Select a classifier that best fits the expected species in your study area. Even with high quality recordings, and because of overlapping call characteristics for many species*, SonoBat still may only classify with a 95-98% correct classification rate and you should never rely on the autoclassification results to base id’s for out of range or rare species. Any species that you can exclude from consideration will increase the performance in classifying the remaining species that you expect to have in a your area. For further information regarding classification of western North America species see the SonoBat 3.1 NW classification notes.
1 Coto vocalizes low amplitude calls that seldom provide robust signals unless they happen to fly very close to a microphone. SonoBat outputs Coto results somewhat liberally with the intent to serve as an alert for a potential detection, but with the expectation that these results require manual confirmation, and the relative rarity of Coto does not present much of a burden to do that. Do not accept an automated classification result for Coto without manual confirmation. Coto produces feature-thin calls that just run from about 40 to 20 kHz, and out of range bats can easily leave fragments that leave pieces like that, as well as some insects and noise. To manually confirm Coto classification, check that they appear associated with a normal sequence of calls, and have a complete sequence of search phase calls; sequences with harmonics typical of Coto (compare with reference views) provide the highest confidence for confiming identification. Approach phase calls from Tabr can mimic Coto search phase calls. If Tabr approach phase you would see a transition of call types, and theCoto-like calls would likely have a call to call interval less than the 50–100 msec more typical of Coto search phase calls.
2 As with Coto, SonoBat produces Euma results somewhat liberally with the intent to serve as an alert for a potential detection, but with the expectation that these results require manual confirmation, and the relative rarity of Euma does not present much of a burden to do that. Some insects and noise sources can generate signals that overlap with Euma call characteristics and result in misclassifications. Check Euma classification results to make sure they appear associated with a normal sequence of calls and have consistent characteristics with Euma reference views and features listed in the acoustic characteristic table distributed with SonoBat. Do not accept an automated classification result for Euma without manual confirmation.
3 SonoBat also produces Eupe results somewhat liberally with the intent to serve as an alert for a potential detection, but with the expectation that these results require manual confirmation. SonoBat does this because many ambient sounds mimic and overlap the longer, low frequency calls of Eupe and result in misclassifications. Check Eupe classification results to make sure they appear associated with a normal sequence of calls and have consistent characteristics with Eupe reference views and features listed in the acoustic characteristic table distributed with SonoBat. Do not accept an automated classification result for Eupe without manual confirmation.