Guide · static detector data

Auto ID for static bat detector data: cut manual checking without losing rigour.

For UK static detector data, use an auto ID tool that filters noise files, gives every identification a confidence score and records who checked what, then vet the low-confidence calls and the hard groups (Myotis, Plecotus, Barbastella, Nyctalus and Eptesicus) by hand, as Bat Conservation Trust guidance expects. BioSonic is built for exactly this job: it runs in the browser, scored 98.7% F1 in a third-party benchmark by Wilder Sensing, and lets the whole team vet one shared project. Kaleidoscope Pro, the BTO Acoustic Pipeline, SonoBat UK and BatExplorer all have their place, and we say below where each one fits better.

Works with WAV and FLAC from Song Meter SM4BAT and Mini Bat, AudioMoth, Anabat, Batlogger and Pettersson detectors, plus Anabat .zc and .NN# zero-crossing files.

Last updated: 2026-11-17. Written by the BioSonic team.

The problem

Why static data swamps survey teams.

A transect gives you one evening of recordings. A static gives you every night it is out, from dusk to dawn, and most survey designs put several statics out at once. Every detector-night adds files, and every file has to be sorted before anyone can write the report.

A lot of those files are not bats. Triggers fire on wind, rain, insects, vegetation and electrical noise. On one survey night that the UK consultancy Wharton analysed side by side in two tools, around 3,000 of the 4,200 files were noise.7 Opening noise files one at a time is where most manual checking time goes.

3,000 of 4,200
files on one survey night were noise, not bats
Wharton case study, UK

How much data one static produces

These are the mid-range estimates per detector-night that our storage calculator uses. Real volumes vary with bat activity, trigger sensitivity, sample rate and night length.9

DetectorRecording modeGB per night (estimate)
Song Meter SM4BAT FSTriggered, full spectrum1.5
Song Meter Mini BatTriggered1.0
AudioMothTriggered1.8
AudioMothContinuous, 384 kHz, ten-hour night28
Batlogger M2Triggered0.8
Pettersson D500XTriggered2.5
Anabat SwiftTriggered, zero-crossing0.1

Twenty statics out for ten nights is 200 detector-nights, or roughly 300 GB at 1.5 GB a night. Over a whole season, Wharton collected 822,729 audio files, about 2 TB.7 At that scale, checking every file by hand is not realistic, and checking a random sample misses the rare species that matter most.

Good practice

A sound workflow: filter, identify, vet, report.

Auto ID is not the end of the analysis. Its job is to decide where the ecologist's time goes.

Filter noise first

Remove files with no bat calls before anyone looks at them. This is the biggest single saving, because noise is often the majority of what a static records.

Auto ID with a confidence score

Every bat recording gets a species suggestion and a confidence score, so you can work through the night in order of how sure the classifier is, not in date order.

Targeted vetting

Check low-confidence identifications, the look-alike groups and notable records on the spectrogram, plus at least one recording of each species on each survey date.1

Report the method

State what was automated, what was checked by hand and by whom, the percentage vetted and how many identifications were corrected.

Which calls to vet by hand

Bat Conservation Trust guidance says automated classifications should be manually verified, ideally at least one recording of each species on each survey date, and that this matters most for notable records. It also notes that for some species, such as whiskered bat (Myotis mystacinus), an audio recording generally cannot be verified conclusively, and the next step is a trapping survey under licence.1 In practice, these are the groups where a human should look at every pass, or at least every uncertain one:

GroupWhy it needs a humanVetting rule
MyotisCalls overlap in frequency. A 2024 test of three classifiers, Kaleidoscope Pro among them, found none could reliably identify Myotis to species.5Vet every pass. Report to genus (Myotis sp.) where species cannot be confirmed.
Plecotus (long-eared bats)A 2017 study found automated classifiers identified brown long-eared bat between 0 and 89% of the time, depending on the tool.4Vet every pass.
Barbastella (barbastelle)An Annex II species and usually a notable record. Its Type B call can look like Nathusius' pipistrelle.10Vet every pass.
Nyctalus and EptesicusNoctule, Leisler's bat and serotine all make low hockey-stick calls and are look-alikes of each other.10Vet low-confidence passes and every pass that changes your conclusions.
Pipistrelles and other common speciesThe bulk of most datasets. Errors here rarely change the assessment, but they add up.Vet low-confidence passes and at least one recording per species per survey date.

How BioSonic handles it

Built for static detector volumes.

Upload as recorded

Drag in WAV or FLAC files from Song Meter SM4BAT and Mini Bat, AudioMoth, Anabat, Batlogger and Pettersson detectors, or Anabat zero-crossing files (.zc and .NN# sequence files), which are decoded on upload. Any file length, any folder size, in the browser with nothing to install. For batcorder .raw files, ask us first.

Noise filtered automatically

A deep-learning noise model trained on 3.5 million non-bat recordings removes wind, rain, insects and equipment noise, so empty files never reach your review list.

Species suggestion with confidence

Image recognition on spectrograms, trained on over 2.5 million verified bat calls curated by 12 bat experts, suggests a species with a confidence score and can find up to 7 species in one recording. It covers all UK bat species.3

Feeding buzzes and social calls flagged

Echolocation, feeding buzzes and social calls are separated automatically, which helps when you need to show foraging or possible roost activity, not just presence.

One project, whole team

Field staff upload, analysts vet and the named ecologist signs off in the same shared project, with admin, editor and viewer roles. One organisation account and no per-seat fees.

Every decision logged

Confirm or correct an identification on the spectrogram and the change is recorded with the reviewer and the time. Export every identified recording with date and time to CSV or Excel, with charts and maps for the report.

Recordings and results are stored in the EU, and the model is retrained every weekend on newly verified data. You buy a package of analysis gigabytes. The bigger the package, the lower the price per GB. All features and all users are included, and your recordings are kept for 9 years. Ask us for a quote. Because every user is included, giving a junior analyst access to vet costs nothing extra.

Two honest limits. The benchmark figures were measured on full-spectrum recordings, and no zero-crossing benchmark has been published, so test zero-crossing data on your own labelled files. And BioSonic needs an internet connection: if you must process in the field offline, a desktop tool is the better fit.

AudioMoth

Identifying bat species from AudioMoth recordings.

AudioMoth is popular for statics because it is low cost and easy to deploy in numbers. That is also why AudioMoth projects swamp teams: many units, many nights, and settings that decide how much noise you record. A practical workflow:

  1. Choose triggered or continuous recording on purpose. Our storage calculator estimates about 1.8 GB a night for a triggered AudioMoth. Continuous recording over a ten-hour night is about 18 GB at 250 kHz and about 28 GB at 384 kHz, and most of it contains no bats.9
  2. Keep the files as recorded. Organise them by site and night when you pull the SD cards. There is no need to convert, rename or split them.
  3. Upload the WAV (or FLAC) files to one project per site or survey. BioSonic accepts any file length, so long continuous files do not need cutting into short segments first.
  4. Let noise filtering remove the empty files, then vet in confidence order, starting with the look-alike groups in the table above.
  5. Export species, date and time per recording to CSV or Excel for your activity analysis, or use the built-in charts and maps.

Other routes exist. The BTO Acoustic Pipeline accepts AudioMoth WAV files and is free within a yearly credit allowance if you share results with it.13 Kaleidoscope Pro reads full-spectrum WAV on the desktop. BatDetect2 is a free model trained on UK data, but its licence covers non-commercial use only, so it is not for paid consultancy work. Our AudioMoth bat call analysis guide covers recording settings and compares these options in more detail.

Planning and EIA

How to evidence the method for a planning authority.

Auto ID is defensible when the method is clear. The 4th edition of the BCT Good Practice Guidelines added a section on bat sound analysis and auto-identification, and says that departures from its methods are acceptable when the ecological rationale is clear, supported with evidence where they may face challenge.2 For an ecology report, an Environmental Statement chapter or a DCO application, that means three things on paper:

  • The method statement: which software, the date the analysis ran, what was filtered automatically, and your vetting rules.
  • The numbers: how many recordings were auto-identified, what percentage was checked by hand, and how many automatic identifications were corrected.
  • The audit trail: who vetted what, and when, so a reviewer can trace any record back to its spectrogram.

Example method statement wording

Edit the parts in square brackets to match your survey.

Static detector recordings were analysed in BioSonic (Biosonic AB), a browser-based platform that filters noise files and suggests a species, with a confidence score, for each bat recording. Every recording identified as Myotis, Plecotus, Barbastella, Nyctalus or Eptesicus, every recording below [threshold] confidence, and at least one recording of each species on each survey date were checked on the spectrogram by [name, role], in line with BCT guidance (Collins, 2023). In total [X]% of bat recordings were checked manually and [Y] automatic identifications were corrected. Myotis recordings that could not be confirmed to species are reported as Myotis sp. All confirmations and corrections are logged by reviewer and date.

Then add the evidence for the classifier itself. BioSonic scored 98.7% F1 in a third-party benchmark by Wilder Sensing on 5,325 predictions, and the full confusion matrices are on our benchmarks page.6 A benchmark does not replace your own vetting, but it helps a reviewer judge how far to trust the part you did not vet. List the limits plainly too: Myotis to species, zero-crossing data and very noisy sites.

Options compared

Which auto ID tool fits this job?

Five tools come up most for UK static data. Each is a serious option, and the right one depends on your data volume, team size and whether your recordings can be shared. Facts about other tools come from their makers' own pages, as cited in our Kaleidoscope Pro alternatives guide.

ToolHow it worksFits best whenWatch out for
Kaleidoscope Pro (Wildlife Acoustics)Desktop software for Windows, macOS and Linux with Auto-ID classifiers for the UK, Europe, North America and more. Single-user annual licence with two activations.11You work across continents, need to process offline, or also analyse birds and frogs.One best-match species per recording in the headline output. In the Wilder Sensing benchmark it scored 83.0% F1, with 174 false positives against BioSonic's 13.6 Wildlife Acoustics itself says to always verify a suggested ID manually.
BTO Acoustic Pipeline (British Trust for Ornithology)Desktop app with cloud processing. Classifiers for 69 European and North African bat species, plus small mammals and bush-crickets. BTO reports a 93 to 100% true positive rate for European bats from its own checks.12Cost is the main concern, results can be shared with BTO, or you also need other taxa.Free within a yearly credit allowance only when results are shared, more credits per GB when not.13 Wilder Sensing could not validate results on the platform and found it splits files longer than 5 seconds into 5-second segments.16
SonoBat UKDesktop full-spectrum call analysis with a classifier set for UK species. Windows (Intel) and macOS.14One experienced analyst wants conservative, full-spectrum vetting on their own machine.UK: annual subscription per person. We found no classifier for mainland European species.
BatExplorer (Elekon)Windows software that proposes species and genus from averaged call parameters against European and UK libraries. Lite and Standard are free; Professional is an annual licence.15You own Batloggers and want free species proposals plus manual analysis.Designed around Batlogger files. Parameter matching needs clean, well-separated calls, and there is no large public benchmark.
BioSonic (Biosonic AB)Browser platform. Deep learning on spectrograms with noise filtering, confidence scores, behaviour flags and shared projects with logged review.A team handles many detector-nights of UK or European data and wants auto ID, vetting and reporting in one place.Needs an internet connection, covers European species only, and zero-crossing accuracy has not been benchmarked separately. Myotis and Plecotus still need a human, as with every tool.

Many teams combine tools: BioSonic for bulk sorting and vetting, and Anabat Insight, BatExplorer or BatSound for measuring a difficult call by hand. For more detail, see the one-to-one comparisons with Kaleidoscope Pro, the BTO Acoustic Pipeline, SonoBat UK and BatExplorer.

Worked example

What 200 detector-nights cost in analyst time.

This example uses only figures published on this site: the rates in our time calculator, taken from Wharton's customer data (0.245 analyst hours per GB for a traditional Kaleidoscope-class workflow and 0.0815 hours per GB with BioSonic), and the storage calculator's estimate of 1.5 GB a night for a triggered full-spectrum Song Meter SM4BAT.8,9

StepCalculationResult
Detector-nights20 statics for 10 nights each200
Data volume200 nights at 1.5 GB eachabout 300 GB
Traditional workflow300 GB at 0.245 hours per GBabout 74 hours (2.1 weeks at 35 hours)
With BioSonic300 GB at 0.0815 hours per GBabout 24 hours (0.7 weeks)
Analyst time saved74 minus 24, roundedabout 49 hours

The rates match the single night Wharton tested side by side: 10.3 GB took about 2 hours 30 minutes in Kaleidoscope Pro and about 50 minutes in BioSonic, a 66% saving, most of it from automatic noise filtering.7 Your own figures will differ with site noise, detector settings and how much you vet, so put your volumes into the time calculator or upload one night you have already analysed and compare.

FAQ

What survey teams ask about auto ID.

What auto ID tool would you recommend for UK static bat detector data?

For a team with a lot of static data, pick a tool that filters noise files, gives every identification a confidence score and records who checked what. BioSonic does all three in the browser and scored 98.7% F1 in a third-party benchmark by Wilder Sensing.6 The BTO Acoustic Pipeline is the low-cost choice if your results can be shared, Kaleidoscope Pro suits offline or multi-continent work, SonoBat UK suits one experienced analyst on one machine, and BatExplorer suits Batlogger owners. We make BioSonic, so test any shortlist on your own labelled recordings.

Does auto ID replace manual vetting?

No. Bat Conservation Trust guidance says automated classifications should be manually verified, ideally at least one recording of each species on each survey date and especially notable records, and some species, such as whiskered bat, cannot generally be verified from a recording at all.1 Good auto ID changes how much you check, not whether you check: noise files are removed and a confidence order sends your time to the hard calls.

How much of the auto ID output should we vet?

The guidance does not set one percentage, so set clear rules and report the result. A defensible set of rules is to vet every Myotis, Plecotus, Barbastella, Nyctalus and Eptesicus pass, every low-confidence identification and every notable record, plus at least one recording of each species on each survey date. Then state the percentage of recordings vetted and the number of corrections in your report.

How do I identify bat species from AudioMoth recordings?

Copy the WAV files from the SD card as recorded, organised by site and night, and upload them to an auto ID tool. In BioSonic, noise files are filtered out, each bat recording gets a species suggestion with a confidence score, and you vet the uncertain and look-alike calls on the spectrogram before exporting to CSV or Excel. No conversion or file splitting is needed. If you record continuously, expect far more data and noise: about 17 to 28 GB a night, against about 1.8 GB for a triggered AudioMoth.9

How do we justify auto ID to a planning authority or in an Environmental Statement?

Write the method down and back it with numbers. The method statement should name the software, explain what was automated and what your vetting rules were, and give the percentage of recordings checked by hand and the number of corrections. Keep the audit trail of who confirmed or corrected each record. The BCT Good Practice Guidelines accept departures from their methods when the ecological rationale is clear and supported with evidence, so cite the classifier's published benchmark alongside your own vetting results.2

Which detectors and file formats does BioSonic accept?

WAV and FLAC from Song Meter SM4BAT and Mini Bat, AudioMoth, Anabat, Batlogger, Pettersson and other full-spectrum detectors, uploaded as recorded. Anabat zero-crossing files (.zc and .NN# sequence files) are decoded on upload, but the published benchmark was measured on full-spectrum recordings, so test zero-crossing data on your own labelled files. For batcorder .raw files, ask us first. See detector compatibility for what to upload from each detector.

Is BioSonic more accurate than Kaleidoscope Pro or the BTO Acoustic Pipeline?

Against Kaleidoscope Pro, yes, in the third-party head-to-head we know of: in Wilder Sensing's benchmark BioSonic scored 98.7% F1 to Kaleidoscope Pro's 83.0%, with 13 false positives against 174.6 The BTO Acoustic Pipeline could not be benchmarked the same way, because its platform does not support manual validation of results; BTO reports a 93 to 100% true positive rate for European bats from its own checks, which is a different metric.12,16 Myotis and Plecotus stay hard for every classifier.

How is BioSonic priced?

You buy a package of analysis gigabytes. The bigger the package, the lower the price per GB. All features and all users are included, so the whole team, from field staff to the ecologist who signs off, works in one organisation account, and your recordings are kept for 9 years. Ask us for a quote, or start with a free trial on your own recordings.

Sources

Every claim, referenced.

Figures about BioSonic come from pages on this site. Guidance and other tools are cited to their own publishers.

  1. Bat Conservation Trust, Guidelines for passive acoustic surveys of bats in woodland: automated classifications should be manually verified ("ideally at least one recording of each species on each survey date"), especially notable records; for some species, such as Myotis mystacinus, an audio recording generally cannot be verified conclusively. Accessed 2026-11-17.
  2. Collins, J. (ed.) (2023) Bat Surveys for Professional Ecologists: Good Practice Guidelines (4th edn). The Bat Conservation Trust, London. The publisher's page lists new sections including bat sound analysis auto-identification, and says departures from the methods are acceptable when the ecological rationale is clear and supported with evidence. Accessed 2026-11-17.
  3. Bat Conservation Trust, UK bats: 18 species of bat in the UK, 17 of them known to breed here.
  4. Rydell, J. et al. (2017) Testing the performance of automated identification of bat echolocation calls. Ecological Indicators 78, 416 to 420. Automated classifiers identified Plecotus auritus between 0 and 89% and Myotis species between 4 and 93% of the time.
  5. Solick, D.I. et al. (2024) Automated echolocation classifiers vary in accuracy for northeastern U.S. bat species. PLOS ONE 19(6): e0300664. None of three classifiers tested, Kaleidoscope Pro among them, reliably identified calls to individual Myotis species.
  6. Third-party benchmark by Wilder Sensing, 5,325 predictions for BioSonic. BioSonic 98.7% F1 against Kaleidoscope Pro 83.0% F1 (13 against 174 false positives). Full method on our benchmarks page.
  7. Wharton case study: one survey night of 4,200 files (10.3 GB), around 3,000 of them noise, took about 2 hours 30 minutes in Kaleidoscope Pro and about 50 minutes in BioSonic (66% less time); 822,729 files, about 2 TB, in the 2025 season.
  8. BioSonic time calculator: 0.245 analyst hours per GB for a traditional workflow and 0.0815 with BioSonic, from Wharton's 2 TB season (490 against 163 hours).
  9. BioSonic storage calculator: mid-range GB-per-night estimates by detector, and AudioMoth continuous recording at about 18 GB (250 kHz) and 28 GB (384 kHz) over a ten-hour night.
  10. BioSonic Bat School: look-alike species for noctule, Leisler's bat, serotine and barbastelle.
  11. Wildlife Acoustics, Automatic bat identification (classifier regions; always manually verify a suggested ID) and Kaleidoscope Pro (two activations for a single user for 12 months).
  12. British Trust for Ornithology, Ultrasonic classifiers: 69 bat species; 93 to 100% true positive rate for European bats.
  13. British Trust for Ornithology, Acoustic Pipeline pricing: yearly free credits when results are shared; more credits per GB when they are not.
  14. SonoBat, Purchase and licensing (supported systems) and SonoBat UK.
  15. Elekon, BatExplorer (editions and species libraries).
  16. BioSonic vs BTO Acoustic Pipeline: Wilder Sensing's findings that results could not be validated manually on the BTO platform and that files longer than 5 seconds were split into 5-second segments.

Last verified: 2026-11-17.

Try it on one night you have already analysed.

Upload a night of static recordings and compare the result, and the time it took, with your current workflow. Free, in the browser, no install.

support@biosonic.se · +46 72 744 65 85