Call me maybe? Kuhl’s and Nathusius’ pipistrelles’ calls and the future challenges for classifiers

This project aimed to collect recordings of two species of bat within Jersey, Nathusius’ pipistrelle (Pipistrellus nathusii) and Kuhl’s pipistrelle (Pipistrellus kuhlii), aiding identification of these species which overlap in call frequency. Recordings were uploaded through the BTO pipeline, a classifier which includes bat calls alongside small mammals and insects, so to be able to mitigate the issue of by-catch from using static recorders. The BTO pipeline is unique as a classifier as it also standardises confidence levels across the recorded species.
A sample of randomly selected calls which the BTO pipeline identified as from these species were then moderated with the use of an rshiny app and audited manually with assistance from local experts.
Overall, 507858 of recordings were identified, 268,595 of which were identified as bat species through the classifier.
More Kuhl’s recordings (n= 3230) were collected than Nathusius’ (n= 468), which fits with the expectations of the populations in Jersey, with Kuhl’s pipistrelle thought to be more common.
The average kHz of the randomly selected Nathusius’ recordings was 41.01146(n=165), compared with the average kHz of the randomly selected Kuhl’s recordings was 39.52122 (n=165). Nathusius’ pipistrelle call duration was found to be significantly longer than that of Kuhl’s pipistrelles (p<0.001, df=2389, t=-3.438).
The difference between the static location and number of calls was significant (p<0.001, df=7, F= 94.377). The static location with the most of these species were for both species, St Andrews Park. Followed by Blanc Pignon (Kuhl’s) and Oakwood (Nathusius’). In terms of transects, Queens Valley had the highest frequency of calls of Nathusius’, whereas St Brelade transect had the highest frequency of calls of Kuhl’s.
The classifier error rate was found to be 12.06% when the randomly sampled subset was manually audited. In the future, classifiers could become more locally relevant with the addition of recording data from the locality, which would increase the accuracy.
Although no new definite differences between the two species were found using the recordings, locations were recommended for trapping in the future to confirm the species and their relative activity levels. The amount of data collected from this project will go towards future projects; for example, there was a large amount of cricket recordings collected, some of which (Grey bush cricket) are deemed rare for the island.