Hibernacula microclimate and declines in overwintering bats during an outbreak of white‐nose syndrome near the northern range limit of infection in North America

Abstract We document white‐nose syndrome (WNS), a lethal disease of bats caused by the fungus Pseudogymnoascus destructans (Pd), and hibernacula microclimate in New Brunswick, Canada. Our study area represents a more northern region than is common for hibernacula microclimate investigations, providing insight as to how WNS may impact bats at higher latitudes. To determine the impact of the March 2011 arrival of Pd in New Brunswick and the role of hibernacula microclimate on overwintering bat mortality, we surveyed bat numbers at hibernacula twice a year from 2009 to 2015. We also collected data from iButton temperature loggers deployed at all sites and data from HOBO temperature and humidity loggers at three sites. Bat species found in New Brunswick hibernacula include Myotis lucifugus (Little Brown Bat) and M. septentrionalis (Northern Long‐eared Bat), with small numbers of Perimyotis subflavus (Tricolored Bat). All known hibernacula in the province were Pd‐positive with WNS‐positive bats by winter 2013. A 99% decrease in the overwintering bat population in New Brunswick was observed between 2011 and 2015. We did not observe P. subflavus during surveys 2013–2015 and the species appears to be extirpated from these sites. Bats did not appear to choose hibernacula based on winter temperatures, but dark zone (zone where no light penetrates) winter temperatures did not differ among our study sites. Winter dark zone temperatures were warmer and less variable than entrance or above ground temperatures. We observed visible Pd growth on hibernating bats in New Brunswick during early winter surveys (November), even though hibernacula temperatures were colder than optimum for in vitro Pd growth. This suggests that cold hibernacula temperatures encountered near the apparent northern range limit for Pd do not sufficiently slow fungal growth to prevent the onset of WNS and associated bat mortality over the winter.


| INTRODUC TI ON
The emergence and spread of multiple wildlife fungal diseases during recent decades has had devastating consequences for biodiversity (Fisher et al., 2016). When a disease enters a naive host population the initial wave of infection often causes epizootics resulting in mass mortality, which may extirpate local host populations or even cause species extinction (Fisher et al., 2016). Factors not directly associated with interactions between host and pathogen, such as environmental conditions, can considerably influence the manifestation of a disease by acting on either the pathogen or the host (Fisher et al., 2016;Scholthof, 2007).
White-nose syndrome (WNS), a disease caused by the fungus Pseudogymnoascus destructans (Pd), is responsible for the deaths of >6.7 million bats in North America (Lorch et al., 2011;US Fish and Wildlife Service, 2019). First reported in 2006 in New York, WNS has spread throughout eastern North America and has also been found in parts of the western USA (Lorch et al., 2016). For unknown reasons, WNS-associated mortality varies greatly among bat species (Frank et al., 2014;Turner et al., 2011). The WNS-associated mortality levels of bats in individual hibernacula may be influenced by multiple variables, including distance from the nearest known Pd-positive site, hibernaculum microclimate, initial bat colony size, bat species composition, and latitude (Flory et al., 2012;Langwig et al., 2012;Wilder et al., 2011). However, data reporting WNS-associated bat declines in North America in relation to the above variables remains scarce in the literature, and no such data exists for Canada (Powers et al., 2015;Turner et al., 2011).
In the USA, Myotis lucifugus (Little Brown Bat), M. septentrionalis (Northern Long-eared Bat), and Perimyotis subflavus (Tricolored Bat) have suffered up to 100% mortality in multiple caves, with an average of 91%, 98%, and 76% mortality, respectively, in those northeastern hibernacula surveyed (Turner et al., 2011). The arrival of WNS has placed Canadian populations of these species at risk, and all three are now listed as endangered under the Canadian Species-at-Risk Act (Environment Canada, 2015). The impact of WNS on Canadian populations of P. subflavus may be particularly severe as southeastern Canada represents the northern extent of P. subflavus range and this species was considered rare to uncommon in Canada pre-WNS (Forbes, 2012;Hitchcock, 1965;Mainguy & Derosiers, 2011).
Microclimate is important in hibernaculum selection by bats, among other factors , and may also be an important factor influencing WNS-associated bat mortality rates. Bats in North America often hibernate at sites with high relative humidity and temperatures that range from −10 to 21°C, depending on the bat species and latitude, although the typical range for most cave-hibernating species is 3.0-10.0°C (Perry, 2013). Cryan et al. (2013) suggested that temperature and relative humidity differences in hibernacula environments could have major effects on which bat species are affected by WNS and infection severity. Indeed, WNS-associated mortality rates of M. lucifugus have been positively associated with hibernacula temperatures in North America (Hayman et al., 2016;Johnson et al., 2014;Langwig et al., 2012Langwig et al., , 2016. However, in Europe, Bandouchova et al. (2018) found that M. myotis had the highest infection intensity when hibernating at low temperatures. WNS may change the microclimate preferences of bats. Following the arrival of WNS in Pennsylvania, the majority of M. lucifugus, P. subflavus, and Eptesicus fuscus roosted in colder sections of hibernacula compared to pre-WNS aggregations (Johnson et al., 2016). This behavior may increase energy conservation or affect disease progression (Johnson et al., 2016).
Cave and mine microclimates are influenced by multiple factors, such as local climate, internal hydrology, and internal topology (e.g., length and complexity of passages, number of entrances; Perry, 2013). Generally, cave and mine temperatures reflect mean annual surface temperature of the surrounding area (McClure et al., 2020;Moore & Sullivan, 1997), with the influence of outside weather conditions extending for several hundred meters into sites, depending on topology (Cropley, 1965). Large bodies of water such as subterranean lakes may greatly affect temperatures because of the high specific heat of water (Perry, 2013). Water entering underground sites also affects temperatures; for example warm water entering sites may radiate warmth, even at considerable distances from the entrance (Moore & Sullivan, 1997;Perry, 2013). Most caves and mines have a gradient of increasing relative humidity from the entrance, rising to 100% in the distant reaches of sites (Perry, 2013;Wigley, 1969).
In 2009, anticipating the spread of WNS to Maritime Canada (New Brunswick, Nova Scotia, Prince Edward Island), we initiated surveys of overwintering bats and collected data on species composition, abundance, and microclimate at hibernacula in New Brunswick, Canada (Vanderwolf et al., 2012). Bat populations in New Brunswick hibernacula include M. lucifugus and M. septentrionalis with small numbers of P. subflavus (Vanderwolf et al., 2012).
Eptesicus fuscus has never been observed in New Brunswick caves or mines, although small numbers overwinter in buildings in the province (McAlpine et al., 2002). Our study area is farther north than most previous hibernacula microclimate studies in North America, and temperatures may be lower, which could affect WNS-associated mortality rates of bats. Here we present data on the spread of Pd among known bat hibernacula in the province following the 2011 arrival of WNS in Maritime Canada, WNS-associated mortality at individual hibernacula, and the subsequent pattern of decline in overwintering bats in relation to hibernacula microclimate. We also compare hibernacula air temperatures, as measured by different microclimate loggers in the same sites, to assess which physical factors affect hibernacula temperatures.

| Bat counts
We monitored six limestone caves, two gypsum caves, and three Howes 182 ± 0 Jan 171 ± 4.2 Jan 200.7 ± 6.8 Feb an abandoned copper mine 2011-2015 that was not included in our initial pre-WNS study (Vanderwolf et al., 2012). These sites (Table 1) included all known bat hibernacula in the province at the time, although additional sites were discovered in 2015. We followed the protocol of the US Fish and Wildlife Service (2012) for minimizing the spread of WNS during all visits to caves, and obtained necessary permits from the New Brunswick Department of Natural Resources and Energy Development.
We counted hibernating bats twice annually (November-December and March-April) as in Vanderwolf et al. (2012), ~6-8 weeks after bats had entered hibernacula for the winter and ~3-6 weeks before estimated exit from each site for the summer.
We did this to determine how many bats were present at the start of hibernation and how many survived to the end of hibernation in each site. Typically, each count was undertaken by two individuals. Low visitation rates (e.g., 1-3 visits per winter) are reported to have little detrimental effect as hibernating bats tolerate several forms of natural disturbance, such as sound and predator activity (Boyles, 2017;Kilpatrick et al., 2020). When present, we collected and identified bat carcasses to species and sex. To reduce disturbance, live Myotis were not identified to species or sex. However, we identified Myotis to species and determined sex for ten live bats per hibernaculum in 2010 (n = 8 hibernacula) and species only in 2012 (n = 5 hibernacula) when handling bats was required for other research (Vanderwolf et al., 2013. Hibernating P. subflavus are morphologically distinct and could be identified without removing bats from cave walls, therefore this species was identified throughout the study. To minimize disturbance to hibernating bats we assessed live bats in the field for the presence of characteristic Pd fungal growth by visual inspection of exposed skin surfaces while bats where roosting (i.e., we did not remove bats from cave walls for inspection). This method underestimates the number of bats with WNS, as we could only examine the exposed skin of accessible roosting bats (some bats were roosting too high to be assessed). As well, lack of visible Pd growth does not equate to the absence of WNS (Janicki et al., 2015). One bat per hibernaculum was removed for confirmation of WNS by histology and sequencing at the Canadian Cooperative Wildlife Health Center on Prince Edward Island after our first observation of visible Pd growth at a site.

| Statistical analysis
We converted relative humidity to water vapor pressure because relative humidity is less informative and potentially misleading (Kurta, 2014). We calculated equilibrium vapor pressure using temperature data and the quadratic formula of Tabata (1973), and then determined actual vapor pressure (hPa) by multiplying the saturation vapor pressure by the relative humidity as recorded by HOBOs. We incorporated previous iButton temperature data collected from the same sites October 2009 to October 2010 (Vanderwolf et al., 2012) in the analysis. We conducted all analyses in R (R Core Team, 2020), and generated plots using ggplot2 (Wickham, 2016 with a Wilcox test and a paired Wilcox test, respectively. We compared air temperature (as measured by iButtons) to water temperature data collected during a previous study (Vanderwolf et al., 2017) with a Wilcox test.
To determine which variables affected the number of bats in each cave, we ran a linear model with hibernaculum passage length, presence of water, year, mean dark zone temperature in the winter (November-April), winter standard deviation of dark zone temperature, maximum winter dark zone temperature, and minimum winter dark zone temperature as explanatory variables. The April bat count was used as the number of bats at each site for each winter, and the temperature variables were calculated from iButton measurements.
We used the function AICtab (package bbmle) (Bolker & Team, 2017) to compare model Akaike information criteria (AIC) values. We considered a p < .05 significant.  In all subsequent counts, we distinguished live bats from dead bats (Tables 1 and 2

| Cave microclimate
The number of microclimate readings taken at each site varied depending on how many loggers malfunctioned (Table 5). Overall, we collected 138,061 iButton air temperature readings (counting paired iButtons in the same box as a single measurement) and 21,842 HOBO air temperature readings.

TA B L E 4
The number of live Perimyotis subflavus seen in New Brunswick hibernacula with the month the count was conducted and the percentage of bats with visible fungal growth on exposed skin  (Figure 2). The R 2 for fixed effects was 0.12 compared to 0.67 for the full model, indicating that year and month explain more twilight zone temperature variation than hibernaculum length and the presence of water.
Dark zone air temperatures increased as hibernaculum passage length increased (F 1 = 7,092.3, p < .001) and were also affected by the presence of water (F 2 = 6,808.0, p < .001). Dark zone temperatures were warmer for sites that contained flowing water (estimate = 6.24°C, p < .001) compared to sites with no water (estimate = 4.85°C, p < .001) or standing water only (estimate = 3.65°C, p < .001). The R 2 for fixed effects was 0.12 compared to 0.56 for the full model, indicating that year and month explain more temperature variation than passage length and the presence of water. A Wilcox test indicated that air temperature was not different from water temperature in the dark zone (W = 4,399, p = .203).
Actual water vapor pressure varied with month (F 11 = 2,672.2, p < .001), location within sites (F 2 = 2,239.5, p < .001), year (F 3 = 669.8, p < .001), and site (F 2 = 336.7, p < .001; Figure 3). There was an interaction between month and location (F 22 = 1,144.7, p < .001), site and location (F 22 = 53.8, p < .001), and year and location (F 6 = 2.3, p = .033). Actual water vapor pressure did not vary among months, sites, or years in the dark zone, while the pressure was higher in summer versus winter outside hibernacula and higher in early fall versus winter in twilight zones (Figure 3). Overall, the dark zone had a higher actual water vapor pressure (estimate = 8.01 hPa, p < .001) than either the twilight zone (estimate = 4.98 hPa, p < .001) or above ground (estimate = 3.52 hPa, p < .001). The dark zone and twilight zone were rarely undersaturated, unlike above ground (Figure 3).
Actual water vapor pressure as measured by the Kestral weather meter was often lower than measurements recorded by HOBOs taken the same day (Figure 4). Likely this was because Kestral measurements were taken on the floor of hibernacula while HOBO measurements were taken near the ceiling. Kestral data indicate the air in the dark zone was close to saturation in all sites during winter.
Air temperature measurements taken by iButtons and HOBOs at the same locations within the same hibernacula at the same time were significantly different (W = 262,348,267, p < .001), with a 0.13°C ± 0.07 difference in the dark zone, 0.38°C ± 0.14 in the twilight zone, and 1.39°C ± 0.49 above ground. HOBO measurements were higher than iButton measurements at above ground locations (Table 7)  USA, particularly M. septentrionalis (Frick et al., 2017). Bat species that show resistance or tolerance to WNS, such as E. fuscus (Frank et al., 2014;Frick et al., 2017;Turner et al., 2011) (Puechmaille et al., 2011). In the USA, Lorch et al. (2011) detected WNS lesions in late September, but major mortality was not observed until the end of January, with peak mortality in March.
The pronounced drop in the overwintering bat population in New Brunswick mirrors bat mortality rates observed in the eastern USA (Powers et al., 2015;  respectively (Vanderwolf et al., 2016b). Therefore, the apparent lack of delay in some sites may be due to our inability to visually observe Unlike Myotis spp., we never observed P. subflavus roosting near hibernacula entrances with the onset of WNS. Our bat counts generally decreased from early hibernation to late hibernation, as found in the USA . This may indicate mortality followed by predation (McAlpine et al., 2011) or the movement of bats to unknown hibernacula, although midwinter flights by Myotis spp. are considered rare (Davis & Hitchcock, 1965).
Bats in New Brunswick appear to have survived multiple years after the arrival of WNS, despite an overall major population decrease due to WNS and the continued presence of viable Pd on cave walls at hibernation sites (Vanderwolf et al., 2016a). However, because we did not mark bats individually, we cannot reject the possibility that some of these "survivors" are migrants from unaffected areas, but this seems unlikely because all adjacent provinces and  (Lorch et al., 2013;Puechmaille et al., 2011;Zhang et al., 2014), bats can likely acquire Pd from roosting substrates, as well as other bats . In the eastern USA, some bat species have high Pd-loads, even in caves with <5 hibernating individuals roosting singly (Frick et al., 2017).

This suggests that roosting individually offers no protection from
Pd transmission once Pd is established in a hibernaculum due to environmental reservoirs and cryptic connections among bats (Hoyt et al., 2018).
Bats in our study area did not appear to select hibernacula based on temperature, and bats in all sites experienced high WNSassociated mortality. However dark zone temperatures and water vapor pressure did not differ among our sites, likely because they were all located within a relatively small geographic area. However, bats may be selecting microclimates within sites that differ from our logger measurements. Bat species found in our sites can successfully hibernate in a wide range of temperatures (−4 to 17.8°C), although P. subflavus generally favor warmer temperatures than M. lucifugus or M. septentrionalis (Meierhofer et al., 2019;Webb et al., 1996). Wilder et al., (2011) suggest that bats in mines are less affected by WNS than in caves, possibly due to differing microclimates. Dark zone temperatures did not significantly differ between caves and mines in New Brunswick, nor was there a difference in bat mortality rates. Langwig et al. (2012) found that M. lucifugus population declines were higher in hibernacula with higher temperatures, which Johnson et al. (2014) and Grieneisen et al. (2015) also found in laboratory experiments. Conversely, Flory et al. (2012) found that WNSrelated mortality was higher in hibernacula with lower temperatures.
As found by previous studies, the presence of flowing water and passage length influenced hibernacula temperatures at our study sites (Perry, 2013), especially at the entrance. Hibernacula with longer passages are buffered from outside temperature fluctuations.
Also, relatively warm water from the dark zone flowing toward the entrance may warm air temperatures in the twilight zone during winter. Significant differences in temperature data between paired iButtons and iButtons and HOBOs in the same locations are likely due to the large sample sizes. The dark zone and twilight zone differences between iButton and HOBO measurements and between paired iButtons were within the error range of iButtons (±1°C) and HOBOs (±0.45°C). Although iButtons and HOBOs were attached to the same tree outside hibernacula, HOBOs were always placed above iButtons, which may partially explain why HOBOs measured higher temperatures than iButtons above ground. Additionally, above ground the HOBO sensor was open to the air while iButtons were placed in a box, which may have influenced temperature readings (Terando et al., 2017).
Caves in the southeastern USA have experienced lower WNSrelated mortality compared to the northeast, possibly due to shorter hibernation periods and increased winter insect availability (Bernard & McCracken, 2017;Flory et al., 2012). Winter flights by bats (exiting hibernacula) are common in the southeastern USA (Bernard & McCracken, 2017). The increased fungal growth associated with warmer temperatures may influence disease outcomes less than the length of the hibernation period and food availability. In Canada, the duration of cold weather, and the subsequent length of hibernation, is even greater than in the northeastern USA (Norquay & Willis, 2014). Although optimal growth temperatures for Pd are 12.5-15.8°C in vitro (Verant et al., 2012), optimal conditions may differ for in vivo growth.