1 INTRODUCTION

When developing best practices for conservation, it is important to be able to conduct effective, precise surveys and properly interpret the results. For bats, survey methods have evolved from simple mist net setups (Greenhall & Paradiso, 1968) to complex arrays of acoustic microphones (Corcoran & Weller, 2018). Our capacity for analysis of data has also advanced from straightforward single species and season models to intricate models capable of community level analyses (MacKenzie et al., 2017). In addition to how we collect and analyze data, it is also important to consider the context of our research and interpret how results may differ within a species range. This is especially true for species that range across multiple ecological regions, such as the little brown bat (Myotis lucifugus; Fenton & Barclay, 1980).

The little brown bat has been historically common throughout its range up until the onset of white-nose syndrome (Pseudogymnoascus destructans), which was confirmed in Illinois in 2013 (Kath, 2013). This species has adapted to frequently roosting in human structures such as barns and attics (Fenton & Barclay, 1980). Little brown bats migrate in the fall to caves to spend the winter months, where their survival is dependent on mitigating the impacts of frequent arousal from white nose syndrome (Auteri & Knowles, 2020). Little brown bats have experienced devastating population declines from this fungus (Frick et al., 2010), prompting a review of the species for protection through the United States Fish and Wildlife Service (https://ecos.fws.gov/ecp/species/9051). Given the current interest in little brown bats, it is important that we understand how to effectively survey this species with available technology and understand habitat selection relationships. We selected little brown bats for our analysis because they were frequently captured in the mist nets compared to other species at the time of survey, right at the time of confirmation of white-nose syndrome in Illinois, meeting the minimum threshold for capture to estimate detection probability.

A combination of multiple methods for observational surveys is widely used throughout taxa to account for survey biases (Celis-Murillo et al., 2009; Clare et al., 2017; Guzy et al., 2014; Nichols et al., 2008). While acoustic recorders can passively record vast amounts of data with minimal field effort, these files can be difficult to use for analysis due to inability to estimate populations and difficulty with some species identification (Lemen et al., 2015; Nocera et al., 2019; Russo et al., 2017). Mist netting requires greater survey effort in the field by a permitted biologist and typically yields lower detection of species (Murray et al., 1999; O'Farrell & Gannon, 1999). This technique is most effective in forested corridors that bats use to commute between roost and foraging habitat (Kunz & Kurta, 1998), limiting ability to survey different habitat types, including agricultural fields, grasslands and residential or developed areas. In order to provide data for multiple land cover types necessary for species habitat modeling, acoustic surveys are necessary. Although mist netting and acoustics are commonly used to survey bats, detection rates need to be regularly assessed as populations are in flux and available software and technology is constantly changing. The development of false-positive occupancy modeling (Miller et al., 2011) lends itself well to the survey options available for bats (Clement et al., 2014). Mist netting provides a certain detection method, fitting the typical occupancy modeling framework (MacKenzie et al., 2002). While acoustics can provide additional data in land cover types where mist netting is less effective, with uncertain (possible false-positive) species identification.

We used both mist-netting and acoustic surveys to estimate the detection probability for the little brown bat, a species once common throughout its range but currently under review for federal protection. In addition to investigating factors influencing detection of this species, we also considered how local and landscape level metrics may impact occupancy. Little brown bats have been associated with edge habitat (Nelson & Gillam, 2017), urban environments (Coleman & Barclay, 2011), and open water (Bergeson et al., 2013). In our study we used the paired survey design in habitats where mist netting is most effective—forested corridors typically near water. With additional acoustic recorders in other types of land cover such as grassland, residential, and agricultural fields, to comprehensively survey the available landscape for occupancy analysis. We hypothesized that date would influence detection as maternal condition changes through the active season and young become volant (Deeley et al., 2022; Fenton & Barclay, 1980) increasing the number of individuals in the population for detection (MacKenzie et al., 2017). This was also a consideration for detection due to bats using waterways as flight corridors and foraging on aquatic insects (Hagen & Sabo, 2011). Since vegetation can minimize acoustic detection range (Weller & Zabel, 2002), we considered that density of trees at the survey location might have an interactive effect with survey method. We also included survey night in our analysis as these factors have been proposed to impact detection (Kaiser & O'Keefe, 2015; La Val, 1970). Our main goal is to determine detection probability of the two survey methods with secondary interest in other factors that influence detection and occupancy of the little brown bat.

2 METHODS

2.1 Study area

We sampled for bats in McHenry County in northeastern Illinois, USA during the summers of 2013 and 2014 (Figure 1). McHenry County has a heterogeneous landscape providing a diversity of survey locations including developed, agriculture, grassland/prairies, and forested sites. This was crucial to our study design, as we wanted to avoid surveying in a single landcover type to understand how bats utilize habitat across a Midwestern landscape. All mist netting occurred on McHenry Country Conservation District property, most sites containing bottomland forest and water features. We surveyed at 14 unique parks with a mean size of 360 hectares, the smallest being 4 hectares and the largest 1375 hectares. Eight species of bats have been recorded in McHenry County: the big brown bat (Eptesicus fuscus), evening bat (Nycticeius humeralis), Eastern red bat (Lasiurus borealis), hoary bat (Lasiurus cinereus), little brown bat, northern long-eared bat (Myotis septentrionalis), silver-haired bat (Lasionycteris noctivagans), and tri-colored bat (Perimyotis subflavus; Feldhammer et al., 2015).

3 RESULTS

We captured 59 little brown bats at four of the 16 mist net sites for a naïve occupancy of 0.25, with all of the captures occurring on 11 of 56 survey nights. We deployed acoustic recorders in 29 forested sites, 32 grasslands, 13 developed/residential sites, 10 agricultural fields and 4 wetlands. The acoustic recorders detected little brown bats on 100 of 220 survey nights at 48 of 88 sites for a naïve occupancy of 0.55.

Of the 10 detection models, the top model (w = 0.997) included method, survey date, and distance to water (Table 1). The method and date model was second (ΔAICc = 13.33, w = 0.001), and the method only model was ranked seventh (ΔAICc = 20.74, w = 0.001). The null model was ranked last with a ΔAICc of 50.74. At the mean survey date (189) and mean distance to water (205 m), the estimated detection probability for mist netting was 0.006 ± 0.007 and 0.332 ± 0.144 for acoustics (Figure 2). The relationship between detection and distance to water was negative, with higher detection closer to water and positive with day of year (Figure 3). Around mid-July is when mist netting detection surpasses 0.50 at a 0 m distance to water. The estimated detection probability at the mean survey date (189) at 0 m from water is 0.31, while at 100 m away, the detection drops to 0.04. The false-positivity rate was estimated at 30.8 ± 0.05%.

In the full candidate set of models including detection and occupancy, there were nine models within 2 ΔAICc, including the top detection only model, p(Method + date + Water), which ranked third (Table 2). The top model included only forested largest patch index and the second model included forested largest patch index and landscape patch richness. The model-averaged estimates for occupancy using the models within 2 ΔAICc identified only slight positive relationships with largest forested patch, landscape patch richness, and average DBH (Figure 4).

4 DISCUSSION

The use of acoustic data for bats is critical to understanding bat distribution patterns, due to the limitations of mist netting to provide adequate capture data with reasonable survey effort for statistical analysis (Weller & Lee, 2007) and to survey outside of forested corridors. The paired method survey design provides necessary data for effective calculation of detection probability and false-positive detection rate. With little brown bats under review by the United States Fish and Wildlife Service for federal protection, it is imperative we understand best survey practices to monitor these populations. Our detection probability estimates highlight the low success rate for capture of little brown bats, a species that during the time of survey was still relatively abundant before the onset of white-nose syndrome. This corroborates similar studies conducted in the western United States confirming low detection for this species (Rodhouse et al., 2012; Weller, 2008). Given population trends since the time of survey, current detection rates are probably much lower since abundance impacts detection rate (MacKenzie et al., 2017) and we know this species has experienced significant declines in areas impacted by white-nose syndrome (Deeley et al., 2021; Frick et al., 2010). According to simulation models by MacKenzie et al. (2002) for accurate estimates of occupancy, detection probability needs to be >0.3 for only two site visits where occupancy is at least >0.7. Our estimates for detection, especially for mist netting were low, as well as occupancy estimates, with most values less than 0.6. The calculation of these detection rates would not be possible from mist-netting in the Midwest post white-nose syndrome as capture rates of little brown bats are so low (Pettit & O'Keefe, 2017).

The negative relationship with distance to water and positive relationship with day of year for detection suggests these should be factors in calculating effective survey effort. We suggest that there may need to be greater survey effort at sites further from water due to the decreasing detection rate as distance to water increases if little brown bats are added to current United States Fish and Wildlife guidance (USFWS, 2023). Site selection for acoustics should prioritize locations near water or may require longer survey effort. All of our survey sites were within 1 km of water so additional data on greater distances away would be beneficial. With the United States Fish and Wildlife guidelines identifying the survey season as 15 May to 15 August (USFWS, 2023), there may be benefits to increasing levels of survey effort early in the field season. Detection probability increasing later in the summer is likely due to the young of the year taking flight (Kaiser & O'Keefe, 2015; Pauli et al., 2017)., increasing abundance and likelihood of detection (Clement et al., 2014) or possibly changes in foraging patterns through maternal progress (Deeley et al., 2022).

Our estimated false-positive detection rate of 30% is higher than the 14% reported by Clement et al. (2014). While they used manual review to verify detections, we chose to only use auto-identification output for our estimates as many organizations do not have access to trained biologists and manual review introduces additional bias to be estimated (Farmer et al., 2012; Fritsch & Bruckner, 2014), including variability among reviewers of bat echolocation files.

The lack of strong associations for occupancy supports the little brown bat being a habitat generalist, which may explain why there are conflicting studies on habitat preference for this species. For example, being identified as abundant (Coleman & Barclay, 2011) and not abundant in urban areas (Kurta & Teramino, 1992). This species has adaptive roosting strategies compared to other Myotis which allows more flexibility in habitat preference (Bergeson et al., 2015). Given the vast species range of the little brown bat, it makes sense that it would be capable of using different habitat types depending on availability of resources (Fenton & Barclay, 1980). We recommend that these estimates of detection and occupancy are considered when determining survey effort required for this species. We also suggest that research continues to use a paired method design to continue to explore how detection rates change with species, abundance, region, equipment and software used (Adams et al., 2012; Lemen et al., 2015).

AUTHOR CONTRIBUTIONS

Tara C. Hohoff: Data curation (lead); formal analysis (lead); writing – original draft (lead). Jill L. Deppe: Conceptualization (equal); funding acquisition (lead); methodology (equal).

CONFLICT OF INTEREST STATEMENT

There are no competing interests to disclose regarding this manuscript.