (Simons‐Legaard et al. 2016) and the lynx range in British Columbia has been stable since 1935 (Gooliaff and Hodges 2018). It is important to remember however, that the Canada lynx range across all of North America has contracted substantially from its presettlement extent (Laliberte and Ripple 2004).
The extensive contraction in the early part of the time series may have extended from 1938 to 1951 (Figure 2). In fact, de Vos and Matel (1952) noted that lynx occurrences were rare at this time and the range was also gradually shrinking. They attributed this decline to ecological changes and overharvesting. The decline prompted the closing of lynx trapping during 1951-1952 season and a quota system for lynx was established and trapping was reopened the next year (de Vos and Matel 1952). At the same time, in all of Canada, harvest dropped from 33,054 pelts in 1925 to only 3,734 lynx pelts in 1949 (de Vos and Matel 1952). Lynx fur returns for each jurisdiction in Canada were an order of magnitude lower during the population crash. In approximately the same period, lynx occurrences and harvest in Wisconsin, Minnesota and Michigan also dropped (McKelvey 2000).
Immediately after this large continental wide population crash and subsequent range contraction, the southern range in Ontario expanded almost 8-fold (Figure 4). The ranged peaked in 1963-1964 and lynx were being harvested more than 100 km south of the boreal forest in southern Ontario for almost 10 years (Figure 3). At the same time there was an increase in fur returns and occurrences of lynx in the Great Lakes states immediately south (McKelvey 2000). Similar range expansion and population increase were also present in Alberta, British Columbia, Saskatchewan, Manitoba and Quebec during this period (Todd 1985, McKelvey 2000).
These earlier large fluctuations of the southern lynx range in Ontario and harvest in the Great Lakes states were likely driven by immigration of lynx from the boreal forest (McKelvey 2000, Steury and Murray 2004, Murray et al. 2008). We do see this pattern in our analysis; the southern lynx range changes with the population dynamics of the boreal lynx and this influence decays away from the boreal forest (Table 1). Density dependent dispersal from the boreal forest likely drives the southern lynx range in the northern Great Lakes region. Consequently, southern populations appear to be maintained by emigration from sources in the range core (Steury and Murray 2004, Murray et al. 2008). During peak years, individuals venture south and colonize marginal habitat outside of the boreal forest in Ontario and eventually reach the northern Great Lakes states (Mech 1973, Mech 1980). In more recent times in Ontario, the boreal lynx cycle did not reach extremes as it once did (Figure S5), therefore lynx populations south were no longer being rescued.
There was a period of slow range contraction from 1970 to the late 1990s, where lynx appeared and quickly disappeared from their southern Ontario range. This period is not unique to Ontario, most jurisdictions followed the same pattern (McKelvey 2000). In an earlier study, Koenet al. (2014) noted that the largest range loss happened in this period, but we did not see a continuous decline after 1991 as they did. In fact, the range expanded, and the west and northeast zones were at their largest possible extent and occupied a combined area that was previously unforeseen (Figure 4). Our results probably differ because we were able to assess a longer time series (1972-2010 vs. 1948-2017) and we examined a much larger area.