Discussion
We used camera trap by-catch observations to provide the first quantitative assessment of how large-scale environmental variables correlate with mountain hare moult timing. As predicted, hares at higher altitudes and latitudes moulted later in spring and earlier in autumn, keeping their winter white coats for longer, compared to their lowland and low altitude conspecifics. Contrary to our prediction, we found support for a later spring moult and earlier autumn moult in areas characterised by coastal climates, rather than inland climates, resulting in hares that live in areas with coastal climates keeping their winter white coats for longer than their inland conspecifics.
The unexpected correlation between climate zone and moult timing (Figure 2.c.) could result from variation explained by climate zone also being partially explained by altitude (Pearson correlation coefficient > 0.5 in both seasons) and latitude (Pearson correlation coefficient < -0.3 in both seasons) (Suppl 6). Additionally, all camera traps from 63° North are close to the coast which may confound results. However, in southern Norway there is a coastal to continental moult timing gradient shown in the prediction maps (Figure 3.a. and 3.f.). Furthermore, it is likely that there will be increased variability in moult timing in coastal climates as there is increased among year variation in snow cover and duration compared to continental climates (senorge.no).
Both altitude (Figure 2.a) and latitude (Figure 2.b) had the predicted correlation with moult timing. Mountain hares that inhabit areas characterised by higher altitude and latitude moulted later in the spring and earlier in the autumn, keeping their winter white coats longer, likely due to an expectation of longer snow cover duration. These results are consistent with previous studies of other mountain hare populations and other hare species, which found that increased elevation correlated with mountain hares keeping their winter coats for longer (Watson 1963) and increased latitude correlated with snowshoe hares keeping their winter white coats for longer (Grange 1932). However, a more recent study on snowshoe hares found no evidence of moult timing variation across a latitudinal gradient in autumn and, in spring, hares that live at high latitudes became brown earlier than those at lower latitudes (Zimova et al. 2019). The correlations between moult timing and altitude and latitude indicates that mountain hare populations have adapted to local conditions, which suggests that gene flow between populations is insufficient to dilute local adaptations. This is particularly true for altitude as this variable can change significantly over a short geographic distance.
The non-overlapping CIs between some years in the spring moult (Figure 4) indicate that there is some between year phenotypic variation. However, the between year difference in moult timing is small, which is consistent with photoperiod rather than climate being the main driver of mountain hare moult timing (reviewed in Zimova et al. 2018). This is consistent with similar studies conducted on snowshoe hares (Mills et al. 2013, Zimova et al. 2014) and least weasels (Atmeh et al. 2018), which found evidence of between year phenotypic variation in spring, but not in autumn. The limited moult timing variation may reduce fitness as an inability to change moulting patterns in response to among year variation in snow extent and duration will increase camouflage mismatch and, consequently, decrease survival probability (Zimova et al. 2016). Behavioural responses to mismatch are possible, micro-habitat patch selection, changed diurnal activity patterns, or manually removing winter fur during the spring moult. However, there is currently no evidence of snowshoe hares modifying their behaviour in response to being mismatched (Zimova et al. 2014).
Moult timing variation could result from population level phenotypic variation, individual level phenotypic plasticity, or a combination of both factors. As our methodology did not facilitate monitoring specific individuals, we could not disentangle the relative importance of population and individual level variation. Observations were obtained at specific locations in multiple years (Suppl 8). This makes it likely that some individuals were recorded in multiple years, increasing the probability of individual phenotypic plasticity influencing results.
As the camera traps were deployed to maximise the probability of detecting lynx their locations within the surveyed area were not randomised and some areas of Norway were not monitored. Additionally, we obtained three times as many observations in spring compared to autumn (Suppl 4), which is probably caused by increased hare activity patterns during the mating season (Pettigrew et al. 2021) and the spring dataset containing observations for 58 more days. This resulted in the intercept CIs (Figure 4) being larger in autumn than in spring. Despite these issues, the prediction maps (Figure 3 and Suppl 10) indicate that the model produces accurate results.
Understanding the role that environmental characteristics have on moult timing is vital when assessing the impact that climate change may have on species that express a seasonal coat colour change. Our study sites span 1,300 km and 11 latitudinal degrees, from sea level to 841 m.a.s.l. making this is the first study that investigates moult phenology over a large, continuous climatic gradient spanning three biomes (temperate forest, boreal forest, and alpine tundra). Analysing the correlation between the explanatory variables and mountain hare moult timing will enable us to predict how the species will react to climate change induced reductions in snow cover extent and duration. We plan to investigate this issue in a future paper.