The classic 10-year population cycle of snowshoe hares (Lepus americanus, Erxleben 1777) and Canada lynx (Lynx canadensis, Kerr 1792) in the boreal forests of North America has drawn much attention from both population and community ecologists worldwide; however, the ecological mechanisms driving the 10-year cyclic dynamic pattern are not fully revealed yet. In this study, by the use of historic fur harvest data, we constructed a series of generalized additive models to study the effects of density dependence, predation, and climate (both global climate indices of North Atlantic Oscillation index (NAO), Southern Oscillation index (SOI) and northern hemispheric temperature (NHT) and local weather data including temperature, rainfall, and snow). We identified several key pathways from global and local climate to lynx with various time lags: rainfall shows a negative, and snow shows a positive effect on lynx; NHT and NAO negatively affect lynx through their positive effect on rainfall and negative effect on snow; SOI positively affects lynx through its negative effect on rainfall. Direct or delayed density dependency effects, the prey effect of hare on lynx and a 2-year delayed negative effect of lynx on hare (defined as asymmetric predation) were found. The simulated population dynamics is well fitted to the observed long-term fluctuations of hare and lynx populations. Through simulation, we find density dependency and asymmetric predation, only producing damped oscillation, are necessary but not sufficient factors in causing the observed 10-year cycles; while extrinsic climate factors are important in producing and modifying the sustained cycles. Two recent population declines of lynx (1940–1955 and after 1980) were likely caused by ongoing climate warming indirectly. Our results provide an alternative explanation to the mechanism of the 10-year cycles, and there is a need for further investigation on links between disappearance of population cycles and global warming in hare–lynx system.