Population dynamics of little brown bats (Myotis lucifugus) at summer roosts: apparent survival, fidelity, abundance, and the influence of winter conditions
Data files
Nov 18, 2022 version files 297.29 KB
-
MYLU_robust_2014-2018_Dryad.txt
-
NCEI_NOAA_monthly_weather2013-2019_Dryad.txt
-
README_Dryad.txt
Abstract
- White-nose syndrome (WNS) has caused the death of millions of bats, but the impacts have been more difficult to identify in western North America. Understanding how WNS, or other threats, impact western bats may require monitoring other roosts, such as maternity roosts and night roosts, where bats aggregate in large numbers.
- Little brown bats (Myotis lucifugus) are experiencing some of the greatest declines from WNS. Estimating survival and understanding population dynamics can provide valuable data for assessing population declines and informing conservation efforts.
- We conducted a 5-year mark-recapture study of two M. lucifugus roosts in Colorado. We used the robust design model to estimate apparent survival, fidelity, and abundance to understand population dynamics, and environmental covariates to understand how summer and winter weather conditions impact adult female survival. We compared the fidelity and capture probability of M. lucifugus between colonies to understand how bats use such roosts.
- Overwinter survival increased with the number of days with temperatures below freezing (β > 0.100, SE = 0.003), and decreased with the number of days with snow cover (β < -0.40, SE < 0.13). Adult female fidelity was higher at one maternity roost than the other. Overwinter and oversummer adult female survival were high (>0.90), and based on survival estimates and fungal-swabbing results we believe these populations have yet to experience WNS.
- Recapture of M. lucifugus using antennas that continuously read passive integrated transponder tags allows rigorous estimation of bat population parameters that can elucidate trends in abundance and changes in survival. Monitoring populations at summer roosts can provide unique population ecology data that monitoring hibernacula alone may not. Because few adult males are captured at maternity colonies, and juvenile males have low fidelity, additional effort should focus on understanding male M. lucifugus population dynamics.
Methods
We conducted this study at two roosts within the Yampa Valley of northwest Colorado (Figure 2). One colony (herein called “house roost”) is in a 120-m2 ranch house at the Rehder Ranch (elevation 2,150 m) built in 1900. The house roost is adjacent to Catamount Lake, and approximately 17 km south of Steamboat Springs, Colorado. Bats roost in the spaces between the interior walls and exterior metal roofing material. The other colony (herein called “barn roost”, elevation 1,930 m) is in the Carpenter Ranch barn, which was built in 1903, is adjacent to the Yampa River, and is approximately 7 km east of Hayden, Colorado. Bats roost between timbers in the hayloft and between timbers and the metal roof.
We captured bats using harp traps and mist nets around the outside of the house roost, and within the barn roost. We captured bats in early summer (June) and when young-of-the-year (juveniles) were believed to be volant, but had not dispersed (late summer: late July/early August). We captured and marked bats twice a year from 2014 – 2018, except for late summer of 2017 at the house roost. Because the timing of M. lucifugus parturition varies each year, the timing of the late-summer capture event was adjusted to optimize juvenile captures. For each bat captured, we recorded the mass, sex, age as juvenile or adult (Brunet-Rossinni & Wilkinson, 2009), palpated and assessed the individual’s reproductive condition, and marked it with HPT9 (9 mm x 2 mm) 134.2 kHz passive integrated transponder (PIT) tag (Biomark, Inc., Boise, Idaho). We inserted tags subcutaneously below the scapula, and we sealed insertion sites with a biomedical glue (VetBond Tissue Adhesive, 3M Science, St. Paul, Minnesota). We conducted facial and wing swabbing and guano collection to assess the presence of the fungus (Pseudogymnoascus destructans; Pd) that causes WNS (analysis conducted by U.S. Geological Survey, National Wildlife Health Center, Madison, WI).
At the house roost, we detected tagged bats using a 7.6-m cord PIT-tag antenna system (IS1001, Biomark, Inc., Boise, Idaho; herein called “antenna”) that was stretched under the eaves. We installed this antenna in the summer of 2015, and it was disrupted for several weeks late in summer of 2015 and early 2016 because the solar panel became disconnected from the battery. At the barn roost, we varied the PIT-tag reading system’s configuration and placement over time. In 2014, we placed two 0.6 m x 0.6 m window-style PIT-tag readers (FS2001, Biomark, Inc., Boise, Idaho) at the top of the primary entrance (3.5 m x 3.5 m barn door opening), while netting was stretched along the lower section of the entrance to channel bat flight through or near the window-style readers. This arrangement produced few detections (< 60 detections/2 weeks), so in 2015, we installed a cord antenna system like that used at the house roost. We weaved the cord antenna across the barn door opening with approximately 0.25-m gap, then in August 2015 moved the antenna within the hayloft to reduce conflicts with on-going cattle operations. At both locations, we acquired additional detections using handheld PIT-tag readers to scan roosting bats in the hayloft of the barn roost, and at bat houses near the house roost.
We created encounter histories based on weeks of the year when individuals were captured or detected, thus there were 52 encounter periods throughout the year. We structured the data into weekly intervals because it allowed multiple reencounter (recapture) occasions throughout the summer. These multiple opportunities to reencounter tagged animals increase the opportunities to detect bats and improve estimates of detectability. We analyzed mark-recapture data using a Huggins robust design model in Program MARK (Kendall, 2001). The robust design model allows estimation of parameters, such as abundance, during the closed sessions when we assume there are no births, deaths, immigration, or emigration. Also, the robust design model allows estimation of other parameters, such as survival, immigration, emigration, and fidelity, during open sessions. Because the fate of juvenile bats may be strongly associated with the fates of their mothers, we estimated an overdispersion scaling parameter (ĉ), from the global model in a Cormack-Jolly-Seber analysis (White & Burnham, 2001). We compared competing models using quasi-likelihood Akaike’s information criterion with small sample size bias correction (QAICc) and the probability of a model being the most-parsimonious model (QAICc weight; wi – Burnham & Anderson, 2002).
We estimated oversummer and overwinter survival, because if WNS was impacting M. lucifugus populations, we expected overwinter survival to be lower than oversummer survival. To estimate oversummer survival, we structured the encounter data such that there was an open season between summer tagging events. To estimate overwinter survival, we used the open session between late summer and early summer tagging events. The two periods of population closure each year were based on the arrival of a substantial number of individuals (>30 individuals/week) after hibernation, and the departure of most bats before hibernation (<30 individuals/week remaining at the roost). The selection of 30 individuals is arbitrary, but chosen to designate when we believed bats were returning to the roost. The “early-summer closed period” began after the week when greater than 30 individuals were detected (late April to mid-May) and extended until the week after the capture and tagging event in early June. The time between early-summer (June) and late-summer (late July/early August) capture events was considered open because young were being born. The “late-summer closed session” began after the July/August capture event and ended the week when fewer than 30 individuals were detected in early fall (mid-August to mid-September).
Using the robust design model, we estimated capture probability (p), recapture probability (c), overwinter and oversummer apparent survival (φ), abundance (N), temporary immigration (1 – γ’), temporary emigration (γ”), and site fidelity (1 – γ”; the probability of being detected at time interval i, given you were detected in time interval i - 1). As a modeling approach, we first modeled capture and recapture probability, keeping apparent survival and movement parameters (γ’, γ”) as varying temporally. Once the best set of models of p and c were identified, we used those p and c configurations to model φ, γ’, γ”. Seasonal parameter estimates were transformed based on weekly estimates, and variances were estimated using the Delta method (Powell, 2007). Abundance at each roost was estimated for each of the two closed periods per year. For estimating juvenile φ we transitioned juveniles into adults the first year after they were captured, when they would be sexually mature (Humphrey & Cope, 1976). We ran 27 models for p and c, and we ran 89 models for φ and the movement parameters using the most parsimonious model of p and c.
Because weather can impact bat activity and survival, we used weather data from a Steamboat Springs weather station (National Oceanic and Atmospheric Administration National Centers for Environmental Information, Station No. GHCND:USCOOO57936; ncdc.noaa.gov) to model capture and recapture probabilities, and monthly weather summaries to model seasonal survival. We used number of days with precipitation >0.25 cm and >2.5 cm, number of days with maximum temperature >21°C and >32°C, and total monthly precipitation (in) to model oversummer φ. We used measures of winter severity, including number of days with low temperature < 0°C, number of days with low temperature < -32°C, number of days with high temperature < 0˚C, number of days with precipitation > 0.25 cm, number of days with precipitation >2.5 cm, number of days with snow depth ≥ 2.5 cm, and total precipitation as covariates for modeling overwinter survival. Additionally, we used the variance of winter covariates to assess how the variability of winter severity impacted parameters.
We used individual covariates of mass and body condition index (BCI: forearm length/mass) as covariates for seasonal survival (Pearce et al., 2008). Mass was measured at each capture event, but we did not begin measuring forearm lengths until 2016.