Operator bias in software-aided bat call identification

Factors influencing the detection and occupancy of little brown bats (Myotis lucifugus)

Using acoustics to survey for bats has increased as the need for data on increasingly rare species has also increased. We set out to better understand the difference between mist netting and acoustic detection probabilities between these two methods for the little brown bat (Myotis lucifugus), a species highly impacted by white-nose syndrome and currently considered for federal listing in the United States. We also analyzed occupancy relationships with local and landcover variables. We surveyed 15 sites using mist netting paired with an acoustic recorder for multiple nights to estimate detection probability of this species. We also deployed acoustic recorders at another 73 sites. We found that detection rates for mist netting were very low but increased with day of year and decreased from proximity to water. Acoustic surveys had higher detection rates, but there was an approximately 30% likelihood of false-positive detections. At the mean distance to water and day of year, acoustic surveys had a detection rate 55 times higher than mist netting. There were not significant factors influencing occupancy of little brown bats, only a slight positive relationship between forested largest patch, landscape patch richness and forest basal area. Given the declines in little brown bat populations since white-nose syndrome, it is even more critical that we consider the very low detection rate of mist netting compared to acoustic surveys.

Sampling methodology influences habitat suitability modeling for chiropteran species

Technological advances increase opportunities for novel wildlife survey methods. With increased detection methods, many organizations and agencies are creating habitat suitability models (HSMs) to identify critical habitats and prioritize conservation measures. However, multiple occurrence data types are used independently to create these HSMs with little understanding of how biases inherent to those data might impact HSM efficacy. We sought to understand how different data types can influence HSMs using three bat species (Lasiurus borealis, Lasiurus cinereus, and Perimyotis subflavus). We compared the overlap of models created from passive-only (acoustics), active-only (mist-netting and wind turbine mortalities), and combined occurrences to identify the effect of multiple data types and detection bias. For each species, the active-only models had the highest discriminatory ability to tell occurrence from background points and for two of the three species, active-only models preformed best at maximizing the discrimination between presence and absence values. By comparing the niche overlaps of HSMs between data types, we found a high amount of variation with no species having over 45% overlap between the models. Passive models showed more suitable habitat in agricultural lands, while active models showed higher suitability in forested land, reflecting sampling bias. Overall, our results emphasize the need to carefully consider the influences of detection and survey biases on modeling, especially when combining multiple data types or using single data types to inform management interventions. Biases from sampling, behavior at the time of detection, false positive rates, and species life history intertwine to create striking differences among models. The final model output should consider biases of each detection type, particularly when the goal is to inform management decisions, as one data type may support very different management strategies than another.

Implementing and Assessing the Efficacy of the North American Bat Monitoring Program

Bats are under threat from habitat loss, energy development, and the disease white-nose syndrome; therefore, an efficient and effective means to monitor bat populations is needed. The North American Bat Monitoring Program (NABat) was initiated in 2015 to provide standardized, large-scale monitoring to benefit bat biologists, managers, and policy makers. Given the recency of this program, our first objective was to determine the efficacy of implementing NABat. Further, because the probability of detecting a bat varies among species and survey conditions, our second objective was to determine factors affecting detection probabilities of bats using NABat acoustic surveys. We conducted surveys across South Carolina from mid-May through July 2015 and 2016. To determine efficacy of NABat, we compared species detections with historical known distributions and predicted distributions based on environmental occupancy models. To determine factors that affected detection probability, we evaluated support for predictive detection models for each species or species grouping. In general, we found that predicted distributions closely matched known distributions. However, we detected some species in ≤50% of cells within their ranges and others outside their ranges, suggesting NABat may also reveal new information about species distributions. Most species had higher detection probabilities at stationary points than mobile transects, but the influence of interrupted surveys, environmental conditions (e.g., temperature, rainfall, and wind) and habitat conditions often varied among species. Overall, our results suggest NABat is an effective and efficient method for monitoring many bat species, but we suggest that future efforts account for species-specific biological and behavioral characteristics influencing detection probability.

Improving geographically extensive acoustic survey designs for modeling species occurrence with imperfect detection and misidentification

Acoustic recording units (ARUs) enable geographically extensive surveys of sensitive and elusive species. However, a hidden cost of using ARU data for modeling species occupancy is that prohibitive amounts of human verification may be required to correct species identifications made from automated software. Bat acoustic studies exemplify this challenge because large volumes of echolocation calls could be recorded and automatically classified to species. The standard occupancy model requires aggregating verified recordings to construct confirmed detection/non-detection datasets. The multistep data processing workflow is not necessarily transparent nor consistent among studies. We share a workflow diagramming strategy that could provide coherency among practitioners. A false-positive occupancy model is explored that accounts for misclassification errors and enables potential reduction in the number of confirmed detections. Simulations informed by real data were used to evaluate how much confirmation effort could be reduced without sacrificing site occupancy and detection error estimator bias and precision. We found even under a 50% reduction in total confirmation effort, estimator properties were reasonable for our assumed survey design, species-specific parameter values, and desired precision. For transferability, a fully documented r package, OCacoustic, for implementing a false-positive occupancy model is provided. Practitioners can apply OCacoustic to optimize their own study design (required sample sizes, number of visits, and confirmation scenarios) for properly implementing a false-positive occupancy model with bat or other wildlife acoustic data. Additionally, our work highlights the importance of clearly defining research objectives and data processing strategies at the outset to align the study design with desired statistical inferences.

Manual analysis of recorded bat echolocation calls: summary, synthesis, and proposal for increased standardization in training practices

Automated recording units are frequently used for passive acoustic monitoring of taxa, including bats. Detection and species-level identification of bat echolocation calls using manual techniques is a common practice, often supplementing automated analysis by software. However, few standardized protocols exist for manual analysis, which is challenging for novices and impedes comparisons among research groups. In this two-part review, I first summarize and synthesize current approaches to manual call analysis. Three observations about the processes used to conduct manual call identification emerge: (1) there are significant knowledge gaps and few comparisons of interoperator variability; (2) they are individual- and location-specific, with no standardized underlying framework; and (3) they are often not well-described in the peer-reviewed literature. In response to these observations, I then conduct a comparative analysis of the fields of clinical reasoning (the study of medical decision-making) and the identification of bat echolocation calls. Clinical reasoning is a mature area of research and findings from this field may inform practices and instructional strategies for manually identifying echolocation calls. I demonstrate similarities between clinical reasoning and call identification processes and then make recommendations on how to apply findings from the clinical reasoning literature to call identification practices and training.

Weather conditions determine attenuation and speed of sound: Environmental limitations for monitoring and analyzing bat echolocation

Echolocating bats are regularly studied to investigate auditory-guided behaviors and as important bioindicators. Bioacoustic monitoring methods based on echolocation calls are increasingly used for risk assessment and to ultimately inform conservation strategies for bats. As echolocation calls transmit through the air at the speed of sound, they undergo changes due to atmospheric and geometric attenuation. Both the speed of sound and atmospheric attenuation, however, are variable and determined by weather conditions, particularly temperature and relative humidity. Changing weather conditions thus cause variation in analyzed call parameters, limiting our ability to detect, and correctly analyze bat calls. Here, I use real-world weather data to exemplify the effect of varying weather conditions on the acoustic properties of air. I then present atmospheric attenuation and speed of sound for the global range of weather conditions and bat call frequencies to show their relative effects. Atmospheric attenuation is a nonlinear function of call frequency, temperature, relative humidity, and atmospheric pressure. While atmospheric attenuation is strongly positively correlated with call frequency, it is also significantly influenced by temperature and relative humidity in a complex nonlinear fashion. Variable weather conditions thus result in variable and unknown effects on the recorded call, affecting estimates of call frequency and intensity, particularly for high frequencies. Weather-induced variation in speed of sound reaches up to about ±3%, but is generally much smaller and only relevant for acoustic localization methods of bats. The frequency- and weather-dependent variation in atmospheric attenuation has a threefold effect on bioacoustic monitoring of bats: It limits our capability (1) to monitor bats equally across time, space, and species, (2) to correctly measure frequency parameters of bat echolocation calls, particularly for high frequencies, and (3) to correctly identify bat species in species-rich assemblies or for sympatric species with similar call designs.

Bat detective-Deep learning tools for bat acoustic signal detection

Passive acoustic sensing has emerged as a powerful tool for quantifying anthropogenic impacts on biodiversity, especially for echolocating bat species. To better assess bat population trends there is a critical need for accurate, reliable, and open source tools that allow the detection and classification of bat calls in large collections of audio recordings. The majority of existing tools are commercial or have focused on the species classification task, neglecting the important problem of first localizing echolocation calls in audio which is particularly problematic in noisy recordings. We developed a convolutional neural network based open-source pipeline for detecting ultrasonic, full-spectrum, search-phase calls produced by echolocating bats. Our deep learning algorithms were trained on full-spectrum ultrasonic audio collected along road-transects across Europe and labelled by citizen scientists from www.batdetective.org. When compared to other existing algorithms and commercial systems, we show significantly higher detection performance of search-phase echolocation calls with our test sets. As an example application, we ran our detection pipeline on bat monitoring data collected over five years from Jersey (UK), and compared results to a widely-used commercial system. Our detection pipeline can be used for the automatic detection and monitoring of bat populations, and further facilitates their use as indicator species on a large scale. Our proposed pipeline makes only a small number of bat specific design decisions, and with appropriate training data it could be applied to detecting other species in audio. A crucial novelty of our work is showing that with careful, non-trivial, design and implementation considerations, state-of-the-art deep learning methods can be used for accurate and efficient monitoring in audio.

Insectivorous bats respond to vegetation complexity in urban green spaces

Structural complexity is known to determine habitat quality for insectivorous bats, but how bats respond to habitat complexity in highly modified areas such as urban green spaces has been little explored. Furthermore, it is uncertain whether a recently developed measure of structural complexity is as effective as field‐based surveys when applied to urban environments. We assessed whether image‐derived structural complexity (MIG) was as/more effective than field‐based descriptors in this environment and evaluated the response of insectivorous bats to structural complexity in urban green spaces. Bat activity and species richness were assessed with ultrasonic devices at 180 locations within green spaces in Vienna, Austria. Vegetation complexity was assessed using 17 field‐based descriptors and by calculating the mean information gain (MIG) using digital images. Total bat activity and species richness decreased with increasing structural complexity of canopy cover, suggesting maneuverability and echolocation (sensorial) challenges for bat species using the canopy for flight and foraging. The negative response of functional groups to increased complexity was stronger for open‐space foragers than for edge‐space foragers. Nyctalus noctula, a species foraging in open space, showed a negative response to structural complexity, whereas Pipistrellus pygmaeus, an edge‐space forager, was positively influenced by the number of trees. Our results show that MIG is a useful, time‐ and cost‐effective tool to measure habitat complexity that complemented field‐based descriptors. Response of insectivorous bats to structural complexity was group‐ and species‐specific, which highlights the need for manifold management strategies (e.g., increasing or reinstating the extent of ground vegetation cover) to fulfill different species’ requirements and to conserve insectivorous bats in urban green spaces.

The effect of call libraries and acoustic filters on the identification of bat echolocation

Quantitative methods for species identification are commonly used in acoustic surveys for animals. While various identification models have been studied extensively, there has been little study of methods for selecting calls prior to modeling or methods for validating results after modeling. We obtained two call libraries with a combined 1556 pulse sequences from 11 North American bat species. We used four acoustic filters to automatically select and quantify bat calls from the combined library. For each filter, we trained a species identification model (a quadratic discriminant function analysis) and compared the classification ability of the models. In a separate analysis, we trained a classification model using just one call library. We then compared a conventional model assessment that used the training library against an alternative approach that used the second library. We found that filters differed in the share of known pulse sequences that were selected (68 to 96%), the share of non-bat noises that were excluded (37 to 100%), their measurement of various pulse parameters, and their overall correct classification rate (41% to 85%). Although the top two filters did not differ significantly in overall correct classification rate (85% and 83%), rates differed significantly for some bat species. In our assessment of call libraries, overall correct classification rates were significantly lower (15% to 23% lower) when tested on the second call library instead of the training library. Well-designed filters obviated the need for subjective and time-consuming manual selection of pulses. Accordingly, researchers should carefully design and test filters and include adequate descriptions in publications. Our results also indicate that it may not be possible to extend inferences about model accuracy beyond the training library. If so, the accuracy of acoustic-only surveys may be lower than commonly reported, which could affect ecological understanding or management decisions based on acoustic surveys.