Climate change affects timings, frequency, and intensity of frost events in northern ecosystems. However, our understanding of the impacts that frost will have on growth and survival of plants is still limited. When projecting the occurrence of frost, the internal variability and the different underlying physical formulations are two major sources of uncertainty of climate models. We use 50 climate simulations produced by a single-initial large climate ensemble and five climate simulations produced by different pairs of global and regional climate models based on the concentration pathway (RCP 8.5) over a latitudinal transect covering the temperate and boreal ecosystems of western Quebec, Canada, during 1955–2099 to provide a first-order estimate of the relative importance of these two sources of uncertainty on the occurrence of frost, i.e. when air temperature is < 0 °C, and their potential damage to trees. The variation in the date of the last spring frost was larger by 21 days (from 46 to 25 days) for the 50 climate simulations compared to the 5 different pairs of climate models. When considering these two sources of uncertainty in an eco-physiological model simulating the timings of budbreak for trees of northern environment, results show that 20% of climate simulations expect that trees will be exposed to frost even in 2090. Thus, frost damage to trees remains likely under global warming.

Climate warming-driven early leaf-out is expected to increase forest productivity but concurrently increases leaf exposure to spring frosts, which could reduce forests' net productivity. We hypothesized that due to their damaging effect on buds, spring frosts exert a stronger control on bud phenology than do growing degree-days. We monitored bud flush phenology of three white spruce seed sources (one local seed source from the boreal mixedwood forest and two seed sources from the temperate forest), one black spruce seed source originating from the boreal mixedwood forest and four nonlocal Norway spruce seed sources in 2016 and 2017 in two plantations located on both sides of the temperate-boreal mixedwood forest ecotone in eastern Canada (Quebec). We aimed to determine inter- and intraspecies variations in bud break timing and sensitivity to air temperature and photoperiod. We expected that bud break timing for boreal species and seed sources would be better synchronized with the decrease in frost probability than for nonlocal species and seed sources. We used mixed binomial regressions and AICc model selection to determine the best environmental variables predicting each transition from one stage of bud phenology to the next. At both plantation sites, white spruce bud flush began and ended earlier compared to black and Norway spruce. Buds of all spruce species were sensitive to frost probability for early phenological stages, whereas growing degree-days controlled the remaining stages. Photoperiod sensitivity was higher for white spruce compared to black and Norway spruce and reached its maximum in the temperate forest. At intraspecies level, the two southern white spruce seed sources opened their buds earlier than the local source and were more sensitive to photoperiod, which increased their exposure to spring frosts. Onset of spruce bud flush is driven by spring frosts and photoperiod, but once started, bud phenology responds to temperature. The high photoperiod sensitivity in white spruces could counterbalance climate warming and limit future premature leaf-out, whereas the low photoperiod sensitivity in black spruce should not restrain leaf-out advancement with climate warming. Our results call for adapting the temperature-driven hypotheses of ecophysiological models predicting leaf-out to include spring frost probability.