Comparing the results of four widely used automated bat identification software programs in the North Sea region

Recently a few papers were published addressing the prudency needed when using automated identification software programs to analyse recordings of bat echolocation sounds. We want to contribute to that discussion by analysing a reference dataset of bat recordings with four widely used and commercially available software programs (BatIdent, BatExplorer, Kaleidoscope and Sonochiro). The reference data were all recorded in Western-Europe with a batcorder. For most of the recordings there was a visual confirmation of the recorded species. In a few other cases certainty was obtained because the specimens were captured and released or because the recordings were made in certain areas which were outside of the range of other species (e.g. at high altitude to separate Eptesicus serotinus and Eptesicus nilssonii). After running the different programs on the reference data, we compared the outputted results. Overall, identification of the recordings to species level in this test was best with Batident (81% correct identifications), followed by Kaleidoscope (71%), Sonochiro (63%) and BatExplorer (53%). We can conclude that each of the tested programs has its own strengths and weaknesses, but none of them should be used unsupervised. Outputted results need to be checked by a trained expert. In this way, our test affirms the conclusions of previous tests in Northern Europe and the USA.

Abstract of an Oral Presentation or a Poster

Poster presented at the European Bat Research Symposium - Donostia, Spain 1-5 August 2017