Automated echolocation classifiers vary in accuracy for northeastern U.S. bat species.

Acoustic surveys of bat echolocation calls are an important management tool for determining presence and probable absence of threatened and endangered bat species. In the northeastern United States, software programs such as Bat Call Identification (BCID), Kaleidoscope Pro (KPro), and Sonobat can automatically classify ultrasonic detector sound files, yet the programs' accuracy in correctly classifying calls to species has not been independently assessed. We used 1,500 full-spectrum reference calls with known identities for nine northeastern United States bat species to test the accuracy of these programs using calculations of Positive Predictive Value (PPV), Negative Predictive Value (NPV), Sensitivity (SN), Specificity (SP), Overall Accuracy, and No Information Rate. We found that BCID performed less accurately than other programs, likely because it only operates on zero-crossing data and may be less accurate for recordings converted from full-spectrum to zero-crossing. NPV and SP values were high across all species categories for SonoBat and KPro, indicating these programs' success at avoiding false positives. However, PPV and SN values were relatively low, particularly for individual Myotis species, indicating these programs are prone to false negatives. SonoBat and KPro performed better when distinguishing Myotis species from non-Myotis species. We expect less accuracy from these programs for acoustic recordings collected under normal working conditions, and caution that a bat acoustic expert should verify automatically classified files when making species-specific regulatory or conservation decisions.

Oceanic records of North American bats and implications for offshore wind energy development in the United States.

Offshore wind energy is a growing industry in the United States, and renewable energy from offshore wind is estimated to double the country's total electricity generation. There is growing concern that land-based wind development in North America is negatively impacting bat populations, primarily long-distance migrating bats, but the impacts to bats from offshore wind energy are unknown. Bats are associated with the terrestrial environment, but have been observed over the ocean. In this review, we synthesize historic and contemporary accounts of bats observed and acoustically recorded in the North American marine environment to ascertain the spatial and temporal distribution of bats flying offshore. We incorporate studies of offshore bats in Europe and of bat behavior at land-based wind energy studies to examine how offshore wind development could impact North American bat populations. We find that most offshore bat records are of long-distance migrating bats and records occur during autumn migration, the period of highest fatality rates for long-distance migrating bats at land-based wind facilities in North America. We summarize evidence that bats may be attracted to offshore turbines, potentially increasing their exposure to risk of collision. However, higher wind speeds offshore can potentially reduce the amount of time that bats are exposed to risk. We identify knowledge gaps and hypothesize that a combination of operational minimization strategies may be the most effective approach for reducing impacts to bats and maximizing offshore energy production.

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.