Results

The model that could best account for the effort of harvesting a Canada lynx and the spatial temporal process in the southern periphery of Ontario, Canada included the log transformed total number of furbearers harvested and a thin plate smoother on the spatial coordinates (Table S1). This model was 20.914 AIC units lower than all other models and its AICw was 1.000. The spatiotemporal effort model had an adjusted R2 of 0.586 and a deviance explained of 54.1%. Other than the total number of furbearers harvested, the 3 other effort related predictors followed linear relationships (Figure S1). The additive effect of the number of trapping units, the total area, and the average price were not as important than the total number of animals harvested. The probability of harvesting a lynx decreased with the total area harvested while the 3 other predictors had a positive relationship. Also, the influence of lynx pelt price was weak compared to the other predictors.
The probability of harvesting a Canada lynx south of the boreal forest across Ontario changed through time (Figure 3). During the late 1940s and the early 1950s, the likelihood of harvesting a lynx was at its lowest. However, in the mid-1960s the odds peaked across the southern range and even trapping units found in the east had a high probability. After this peak lynx period, it then became unlikely to harvest a lynx in the east and this pattern continued to 2017. The odds of harvesting a lynx peaked in both the west and central zones in the early 1960s and again in the mid-1970s, then declined until the 2000s and increasing slowly until 2017 to an overall probability of harvest higher than in previous years.
To get a better idea of the range dynamics, we calculated the occupied area of the southern range of each zone for each year (Figure 4). In 1950 the total extent of the southern lynx range was at its lowest and occupied a total are of 19,118.2 km2. The extent of the range peaked between 1963-1964 and occupied a maximum area of 147,483.5 km2. This was an area 7.7x larger than during the crash in the late 1940s. From 1970 onwards, the southern range varied much less in size compared to previous years. It declined between 1970 and the late 1980s, but gradually increased until 2017 to a size comparable to the early 1970s. There were also a few notable decreases in range in the periods 1965-1972, 1983-1992 and 1995-2002.
In general, all 3 zones (west, central, and, east) followed similar patterns. However, from 1957 to 1964 the east zone increased from 5,462.4 to 30,042.9 km2, which was a 6-fold increase and occupied most of the Lanark and Renfrew County just east of Ottawa (Figure 3). This increase was not as dramatic in the west and central zones, where there was only a 1.5- and 1.3-fold increase. Although, these two northernmost ranges were already closer to their maximum extent of 65,548.2 and 54,623.5 km2, consequently they could not have increased as intensely during this period. A smaller range contraction in this same Lanark and Renfrew County area also occurred in 1971 to 1973.
From the late 1950s to 2017, the west and central zones varied by 24,580.6 [40,967.6-65,548.2] and 16,387.0 km2[38,236.5-54,623.5]. The west zone reached its maximum area more recently in 2013 and 2017, whereas the central zone reached its maximum area multiple time in the periods 1960-1967 and 1970-1976. The east zone varied quite differently. It increased dramatically twice in the period 1959-1973 and never reached these levels again. After this point the range varied between 0 and 8,193.5 km2.
We calculated the number of years each sampling unit was within the lynx southern range. Sampling units in the south were less frequently part of the Canada lynx range (Figure S2). In the east zone, sampling units were only part of the range on average 8.7 years over the 70-year time period. In contrast, the west and central zones were part of the range 48.2 and 58.1 years, respectively.
We predicted that undisturbed areas with deep snow, an absence of competitors, and close proximity to the boreal forest were more likely to be part of the southern range. We found that 2 of 5 of these relationships met our initial expectations (Table 1). Sampling units that were more frequently found within the Canada lynx range were closer to the boreal forest and had deeper average annual snow. We also predicted that years with large numbers of hare and lynx in the boreal forest, low number of competitors, and deep snow increased the extent of the southern Canada lynx range. We found that only 1 relationship met our initial expectations (Table 1); when the number of Canada lynx in the boreal forest increased, the area of southern range increased the following year.