Supplemental Material

Table S1. Summaries of models used to explain the probability of harvesting a lynx while accounting for harvest effort across sampling units in the southern lynx range between 1948-2017 in Ontario, Canada. Covariates unrelated to animals harvested were also included in each of these models. These were the total number of trapping units, the area occupied by those trapping units and the average pelt price. The spatiotemporal pattern was modelled with a tensor product smoother. We modelled the year of harvest with a cubic regression smoother and the spatial process was modelled with a thin plate smoother on the geographic coordinates or with a Markov random field smoother using the neighborhood structure. Harvest-based effort covariates were ln transformed. MRF is Markov Random Field and TP is Thin Plate smoother.