Back to Search
Start Over
Stronger together: Combining automated classifiers with manual post-validation optimizes the workload vs reliability trade-off of species identification in bat acoustic surveys
- Source :
- Ecological Informatics. 49:45-53
- Publication Year :
- 2019
- Publisher :
- Elsevier BV, 2019.
-
Abstract
- Owing to major technological advances, bioacoustics has become a burgeoning field in\ud ecological research worldwide. Autonomous passive acoustic recorders are becoming widely\ud used to monitor aerial insectivorous bats, and automatic classifiers have emerged to aid\ud researchers in the daunting task of analyzing the resulting massive acoustic datasets.\ud However, the scarcity of comprehensive reference call libraries still hampers their wider\ud application in highly diverse tropical assemblages. Capitalizing on a unique acoustic dataset\ud of more than 650,000 bat call sequences collected over a 3-year period in the Brazilian\ud Amazon, the aims of this study were (a) to assess how pre-identified recordings of free-flying\ud and hand-released bats could be used to train an automatic classification algorithm (random\ud forest), and (b) to optimize acoustic analysis protocols by combining automatic classification\ud with visual post-validation, whereby we evaluated the proportion of sound files to be postvalidated\ud for different thresholds of classification accuracy. Classifiers were trained at species\ud or sonotype (group of species with similar calls) level. Random forest models confirmed the\ud reliability of using calls of both free-flying and hand-released bats to train custom-built\ud automatic classifiers. To achieve a general classification accuracy of ~85%, random forest\ud had to be trained with at least 500 pulses per species/sonotype. For seven out of 20 sonotypes,\ud the most abundant in our dataset, we obtained high classification accuracy (>90%). Adopting\ud a desired accuracy probability threshold of 95% for the random forest classifier, we found that\ud the percentage of sound files required for manual post-validation could be reduced by up to\ud 75%, a significant saving in terms of workload. Combining automatic classification with\ud manual ID through fully customizable classifiers implemented in open-source software as\ud demonstrated here shows great potential to help overcome the acknowledged risks and biases\ud associated with the sole reliance on automatic classification.
- Subjects :
- 0106 biological sciences
Bioacoustics
Computer science
Trade-off
Machine learning
computer.software_genre
010603 evolutionary biology
01 natural sciences
Field (computer science)
Task (project management)
Software
46 Information and Computing Sciences
Ecology, Evolution, Behavior and Systematics
Reliability (statistics)
Ecology
business.industry
010604 marine biology & hydrobiology
Applied Mathematics
Ecological Modeling
Workload
Computer Science Applications
Random forest
3109 Zoology
Computational Theory and Mathematics
Modeling and Simulation
Artificial intelligence
business
computer
31 Biological Sciences
Subjects
Details
- ISSN :
- 15749541
- Volume :
- 49
- Database :
- OpenAIRE
- Journal :
- Ecological Informatics
- Accession number :
- edsair.doi.dedup.....622e9e7e79fcb088f3ecb672ae6c43a7