Figure 1. Map of study area, located in east-central Illinois, USA, indicating the location of the 50 bat detectors (white circles) placed along 10 transects. Transects followed roads, starting at a forested area (green) and into uninterrupted row crop agriculture (brown) at 0m, 500m, 1000m, 2000m and 4000m from the forest edge.
Illinois Bat Community
Illinois has 13 insectivorous bat species (Whitaker and Hamilton, 1998), of which four are known to forage in open habitats such as agricultural landscapes: big brown bats (Eptesicus fuscus , EPFU), silver-haired bats (Lasionycteris noctivagans , LANO), eastern red bats (Lasiurus borealis , LABO), and hoary bats (Lasiurus cinereus , LACI). These species have high wing aspect ratios or long and narrow wings, resulting in decreased maneuvering abilities, high flight speeds, and sustained flight endurance capabilities, all traits that facilitate foraging in open habitat (Loeb and O’Keefe, 2011). These larger-bodied bats have been documented to consume a range of agricultural pests such as June beetles (Scarabidae ), cucumber beetles (Chrysomelidae ), Asiatic oak weevil (Curculionidae ), corn earworm moth (Helicoverpa zea ), and stink bugs (Pentatomidae ; reviewed by (Kunz et al., 2011). While sustained flight endurance should facilitate foraging in open expanses of agricultural landscapes, all open-foraging species rely on trees as roosting habitat and, thus, forest availability is a factor that could limit bat presence and activity on the landscape. Eastern red bats and hoary bats are foliage-roosting species that roost in tree canopies, hanging from leaves, twigs, and branches (Mager and Nelson, 2001; Willis and Brigham, 2005). While big brown bats use tree cavities and crevices as roosts (Willis and Brigham, 2004), this species often roosts in buildings, including barns (Benedict et al., 2017). Small-bodied bats, consisting mainly of Myotis species, generally remain in or close to forested habitat (Beilke et al., 2021; Henderson and Broders, 2008). Although they consume fewer large agricultural pest species, small-bodied bats are extremely important in consuming defoliating arthropods in forests (Hughes et al., 2021; Kalka et al., 2008).
Study design
We selected 10 transects, with each beginning at a prominent riparian corridor and heading perpendicularly into agricultural landscapes devoid of any large (> 1ha in size) natural features such as patches of forest or prairie (Fig. 1). For each transect, we deployed 5 passive acoustic bat detectors placed between roads and crop field edges at positions approximately 0m, 500m, 1000m, 2000m, and 4000m from the riparian corridor assumed to be frequented by bats (Fig. 1). The bat detectors were placed at a height of 3m, adjacent to corn and soy fields and facing the field interior.
To determine our study design, we piloted detector placement to test whether detectors placed adjacent to fields would differ in the number of bat recordings from those placed deeper within fields. We placed 7 pairs of detectors on properties, one adjacent to the field and one 100 m into the field, perpendicular to the road, for a total of 41 days. There was no difference in bat activity between detectors adjacent and those placed 100m inside of a field (z = 0.29, df = 170, p = 0.772).
Acoustic data
Acoustic detectors were deployed from May 29th to September 29th 2021. We used OpenAcoustics’ AudioMoth bat detectors (v1.1.0 and v1.2.0), deployed in sealed, waterproof cases (AudioMoth IPX7 Waterproof case). We recorded throughout the evening, from 6pm to 6am Central Daylight Time, sampling at a rate of 250kHz with medium gain in full-spectrum format. We used an amplitude threshold at 0.1% and a high-band filter at 40kHz – which reduces the amplitude of frequencies below 40kHz (AudioMoth operation manual, p14), and is recommended when monitoring a bat community where common low frequency calls are 16kHz or above. Settings were based on pilot trials in May, but this amplitude threshold was deemed far too low when analyzing data from July – resulting in more noise being recorded – however this threshold was kept to ensure consistent parameters across the study. Data were collected in 55s recordings with a sleep duration of 5s; these files were later split into segments of 15s or less (i.e., 4 parts or call files).
Acoustic data were analyzed with Kaleidoscope Pro (v.5.4.6; KPro) to allow for the use of automated identification (i.e., autoID). We followed the protocol proposed by the North American Bat Monitoring Program (NABat; Loeb et al., 2015) to analyze the acoustic files, using a minimum of 3 pulses, a pulse length range of 0–50ms, and a call frequency range of 10–120kHz. We used the standard species list for Illinois, USA for this analysis; this list includes EPFU, LABO, LACI, LANO, Myotis austroriparius (MYAU), Myotis grisescens(MYGR), Myotis lucifugus (MYLU), Myotis septentrionalis(MYSE), Myotis sodalis (MYSO), Nycticeius humeralis(NYHU), and Perimyotis subflavus (PESU); two species that occur only in southern Illinois are not on this list.
We further filtered the Kpro output to exclude any species-night-entries that had a maximum likelihood estimate of 1, as these are considered unlikely to be present based on the number of files identified as this species. Furthermore, because of a potentially high proportion of false-positive species identifications when using bat calls to ID bat species, we opted to analyze our data according to phonic groups of bat species, as described in Beilke et al. (2021). This commonly used classification technique is based on the characteristic frequency (Fc) of a bat species call: low frequency bats (Fc range: 18–30 kHz) include LACI, EPFU, and LANO; mid frequency bat species (Fc range: 23–43 kHz) include NYHU, PESU, and LABO; and high frequency bats (Fc range: 38–48 kHz) include all Myotis species. Bat activity was thus differentiated between three phonic groups: low, mid, and high frequency bats (Beilke et al., 2021).
Relative pest abundance
To estimate the relative abundance of crop pest species in our study area, we used a data set collected during our study period by the Crop Protection and Pest Management Program (Grant No. 2021-70006-35476) from the USDA National Institute of Food and Agriculture, in Champaign County, Illinois. The data was collected using pheromone lures to attract insects to metal Hartstack traps or nylon Heliothis traps around agricultural fields, targeting the main agricultural pest species (all large moths, wingspan range 24–45 mm; https://bugguide.net/) in the area: black cutworm (Agrotis ipsilon ), corn earworm (Helicoverpa zea ), European corn borer (Ostrinia nubilalis ), and fall armyworm (Spodoptera frugiperda ). Although these data were not specific for each of our study locations, we assume that the emergence timing of pest species is similar across the county. We applied a best-fitting curve to the data to predict pest abundance (number of individuals per trap night) for our data points based on date (Fig 2). We then standardized these data by dividing pest abundance by maximum pest abundance in the dataset, yielding a relative measure of pest abundance from 0 (no pest species present) to 1 (maximum pest abundance).
Covariates
Distance to forest was measured using remotely sensed images of the study area, through the GoogleEarth engine (images dated April 21st 2019 or later). Weather covariates were sourced from wunderground.com based on a weather station in Champaign, IL. We downloaded data for daily minimum and maximum temperature, mean and maximum daily windspeed, maximum daily humidity, and precipitation. Initial analyses revealed that only minimum daily temperature and precipitation were correlated with both bat activity and bat diversity, therefore only these two weather variables were retained in the statistical analyses.