Stop checking every file and check by risk: let software remove noise, test a random sample of each common species to set a confidence threshold, and check in full the rare, unexpected and oddly timed records. This follows the four steps for auto ID in the BCT guidelines. Software such as BioSonic does the noise removal and lets a team review by species, confidence or random sample. In one published Wharton case study it cut the time for a night of 4,200 files by 66%.
We make BioSonic. The method on this page works with any bat analysis software.
What to check, and how much
| Records | How much to check | Why |
|---|---|---|
| Common species above your threshold | A random sample for each species | BCT uses a random sample to measure the false positive rate and set the threshold. |
| Records below the threshold | Discard them, or check those that matter | BCT describes discarding records below an error threshold. |
| Locally uncommon or rarely detected species | All of them | BCT lists these for manual verification. One error changes the result. |
| Records at an unusual time of night or year | All of them | BCT lists these for manual verification. |
| Each species on each survey date | Ideally at least one | BCT gives this in its woodland acoustic guidelines. |
| Noise files | A random sample for each detector and period | Shows how many bats the filter missed. |
Five ways to save the hours
- Analyse first and look later, so nobody scrolls through wind, rain and insects.
- Set the threshold for each species and project. One score for everything wastes checks.
- Review one species at a time. Similar calls in a row are quicker to judge.
- Split the detectors between reviewers and give the difficult groups to your most experienced analyst.
- Keep the numbers as you go: files checked, files changed and who checked them. The method section then writes itself.
What it saved in one case
Wharton timed one night of 4,200 files, of which about 3,000 were noise. It took about 2 hours 30 minutes in Kaleidoscope Pro and about 50 minutes in BioSonic. That is one team and one night, so time one of your own. See the Wharton case study.
Common question
Do I have to check every bat recording by hand?
BCT does not ask for that when auto ID is validated. It says to verify a proportion of the data and to check by hand the situations where errors are more likely or matter more, such as locally uncommon species.
Sources
- Bat Surveys for Professional Ecologists, 4th edition (Collins, ed., Bat Conservation Trust, 2023, amended March 2024), sections 10.2.9 and 10.2.17 to 10.2.19.
- Guidelines for passive acoustic surveys of bats in woodland (Bat Conservation Trust).
- Wharton case study, published by BioSonic.
Related: Auto ID for static detector data · How much of the noise folder to check · Time calculator
