Study site – This study was conducted at a cave located in the Snake Range, White Pine County, Nevada (2,040 m elevation). The cave is approximately 10 m up a steep inclined rock face and is accessible through a 4 by 13 m opening (Figure 1). The cave opening leads to a large upper first chamber, which descends roughly 20 m below the opening, narrows, and continues to a second large chamber. The internal structure of the cave allows the second chamber to retain heat produced by roosting bats, buffering roosting bats from surface conditions, while the first chamber provides an area for bats to gain altitude before outflight by flying counterclockwise around the room (Figure 1).

The cave was mined for guano during the 1920’s. In 1926, a 66 m long adit was driven from the hillside into the lower chamber (Figure 1). This adit facilitated guano removal by providing direct access to the greatest accumulations of guano at the base of the second chamber. Placement of the adit portal at a lower elevation than the natural cave opening facilitated airflow between the two openings, likely causing a drop in cave temperature and humidity. In 1996, the adit opening was sealed so that the internal climate of the cave could return to historical conditions.

Historical bat use of this roost is largely anecdotal. Bat use is suspected to have increased in response to the adit closure, due to an increase in cave temperature and humidity. Occasional historical surveys suggested that use was limited to late summer and early fall by migrating bachelor males. Beginning in 2010, intensive surveys revealed large numbers of bats occupying the cave from June to October.

Bat capture and assessment – Bats were captured using a 2 by 2 m double-frame harp trap (Austbat, Victoria, Australia http://www.faunatech.com.au/products/harptrap.html, Kunz, Hodgkison, & Weise, 2009). Foam padding was attached to the frame on the outflight side of the harp trap to reduce risk of injury to bats that hit the frame. The harp trap was deployed at the mouth of the cave, perpendicular to the direction of outflight in one of the main flight paths of the emerging column of bats (Figure 1). The trap was opened at dusk before outflight and monitored continuously until it was closed at the end of primary nightly exodus of bats from the cave, or sooner if more bats were caught than could be processed in a timely manner. Although bats continued to fly in and out of the cave all night, trapping focused on the main outflight and typically ended by midnight. After being captured, bats remained in the harp trap bag until removal for processing, which occurred continuously throughout the night as bats were captured. Bats were released immediately following data collection.

For each bat captured, we recorded sex, age (adult or juvenile), and reproductive condition (female – pregnant, lactating, post-lactating, or non-reproductive (parous or nulliparous); male – non-reproductive or scrotal; (Racey, 2009). Sex was determined by observing genitalia. Age was assessed by observing epiphyseal plates of the metacarpal-phalangeal joints (Brunet-Rossinni & Wilkinson, 2009). Bats were classified as juvenile when there was a translucent cartilaginous gap in these joints. We use juvenile synonomously with flying young or volant young of year. Reproductive condition of females was determined by examining nipples and palpating the abdomen. Females were classified as pregnant when the abdomen was enlarged and a fetus could be felt (Gannon, 2003). Bats were classified as lactating if mammary glands were visible under the skin as a yellow disk and/or there was balding around the nipples attributable to the suckling of pups. Post-lactating females retained enlarged nipples, but no milk could be seen through the skin. Additionally, nipples of post-lactating females were dark and hair around the nipples was re-growing (Racey, 2009). Female bats were classified as non-reproductive if they did not appear to be pregnant, lactating, or post-lactating. Non-reproductive females were assessed and recorded as parous if there were signs of past reproduction, or nulliparous if there were no signs of past reproduction. Parous and nulliparous females were combined and classified as non-reproductive for analysis. Enlarged testes indicated that males were scrotal and reproductive while absence of enlarged testes indicated that males were non-reproductive (Sherman, 1937). Presence of a visible gular gland, found in the suprasternal neck region, was another indicator of reproductive males (Krutzsch, Fleming, & Crichton, 2002).

All research was completed under protocols approved by Christopher Newport University’s Animal Care and Use Committee under protocols 2015-9 and 2018-7, and Nevada Department of Wildlife scientific collection permit #39645; and followed the guidelines of the American Society of Mammalogists (Sikes & The Animal Care and Use Committee of the American Society of Mammalogists, 2016).

Data analysis – Bat captures were modeled using binomial generalized linear mixed models with a logit function. The logit function ensures positive fitted values and the binomial distribution is typically used for proportional count data. Response variables were captures of juvenile, adult female, pregnant, lactating, post-lactating, and non-reproductive females relative to total captures by night. Juvenile captures were modeled as the juvenile proportion of all individuals and adult female captures as the female proportion of all adults. Female reproductive condition was modeled as captures by reproductive class relative to all adult female captures. Trap night (Julian day) was used as the explanatory variable, modeled as quadratic and cubic functions to delineate seasonal variation and peaks in proportional abundance. When trapping extended past midnight, night was defined as the Julian day when trapping started. Year was modeled as an intercept-only random effect. Our models followed the general form:

· Ci ~ Binomial (pi, Ni)                                                                  (1)

· E(Ci) = pi                                                                                    (2)

· Logit(pi ) = nightj(i) +(nightj(i) ))2 + (nightj(i) ))3 + yeari             (3)

· Yeari ~ N(θ, σ2Year)                                                                  (4)

Where C is the number of bats captured per nighti for the group of interest, p is the proportion of that group captured per nighti, N is the total number of captures per nighti, night is the Julian date of the sampling night, and year is sample year. Night ranged from 159-299 and year from 2015-2019. Yeari is the random intercept, assumed to be normally distributed with a mean θ, and variance σ2.

To facilitate model convergence, night and year were constrained between -1 and 1 by centering each variable by the mean and scaling by the standard deviation. Scaled values were back-transformed for interpretation and graphical presentation.

Akaike Information Criterion (AICc) was used to guide model selection (Burnham & Anderson, 2002). Model inference included means and standard errors. Graphical results present model fit ± 95% confidence intervals. We used an exact binomial test of proportions to test the null hypothesis that sex, age, and reproductive condition were drawn from Bernoulli distributions with equal group proportions. Analyses were completed with Program R (R Core Team, 2019). The R package ‘nlme’ (Pinheiro et al., 2019) was used for generalized linear mixed models and the R package “MuMIn” (Barton, 2018) was used for AICc.