NEWSFLASH 22/04/2016: To download our “Recommendations for using automatic bat identification software with full spectrum recordings” written with Paola Reason and Kate Jones Click HERE.
The passive real-time detectors (Wildlife Acoustics SM2Bat+’s) used for this project are triggered when they detect sound within a certain frequency range. We define this as one recording, and a recording may contain any number of bat calls, or none if triggered by other species group (e.g. bush-crickets) or ambient high-frequency sounds. Using this technology, bat monitoring on this scale generates an enormous volume of recordings: often over 1,000 sound recordings per detector per night. To process such a volume of acoustic data has necessitated a semi-automated analyses to assign recordings to species followed by additional manual checking of recordings.
STEP 1 – FIRST ANALYSES
A first analysis (Step 1) is currently carried out using the software SonoChiro to assign each recording to species. SonoChiro uses a classification tree developed from call measurements taken from an extensive (>250,000) database of recordings of known species recorded on the same detector type to assign identity to all calls in the first 5 seconds of each recording. From these, recordings not containing bat recordings are identified, and where they do, each is assigned to species and given a confidence index related to the probability of correct classification (as compared with the underlying training database) on a scale of 0–10, but this is poorly defined. It is not linearly related to a probability of correct classification and may be suppressed if acoustically similar species are in SonoChiro’s reference sample, even if those species are known to be absent from the study area. Of particular relevance here are Kuhl’s Pipistrelle Pipistrellus kuhlii, Grey long-eared bat Plecotus austriacus and a number of Myotis species which are not thought to be present in the survey area, but are similar acoustically to species that are. Owing to these uncertainties we perform additional filtering and at the end of the season manual checking of recordings. This is an area of research where tools for processing recordings and knowledge of species identification is continuing to improve.
STEP 2 – RECORD FILTERING
Record filtering (Step 2) is performed in SAS (SAS 2012) using code that we have written to remove the following recordings: recordings with insufficient acoustic information, here judged to be those containing fewer than three bat calls; those containing sounds from other taxa (e.g. bush-crickets) or other ambient noises; or where the confidence index was zero (typically resulting from poor quality recordings). Recordings, with a confidence index of 1 or 2, judged to have “high” error risk (Biotope 2013) are reassigned to species group or genus and are not used further in the analyses.
STEP 3 – MANUAL CHECKING
Manual checking of sonograms (Step 3) using software SonoBat is then used as an independent check of the original species identities assigned by SonoChiro (Step 1) and the filtering (Step 2). For all but two species (Common Pipistrellus pipistrellus and Soprano pipistrelle Pipistrelle pygmaeus, all recordings retained after Step 2 are inspected with SonoBat and identity is checked and re-coded if necessary based on call parameters defined in Russ (2012) and Barataud (2015). See Table A1 for a summary of the most important call parameters used in Step 3). The call shape (hockey-stick shaped) and frequencies of Common and Soprano Pipistrelle Pipistrellus pygmaeus, are characteristic, particularly when considered in combination with other call parameters, so discrimination of these species should be good. For these two species, a sample of 1,000 recordings of each are selected at random and sonograms viewed in SonoBat. Once species identities have been checked by looking at individual recordings in isolation, similar species (e.g. Myotis and Nyctalus), where error is most likely are grouped together and identities looked at again in relation to sequences (timing) of recordings.
Figure 1 below provides an example of how combining peak frequency and band width in an analyses for Common, Soprano and Nathusius’ Pipistrelle, would provide better discrimination than considering peak frequency alone. In practice SonoChiro (Step 1) considers the potential of a wider range of call parameters for improving discrimination. Recordings of Whiskered Myotis mystacinus or Brandt’s bat Myotis brandtii are currently treated as a species pair. Feeding into this process, we have spent time in Paris with colleagues at the Muséum National d’Histoire Naturelle who run the French National Bat Monitoring Programme, to learn from their experience in the acoustic identification of bats, and considering the scarcity of records of Whiskered and Brandt’s bats from Norfolk and Scotland prior to this study, to ensure that they would confidently assign the same recordings to this species pair. Currently about 72% of recordings initially assigned to Whiskered / Brandt’s in Step 1 are reassigned to genus (Myotis species), but we accept that identification of all Myotis is challenging, and we recommend that identification is followed up where possible with more targeted (e.g mist-netting) to ground truth identification. Information on the number of recordings removed or reassigned to genus or species at each step of this process for the first two years of the Norfolk survey can be viewed in Table 2 (from Newson et al. 2015). For more information on the steps involved as applied to the first two years of Norfolk data and first outputs to come out of the project see Newson et al. (2015).