DISCUSSION
Here we examined the spatial variation in hibernation duration and estimated pre-hibernation fat stores across North America and applied these estimates to an updated model of hibernation energetics (Haase et al. 2019) to estimate the overwinter fat necessary for M. lucifugus to survive the hibernation. By comparing the required fat and the fat available, we predicted survival of M. lucifugus across its distribution for two different ecological situations: one in which bats roost within their most preferred conditions and one in which they roost in the best conditions predicted to be available. Finally, we modelled the possible impact of P. destructans infection to understand the implications for M. lucifugus populations where bats have yet to be impacted by WNS.
Winter hibernation duration is a key determinant of the overwintering survival for any hibernating species. While local variation in winter onset due to unique landscape features may create refugia where bats may persist later than or emerge from hibernation earlier than average, the lack of any previous broad-scale estimates for this critical variable highlights the need for this study. Prior work (Humphries et al. 2002, Hayman et al. 2016) had defined the hibernation period a priorias the yearly number of nights with a mean nightly temperature below freezing. Our results suggest that substantial improvements in the estimation of hibernation duration can be made by including elevation, latitude, and the number of days with frost. The number of days of frost, rather than the number of days below freezing, and the counter-intuitive negative coefficients for both Northing (-0.45) and DEM (-0.04) in the model suggest that there is more nuance to the relationship between bats and low temperatures than currently understood. Our model is potentially biased in part by the over-representation of Canadian data, low elevation sites, and the collinearity of elevation and latitude with other explanatory variables. Notably, however, when included in univariate models, the coefficient signs regress as expected with the coefficients increasing with latitude and elevation; yet the univariate models do not predict hibernation duration as well and had higher AIC scores. Because we were interested in prediction, we kept the best model by AIC; however, clearly more work is required to understand what predicts overwinter duration in bats. Also, there were few estimates of hibernation duration from the western-most US states and the more southern latitudes within the species’ distribution. These factors likely impacted model results in unpredictable ways and highlights the need for additional data collection.
Species such as M. lucifugus may not hibernate across their broadest summer distribution, but likely seek out locations with more favourable conditions to overwinter. The maximum winter hibernation duration predicted within the distribution of M. lucifugus was a month and a half longer than the longest observation within our dataset, and this did not consider the portion of the species’ distribution that extended into the Arctic Circle. In all probability, bats likely do not overwinter within these regions for multiple reasons. First, while our survival models predict that an uninfected bat could survive the longest predicted winter hibernation, the necessary roosting microclimate conditions may not exist on the landscape (Perry 2013, McClure et al. 2020). A series of complex interactions between surface features (e.g., slope, aspect, elevation, suitable crevices) or cave features (e.g., number of entrances, air flow, depth) ultimately define when and where suitable hibernacula conditions exist (Perry 2013, McClure et al. 2020). These site level determinants of microclimate conditions make it difficult to define a relationship between landscape-level features and the available subterranean conditions and create challenges in attempting to predict where suitable hibernacula conditions exist (McClure et al. 2020). Despite this, our use of modelled subterranean temperatures offers an improvement over assuming either static optimal conditions or mean annual surface temperature used in Hayman et al. 2016. Second, and perhaps more importantly, regions where extended winters reduce the summer active period to <100 days likely create significant challenges to reproductive success. With a gestation period of ~60 days (O’Farrrell and Studier 1973, Kurta et al. 1989), a female bat would be hard pressed to gestate, nurse, and wean young, while still allowing the young of the year time to fatten sufficiently to survive such an extended hibernation period. These results highlight the fact that only a subset of summer distributions may be suitable for overwinter survival, an idea rarely considered in the definition of bat species’ distributions.
Localized clines in body size and body mass of M. lucifugus have previously been recorded (Lausen et al. 2008, Lacki et al. 2015). When we compare more detailed metrics of body composition (lean mass and body fat content), however, these clines are better understood: the lean mass of bats generally stays consistent, while the difference in body mass is due to increases in fat. Our model selection suggests that variation in the duration of winter hibernation may in part drive variation in both body mass and fat stores across the range of the species. The relationships between mass (and thus fat) and latitude and days below freezing also suggest stronger selection pressure for heavier bats in more extreme conditions. While still useful, the relationship between the selected abiotic variables and body mass of bats showed strong spatial autocorrelations among residuals, and there may be additional continental-scale drivers or local determinants not investigated in this study.
The scaling relationship identified between pre-hibernation fat stores and body mass is a departure from the fixed 30% value used in previous energetic modelling studies (Humphries et al. 2002, Hayman et al. 2016) and is supported by contemporary findings (Cheng et al. 2019). More localized body mass–body fat relationships may exist, yet without increased data resolution drivers of the true relationship will remain difficult to assess. Here we assumed bats did not forage during hibernation, but some bats (primarily from southern hibernacula) have been known to forage over winter (Bernard and McCracken 2017). Without data, we were required to assume bats relied exclusively on pre-hibernation fat stores.
For our modelled roost conditions, virtually all uninfected M. lucifugus could survive the estimated duration of winter. Observed rates of survival among uninfected overwintering bats is high (Boyles and Brack 2009) and bats are likely capable of overwintering across most of their summer distribution where suitable hibernacula exist. These results are an improvement over previous models which did not predict survival where hibernation duration > 6 months (Hayman et al. 2016) despite our prediction that some 51% of the study extent may experience winters longer than that. In previous models, roosting microclimatic space was derived exclusively from surface metrics (i.e., mean annual surface temperature and relative humidity). Our use of a static, optimal roosting scenario served as a baseline for a best available temperature scenario and provided a more biologically relevant conditions for M. lucifugus, as bats are known to preferentially roost in these conditions when available (Thomas and Cloutier 1992, Haase et al. 2019). In areas where WNS has devastated hibernating colonies of M. lucifugus , available microclimates are often warmer with temperatures reaching 10°C (Perry 2013). This warmer microclimate has a significant impact on energy expenditure during hibernation (Figure S4). Despite this, our predictions for overwinter survival using estimates of the best roosting temperatures available suggest that survival may still be possible, if bats use the coldest areas within the cave or mine system. In a rare ray of hope, all of the hibernation temperatures that we recorded in the West were well below the 10°C mark, with hibernating M. lucifugus in northeastern Alberta and Northwest Territories roosting at about 2˚C and 100% relative humidity (CLL, unpublished data ). While the high relative humidity is generally beneficial to the growth of P. destructans (and thereby a promoter of WNS pathology), the cooler temperatures may slow fungal growth in comparison to the warmer roosts of the Eastern United States.
Our model predicted uninfected bats to emerge from hibernation with remaining fat quantities far greater than the amount of fat thought to be needed to hibernate for the entire duration of winter. Some of this may be an artifact of our modelling, as we may be missing additional energetic costs such as flight during the arousal periods or sex-related differences, and we did not model deviance from the most optimal arousal patterns that can result from arousals of individuals sharing the hibernacula (neighbour-initiated arousals; Hayman et al. 2010, Jonasson and Willis 2012, Turner et al. 2015, Czenze et al. 2017). Alternatively, remaining fat stores may be retained as a buffer against adverse abiotic conditions experienced after emergence, especially among females that undergo pregnancy immediately upon exit from hibernation (Johnson et al. 2017, Czenze et al. 2017) and whose reproductive success stands to benefit from a longer growing season. Early emergence would provide a selective advantage to the young of the year, as even one or two extra weeks foraging on the landscape could increase fat stores, making them more prepared for hibernation (Reynolds and Kunz in press). Currently, little is known about the energetic demands of the emergent bats as they return to the landscape, and accurate parameterization of this factor could significantly change our definition of survival capacity and increase our estimates of WNS-related mortality.
Modelled hibernation survival predictions of bats infected with P. destructans at either \(4\)˚C and 98% relative humidity or using the best available temperature do not match observations from the Eastern United States where mass mortality events have occurred due to WNS (Blehert et al. 2009, Frick et al. 2010, 2015) (23,24,54). While our analysis did demonstrate infection with P. destructansdramatically increases the amount of energy expended during hibernation, nearly 95% of all cells analysed predicted survival despite infection. Interestingly, however, the areas with the greatest increase in hibernation energy expenditure using the best available temperature were much more in line with regions where bat populations have experienced the greatest mortality.
Irrespective of the absolute values for survival predicted within this work, our modelling predicts that M. lucifugus populations where WNS is currently absent, especially those along the Rocky Mountains, Alberta, British Columbia, and Alaska will require similar increases in energetic expenditure as populations currently impacted by WNS. Thereby, if the increase in energy expenditure results in the same pattern of mortality, M. lucifugus populations in western North America may be expected to suffer mortality events similar to those experienced in the eastern and central populations (Frick et al. 2010, 2015). The long duration winters, especially in northern British Columbia, northern Alberta, Alaska, Yukon, Northwest Territories, and portions of the Rocky Mountains, may result in severe WNS-associated pathology and disruption to hibernation physiology, although use of cool available hibernacula microclimates in some of these areas may slow fungal growth (Reeder et al. 2012, Verant et al. 2014). Identifying and describing roosting microclimates in these areas will allow for better predictive models at a regional scale. Some northern hibernacula may provide a refugia forM. lucifugus infected with WNS; however, this cannot be defined until the availability of microclimates and spring energy requirements are quantified. For example, if gleaning of arthropod prey in spring is typically required for successful reproduction in northern latitudeM. lucifugus (Talerico 2008, Kaupas and Barclay 2017), then the extra energy expenditure associated with gleaning (Norberg et al. 1987) may require additional fat stores and WNS-related mortality rate may in fact be higher than predicted. Additionally, even if WNS related mortality is not directly observed, reproductive success may decline resulting in more long-term population decline.
The outputs of the energetic model that we applied are sensitive to a number of parameters and assumptions (Haase et al. 2019). The hibernation energetic model is largely derived from first principles and deviations in parameters, especially those defining the frequency or duration of arousal (cluster-based arousal, partial arousals, disturbances, etc. 8) or increasing these costs (e.g. increased number of mid-winter flights or disease pathophysiology), have the potential to alter the amount of resources required to survive hibernation. Roost microclimates are important to survival predictions and impact bat energetic and fungal growth dynamics (Marroquin et al. 2017), and further study is required to understand how relative humidity may vary across the landscape and within hibernacula.
Overall, this work represents an effort to iteratively refine both individual and landscape models of bat hibernation physiology and the impacts of WNS on bat populations. By specifically addressing the spatial heterogeneity of both abiotic phenomena and host traits, we offer some of the most detailed predictions for the potential impacts of WNS on M. lucifugus as P. destructans continues to spread through western North America.