Hibernacula water chemistry and implications for hibernating bats

Abstract

Drinking water is a critical resource for hibernating bats and its importance may be further increased when disease affects water balance. White-nose syndrome (WNS), a fungal disease that causes dehydration and electrolyte imbalance in bats, is associated with high mortality rates of several hibernating bat species in North America. Aside from restoring water balance, hibernaculum water sources may also provide minerals to bats, which could contribute to restoring electrolyte balance and reducing the impacts of WNS. However, hibernacula water sources may also be a source of toxic elements, such as heavy metals. We collected water samples from 12 hibernacula in New Brunswick, Canada, and determined the concentrations of 18 elements in each water sample (n = 103 samples). Aluminum, barium, calcium, chloride, magnesium, manganese, potassium, and sodium were the most common elements detected, with concentrations of aluminum, lead, and manganese above drinking water recommendations (developed for human consumption) in some samples. The concentrations of electrolytes in cave water were orders of magnitude below therapeutic concentrations. Sampling period (early hibernation, late hibernation) did not affect results, but water profiles differed among sites and sample types within a site (running water, standing water, ceiling drip, and ice). The water profiles we recorded in our study suggest little potential for secondary consequences of drinking water, whether positive (i.e., electrolyte or mineral supplementation) or negative (i.e., heavy metal contamination).

cave water, contaminants, electrolytes, underground water, white-nose syndrome

Drinking water is a critical resource for hibernating bats. Hibernation (reduction of body temperature and metabolic rate) allows bats to manage their energy stores over long periods of time, but bats must also maintain water balance. Microclimate selection affects the rate of evaporative water loss, and both microclimate variations and dehydration have been suggested as drivers of spontaneous arousals during hibernation (Speakman and Racey 1989; Thomas and Cloutier 1992; Thomas and Geiser 1997; Ben-Hamo et al. 2013). Regardless of microclimate, hibernating bats must drink water periodically to maintain water balance since oxidative water released from fat mobilization is less than evaporative water loss in hibernating Myotis lucifugus (Thomas and Cloutier 1992; Thomas and Geiser 1997). For bat populations in the southern United States, it may be possible for bats to emerge from hibernacula during the winter to forage and drink due to higher temperatures and higher prey availability compared to northern populations (Whitaker and Rissler 1992). Bat populations in Canada and the northern United States often cannot leave hibernacula during the winter due to below-freezing temperatures outside, or due to the hibernaculum entrance being blocked by ice and snow. Additionally, liquid water and insect prey are difficult or impossible to find outside of caves for bat populations in the north during the winter. A variety of bat species have been observed drinking water from condensation on cave walls, as well as from standing water in hibernacula (Folk 1940; Rysgaard 1942; Twente 1955; Muir and Polder 1960; Codd et al. 1999). These water sources represent the sole intake, of any substance, for hibernating bats unable to emerge from their hibernaculum during winter. While water is a critical resource for hibernating bats, determining the constituents of hibernaculum water sources may be informative in the contexts of disease, contaminants, and applications of variation in water chemistry among sites.

Water balance during hibernation may be affected by the fungal disease, white-nose syndrome (WNS). WNS, a disease caused by the fungus Pseudogymnoascus destructans, is associated with significant mortality events involving 7 hibernating bat species in North America (Frick et al. 2010; Lorch et al. 2011; Turner et al. 2011; Langwig et al. 2012). Pseudogymnoascus destructans grows on the exposed skin of hibernating bats, which initiates a cascade of physiological effects (Warnecke et al. 2013; Verant et al. 2014). Among those effects, such as increased energy expenditure, WNS causes dehydration and electrolyte imbalance (Cryan et al. 2010, 2013; Warnecke et al. 2013; Verant et al. 2014), and therefore may cause bats to drink more during winter (Cryan et al. 2010; Willis et al. 2011). Although there is no evidence for increased drinking in inoculation studies with captive bats (Brownlee-Bouboulis and Reeder 2013; Wilcox et al. 2014; Bohn et al. 2016), free-living bats may have a wider variety of behavioral options to mitigate disease-related dehydration.

Hibernaculum water sources may provide minerals to bats (Codd et al. 1999), which could contribute to restoring electrolyte balance and reducing the impacts of WNS (Cryan et al. 2013). Indeed, the high level of calcium found in water sources from some underground sites has been suggested as a driver of the use of karstic caves by insectivorous bats with limited calcium availability from other sources, such as their insect prey (Barclay 1994; Adams et al. 2003; Racey and Furey 2014). Generally, previous studies on the chemical composition of water sources in caves have focused on the dissolution of carbonate bedrock and precipitation of speleothems, transformation of rain water chemistry when in contact with cave bedrock, and the influence of local pollution sources (Ford and Williams 1989; Mayer 1999; Fairchild et al. 2000; Palmer 2007). Water chemistry in caves is influenced by bedrock petrography and surrounding land use, and also varies seasonally due to changes in water temperature and flow rate related to external weather patterns, particularly rainfall (Mayer 1999; Motyka et al. 2005; Camacho et al. 2006; Kawai et al. 2006). Water chemistry can also vary within a site when comparing different water sources, including dripping, standing, and running water (Mayer 1999; Motyka et al. 2005).

While hibernaculum water sources might provide electrolytes for some hibernating bats, water may also contain toxic compounds. Many bats hibernate in abandoned mines where water sources potentially contain high concentrations of heavy metals. Bats appear to be particularly susceptible to heavy metal accumulation because most hibernating species are long-lived, occupy high trophic levels, feed on aquatic emergent insects, and sustain high metabolic rates and food intake during the active season (Hickey et al. 2001). While the bioaccumulation of heavy metals in bats from insect prey is recognized (Clark and Shore 2001), the additive effects of contaminated drinking water have not been studied. Heavy metal accumulation in bats is of increasing concern (Hickey et al. 2001; Flache et al. 2015; Zukal et al. 2015), and bats limited to poor quality water sources inside contaminated hibernacula may suffer adverse health effects.

We analyzed the water chemistry of samples collected from bat hibernacula to address 2 primary objectives: 1) to determine the levels of electrolytes in hibernaculum water as a resource for hibernating bats with particular reference to WNS, and 2) to document levels of heavy metals and other constituents related to potential health thresholds.

Materials and Methods

We collected water samples in 12 subterranean sites from 2014–2015: 9 caves and 3 abandoned mines in New Brunswick, Canada (Table 1). These represent all bat hibernacula in the province known and accessible at the time, and were used by little brown bats (M. lucifugus), northern long-eared bats (Myotis septentrionalis), and tricolored bats (Perimyotis subflavus—Vanderwolf et al. 2012). A map of the locations of sites can be found in Vanderwolf et al. (2012). The mean winter (1 November to 30 April) air temperature in the dark zone (area deep in caves beyond light penetration) of 11 of the sites is 5.0 ± 1.5°C (Vanderwolf et al. 2012). The entrances of some sites become completely blocked by snow and ice during the winter, such as White Cave, Berryton Cave, and Markhamville Mine. We followed the protocol of the United States Fish and Wildlife Service (2012) for minimizing the spread of WNS during all visits to caves. Water samples were collected early during the bat hibernation season (mid-November to late January), and late in the hibernation season (late March to late April). To test for variation among water sources, we collected samples in the dark zone from water dripping from the ceiling, standing water on the cave floor, flowing water on the cave floor, and from accumulations of ice in the twilight zone (Table 1). Although ice may not immediately be considered a drinking water source for hibernating bats, in other regions ice may represent the only water source (e.g., Manitoba; Q. M. R. Webber, University of Winnipeg, pers. comm.) and it is possible that bats could drink from melting ice. Samples collected from each site were dependent on which of these water sources were available. Two 125 ml high-density polyethylene, screw-cap bottles were filled for each water sample. Pieces of ice were collected in large plastic bottles and allowed to melt at 4°C in a refrigerator. The water was then poured into the screw-cap bottles described above, carefully avoiding disturbance of any sediment that had settled on the bottom of the collecting bottle. Water pH and temperature were measured in the field using a Hanna HI 98127 pH pen (Hanna Instruments, Woonsocket, Rhode Island) according to the recommendations of Sasowsky and Dalton (2005). The pH pen was calibrated at pH 7 and pH 10 the day of each site visit.

Table 1.

Characteristics of underground sites where water samples were obtained in New Brunswick, Canada, 2014–2015. All but 1 of our study sites were known bat hibernacula with hibernating populations of Myotis lucifugus, M. septentrionalis, and Perimyotis subflavus (Vanderwolf et al. 2012). The highest number of bats counted during 1 visit in the years 2009–2011 is shown. Information on bedrock geology is taken from McAlpine (1983). The number of water samples of each type that were taken is indicated. Liquid water samples were collected from the dark zone, whereas ice samples were collected from the twilight zone. ND = no data.

Site	Type	Length (m)	Max hibernating bat population (2009–2011)	Surrounding environment	Number of water samples
					Ceiling drip	Flowing	Ice	Standing	Total
Howes Cave	Precambrian limestone, Greenhead group	80	201	Greenspace in a city, near a road	0	1	1	4	6
Underground Lake Cave	Mississippian gypsum, Windsor group	140	243	Mixed forest, no human habitation nearby	3	2	2	3	10
Kitts Cave	Mississippian limestone, Windsor group	154	24	Mixed forest, near a road	2	4	2	3	11
Harbell’s Cave	Precambrian limestone, Greenhead group	74	32	Mixed forest, urban park	1	4	0	3	8
Berryton Cave	Mississippian limestone, Windsor group	332	6,087	Hardwood forest, hilltop limestone lens	3	4	2	3	12
Dalling’s Cave	Mississippian limestone, Windsor group	98	6	Mixed forest, no human habitation nearby	2	4	2	1	9
Markhamville Mine	Abandoned manganese mine	ND	245	Mixed forest, no human habitation nearby	3	3	2	4	12
Glebe Mine	Abandoned manganese mine	197	192	Mixed forest, no human habitation nearby	1	3	2	3	9
Dorchester Mine	Abandoned copper mine	ND	140	Mixed forest, no human habitation nearby	3	4	1	3	11
White Cave	Mississippian gypsum, Windsor group	515	219	Mixed forest, no human habitation nearby	3	3	2	3	11
Lost Brook Cave	Mississippian limestone, Windsor group	79	ND	Mixed forest, no human habitation nearby	0	1	1	0	2
Markhamville 2 Mine	Abandoned manganese mine	ND	27	Mixed forest, no human habitation nearby	1	0	1	0	2

Samples were kept cool and transported to a commercial laboratory (Analytical Services Laboratory, Analytical Services Section, New Brunswick Department of Environment, Fredericton, New Brunswick) within 1 week of collection. Samples with minimal visible sediment (< 2 mm sediment on bottom of collection bottle) were then preserved with ultra-pure concentrated nitric acid (0.25 ml acid/collection bottle; Seastar Chemicals, Sydney, British Columbia, Canada) and stored at room temperature. Samples containing > 2 mm sediment were centrifuged prior to acidification to prevent dissolution of solid matter. All samples were then poured off, centrifuged in 50 ml tubes (using HERMLE Labnet Z206 A), and acidified (0.1 ml acid/50 ml sample). Samples marked for chloride analysis were not acidified.

Metal parameters (aluminum, antimony, arsenic, barium, cadmium, chromium, copper, lead, manganese, selenium, thallium, uranium, zinc) and potassium were analyzed by inductively coupled plasma mass spectrometry (ICP-MS) following Environmental Protection Agency (EPA) method 200.8 (modified; see below) using a Perkin-Elmer Elan 9000 Inductively Coupled Mass Spectrometer (PerkinElmer Inc., Tempe, Arizona; Standard Methods 3125, approved 2009, editorial revisions 2011, Elan Instrument and Hardware Manuals). Modifications to EPA method 200.8 included: 1) samples were analyzed directly by nebulization, without acid digestion, after being preserved with nitric acid; 2) no laboratory-fortified blank was used, so quality controls were run approximately every 20 samples to monitor instrument performance; and 3) all calibration standards and blanks were prepared in a matrix of 0.2% nitric acid. A commercially prepared multi-element stock standard containing all elements of interest was purchased and diluted to prepare calibration standards.

Mineral parameters (calcium, magnesium, sodium) were analyzed by inductively coupled plasma optical emission spectrometry (ICP-OES) following EPA method 200.7 (modified; see below) using an Optima 3300 DV Spectrophotometer (PerkinElmer Inc.; Standard Methods 3120, approved 1999, editorial revisions 2011—APHA et al. 2005). Modifications to EPA method 200.7 included: 1) samples were analyzed directly by nebulization, without acid digestion, after being preserved with nitric acid; 2) spectral and inter-element interference corrections were not used; and 3) all calibration standards and blanks were prepared in a matrix of 0.2% nitric acid. Mixed calibration standards for Ca, Mg, and Na were prepared from commercially purchased 1,000 ppm stocks.

Chloride levels were analyzed by ion chromatography using a Dionex DX-500 ion chromatograph (Dionex Corp., Sunnyvale, California). Standard methods were followed (APHA et al. 2005; procedure 4110B approved 2000, editorial revisions 2011; https://www.standardmethods.org/), with the exception that the flow rate was 1.20 ml/min, the suppression was 50 mA, and the guard and analytical columns used were designated by the manufacturer for the analyses.

We compared observed electrolyte concentrations to therapeutic concentrations. A 1:1 dilution of unflavored Pedialyte (Abbott Nutrition, Abbott Laboratories, Columbus, Ohio) is used to rehabilitate captive bats (Klug and Baerwald 2010) and has been used for bats with WNS (L. P. McGuire and C. K. R. Willis, pers. obs.). Therefore, we assumed that to have therapeutic value, observed concentrations must approach: sodium 518 mg/l, chloride 620 mg/l, potassium 390 mg/l, and zinc 3.9 mg/l.

To assess potential health impacts of water quality on hibernating bats, we used the National Primary Drinking Water Regulation and National Secondary Drinking Water Regulation of the Environmental Protection Agency, United States (Environmental Protection Agency 2009), and the Guidelines for Canadian Drinking Water Quality (Health Canada 2014). In cases where recommended threshold values differed between agencies, we considered the lower threshold.

We tested for effects of sample type and sampling period with linear mixed effects models. First, we used a principal components analysis to reduce the data set. Only elements detected in > 70% of samples were included in this analysis. Missing values in the data set do not indicate an absence of that element, but rather a concentration below the limit of quantitation (LOQ). Therefore, missing values were replaced with one-half of the LOQ for that element. Six samples were mistakenly analyzed with a higher LOQ for sodium (10 mg/l rather than 0.1 mg/l), representing the only missing values for sodium. Replacing these missing values with 5 mg/l would not be appropriate, rather we used the median of all other sodium values (1.51 mg/l).

We used the prcomp method in R (v3.3.1—R Core Team 2016), and selected which components to retain based on the proportion of variance explained and visual examination of a scree plot. We followed the methods of Zuur et al. (2009) to build our model, using likelihood ratio tests to test for the effects of sample type, location, and period, including location as a random effect to account for repeated sampling. A paired t-test was used to compare changes in water temperature and pH from fall to winter.

Results

Calcium and chloride were the only parameters detected in all 103 water samples, while aluminum, barium, magnesium, manganese, potassium, and sodium were detected in > 70% of samples tested (Table 2). Antimony, arsenic, chromium, and thallium were not detected in any sample, while cadmium and selenium were only detected in 1 sample each. Aluminum, lead, and manganese were all detected at concentrations above drinking water recommendations in some samples (Table 2). The concentrations of electrolytes in cave water were orders of magnitude below therapeutic concentrations and no sample reached therapeutic concentrations (Fig. 1). The complete table of water analysis results is presented in Supplementary Data SD1.

Table 2.

Summary of water chemistry values, including the number of samples with values above recommended thresholds (see text), obtained from water samples taken from underground sites in New Brunswick, Canada, 2014–2015. Mean, median, and maximum values are presented, excluding samples in which the element was not detected. n = 103 samples tested; LOQ = limit of quantitation; n/a = not applicable.

Element	LOQ (mg/l)	Threshold (mg/l)	# above LOQa	# above threshold	Mean (± SE) (mg/l)	Median (mg/l)	Max (mg/l)
Aluminum	0.025	0.2	79	23	0.17 ± 0.024	0.0885	1.4
Antimony	0.001	0.006	0	0	n/a	n/a	n/a
Arsenic	0.0015	0.01	0	0	n/a	n/a	n/a
Barium	0.01	1	75	0	0.084 ± 0.016	0.0305	0.89
Cadmium	0.0005	0.005	2	0	0.0004 ± 0.0003	0.0004	0.0006
Calcium	0.1	n/a	103	n/a	105.97 ± 17.43	31.15	620
Chloride	0.05	250	103	0	5.62 ± 1.61	1.66	105
Chromium	0.01	0.05	0	0	n/a	n/a	n/a
Copper	0.01	1	17	0	0.057 ± 0.012	0.036	0.18
Lead	0.001	0.01	26	6	0.0062 ± 0.0015	0.0026	0.033
Magnesium	0.1	n/a	94	n/a	0.98 ± 0.13	0.54	7.37
Manganese	0.005	0.05	74	27	0.066 ± 0.014	0.0235	0.87
Potassium	0.1	n/a	94	0	0.63 ± 0.077	0.4	4.1
Selenium	0.0015	0.05	1	0	0.0015	0.0015	0.0015
Sodium	0.1a	200	97	0	2.89 ± 0.65	1.515	48
Thallium	0.001	0.002	0	0	n/a	n/a	n/a
Uranium	0.0005	0.02	9	0	0.00074 ± 0.000069	0.0007	0.0011
Zinc	0.005	5	48	0	0.017 ± 0.0013	0.014	0.039

^a6 samples had LOQ = 10 mg/l, so no sodium was detected in these samples.

Fig. 1.

Electrolyte concentrations in water samples from bat hibernacula in New Brunswick, Canada, compared to therapeutic concentrations. Horizontal dashed lines indicate the concentrations of each electrolyte in a 1:1 dilution of unflavored Pedialyte. Note the log scale of the y-axis. No observed hibernaculum water samples approached therapeutic concentrations.

After removing elements detected in < 70% of samples, the remaining data set for testing effects of sample type and sampling period included aluminum, barium, calcium, chloride, magnesium, manganese, potassium, and sodium (n = 103). We retained PC1 and PC2 for analysis, explaining 44% and 20% of the variance, respectively. All factor loadings in PC1 were negative, with loading for chloride, magnesium, potassium, and sodium all < −0.4. Most factor loadings in PC2 were close to zero (≤ |0.15|) with the exception of aluminum and manganese, which both had strong positive loadings (> 0.65).

For PC1, there was no effect of sampling period (likelihood ratio = 1.27, d.f. = 1, P = 0.26), but water profiles varied among locations (likelihood ratio = 38.52, d.f. = 9, P = 0.0001), and among sample types (likelihood ratio = 72.79, d.f. = 3, P < 0.0001). Ice was different from all other sample types (all pairwise P < 0.001), ceiling drip was different compared to flowing water (P = 0.009), but there was no difference between standing water and either ceiling drip or flowing water (Tukey post hoc tests). For PC2, as for PC1, there was no effect of sampling period (likelihood ratio = 0.30, d.f. = 1, P = 0.58), or location (likelihood ratio = 14.70, d.f. = 9, P = 0.20), but water profiles differed among sample types (likelihood ratio = 27.79, d.f. = 3, P < 0.0001). Water collected from ceiling drips differed from other water types, with lower aluminum and manganese concentrations in ceiling water samples (Tukey post hoc tests). The mean water temperature and pH were 5.1°C ± 1.8 and 8.0 ± 0.5, respectively, in early hibernation and 4.1°C ± 1.6 and 6.8 ± 0.6 in late hibernation (Supplementary Data SD2). These seasonal changes were significantly different both for water temperature (t₃₇ = 2.49, P = 0.017), and pH (t₃₇ = 9.70, P = 1.053e-11). Despite this decrease in temperature and pH over the hibernation season, water chemistry did not differ significantly between sampling periods.

Water chemistry varied among sites in predictable ways. Gypsum caves (White Cave and Underground Lake Cave) had higher concentrations of calcium than all other sites, and an abandoned copper mine (Dorchester Mine) had the highest copper levels (Supplementary Data SD1). Harbell’s Cave was notable among sites for high levels of potassium, sodium, magnesium, and chloride.

Discussion

The primary function of drinking water is to maintain osmotic balance, but we were interested in determining whether the chemical profile may have secondary consequences, either positive or negative. The chemical profile of hibernaculum water sources varied among sites, introducing the possibility of positive or negative secondary consequences varying among sites. The potential positive secondary consequence of drinking water that we considered was electrolyte replenishment in bats facing hypotonic dehydration due to WNS. Some sites were notable for much higher concentrations of sodium, potassium, and chloride than other sites (e.g., Harbell’s Cave, Howes Cave), likely due to infiltration of road salt from nearby roads. However, even including these high concentration sites, the concentrations of electrolytes we documented in hibernacula were much lower (orders of magnitude for most samples) than therapeutic concentrations, and therefore are unlikely to contribute to bat survival in the face of WNS infection. Independent of disease effects, it has been suggested that calcium could be a valuable resource for hibernating bats (Barclay 1994; Adams et al. 2003; Racey and Furey 2014), and some of our sites stood out with high calcium concentrations. Underground Lake Cave and White Cave are both gypsum caves and had much higher calcium concentrations than other sites. Yet these sites did not have the highest numbers of bats (Vanderwolf et al. 2012), suggesting that either calcium levels were sufficiently high in all sites, or bats do not choose underground sites based on calcium levels in the water. Thus, it seems unlikely that hibernaculum water sources serve as significant supplementary sources of electrolytes or minerals for hibernating bats based on our results. Water intake and calcium demands also are low for bats during hibernation.

We considered heavy metals and other contaminants as potential negative secondary consequences of drinking hibernaculum water sources. Among heavy metals, arsenic, lead, and mercury are considered the 3 most toxic substances, although other elements such as aluminum and selenium can be toxic at high concentrations (Clark and Shore 2001; Dolara 2014). We did not detect arsenic in our samples, but lead concentrations were above human health limits in several sites. Although aluminum and manganese also were above threshold values, these are not considered threats to human health but are instead listed due to operational and aesthetic purposes, respectively (Health Canada 2014). We did not measure mercury levels because elemental mercury has a low solubility, and water analysis is a poor indicator of mercury pollution (Williams and Coffee 1975). A variety of toxic elements have been documented from bat tissues, fur, and guano, such as lead, cadmium, mercury, selenium, chromium, zinc, arsenic, copper, manganese, and nickel (Clark and Shore 2001; Flache et al. 2015; Zukal et al. 2015). The concentrations of these elements vary with species, sex, age, locality, type of sample, and year of collection, with no clear patterns observed (Luftl et al. 2003; Allinson et al. 2006; Zukal et al. 2015). The lethal and sublethal effects of these elements on bats are largely unknown. There are no studies that quantify uptake and loss rates of toxic elements in bats, although symptoms of lead poisoning and detrimental effects of cadmium and zinc phosphide have been documented (Clark and Shore 2001). In general, both invertebrates and vertebrates excrete the majority or excess of consumed essential and nonessential elements through feces and urine, and indeed, heavy metal concentrations for insectivorous bats generally show higher levels of all elements in guano compared to tissues (Zukal et al. 2015). Local environmental conditions can affect both the bioavailability and toxicity of heavy metals (Mayer et al. 1994; Laskowski et al. 1995). High pH, very hard water, and low water temperatures, such as those found in our study, lowers toxicity and bioavailability for many metals (Mayer et al. 1994), but this protective effect may be negligible once water is ingested (bat stomach pH 5.1–5.6—Strobel et al. 2013). Therefore, although the chemistry of water sources in hibernacula may predictably vary among sites, the bioavailability of heavy metals and other elements, and the acute and chronic health effects to bats are uncertain.

Previous studies on the composition of water in caves have been conducted in more southerly latitudes with warmer water temperatures, but nevertheless the concentrations of heavy metals and other elements we documented show similarities. The calcium, aluminum, manganese, barium, and copper concentrations we recorded in cave waters were generally higher than those found in previous studies, while pH, magnesium, chloride, lead, zinc, and potassium concentrations were similar (Supplementary Data SD3 and SD4). Variation in water chemistry among underground sites is influenced by bedrock petrography and surrounding land use, which can help explain regional differences (Mayer 1999; Motyka et al. 2005; Camacho et al. 2006; Kawai et al. 2006). The sodium concentrations we detected were in the range of previous findings, although 2 previous studies found much higher values (Arizona and Mexico; Supplementary Data SD3). Therefore, hibernacula in North America generally have concentrations of electrolytes below therapeutic concentrations, which consequently are unlikely to contribute to bat survival. Values of pH and concentrations of calcium, magnesium, chloride, aluminum, copper, lead, cadmium, and zinc detected during this study were generally higher compared to freshwater sources outside of caves in eastern Canada (Scheuhammer et al. 1997; Clair et al. 2007). Therefore, underground sites may be important sources for these elements.

Although water chemistry may not affect health, predictable variation among sites suggests the interesting possibility of endogenous markers and trace element analysis for assigning bats captured during the summer active season to a hibernaculum from the previous winter. At high latitudes, water in hibernacula represents the only input for hibernating bats. If water profiles vary among sites, and these elements are incorporated into bat tissues, it may be possible to predict which site a bat originated from. In an exploratory analysis, we used linear discriminant analysis of liquid water samples (excluding ice) and found high classification accuracy, further suggesting water profiles varied predictably among sites (see Supplementary Data SD5 for further description and results of linear discriminant analysis). Further investigation of this idea is required to determine whether bats drink sufficient water, if there is sufficient incorporation of elements into tissues during hibernation, and how effective the technique can be without sampling all hibernacula available to bats (e.g., some hibernacula may be unknown or inaccessible). Many stages of validation are required before water chemistry may be used in studies of endogenous markers, but our results provide the critical first step, confirming that water chemistry (the only inputs for hibernating bats at northern latitudes) varies among sites.

The quantity of water bats consume during hibernation is unknown, as are the main hibernaculum water sources from which bats drink. Bats are known to drink water from condensation on cave walls and their own bodies during hibernation (licking and grooming), as well as from standing water in hibernacula (Folk 1940; Rysgaard 1942; Twente 1955; Muir and Polder 1960; Codd et al. 1999). Standing or flowing water will have had the opportunity to leach compounds from the substrate, but this potential is greatly reduced in condensation. Sampling condensation was not logistically possible in our study (our analyses required 250 ml of water), but it is possible that condensation could differ in concentrations of electrolytes and heavy metals. Regardless of the source which bats drink from, the water profiles we recorded in our study suggest little potential for secondary consequences of drinking water, whether positive (i.e., electrolyte or mineral supplementation) or negative (i.e., heavy metal contamination).

Supplementary Data

Supplementary data are available at Journal of Mammalogy online.

Supplementary Data SD1.—Raw data of chemical analysis of water samples collected in bat hibernacula.

Supplementary Data SD2.—Mean temperatures and pH of water sources in bat hibernacula in New Brunswick, Canada.

Supplementary Data SD3.—Commonly measured parameters of water bodies in caves from previous studies.

Supplementary Data SD4.—Trace metal concentrations in cave water from previous studies.

Supplementary Data SD5.—Linear discriminant analysis of water chemistry samples from New Brunswick, Canada.

Acknowledgments

Thanks to C. Ottens, L. Lamey, and the staff at the Analytical Services Laboratory in Fredericton, New Brunswick, for processing the water samples. Thanks to C. K. R. Willis and Q. M. R. Webber for helpful discussion of this project. Access to hibernacula on private lands was generously provided by D. Roberts, J. Chown, and T. Gilchrist. Sampling at Underground Lake Cave, a Class 1 New Brunswick Protected Natural Area, was made possible through a scientific research permit provided by the New Brunswick Department of Natural Resources Protected Natural Areas Program. Scientific permits for entering sites were provided by the New Brunswick Department of Natural Resources Species-at-Risk Program and the New Brunswick Protected Natural Areas Program. Research funding was provided by the Canadian Wildlife Federation, Crabtree Foundation, New Brunswick Environmental Trust Fund, New Brunswick Wildlife Trust Fund, New Brunswick Department of Natural Resources, and Parks Canada.