From buzzes to bytes: A systematic review of automated bioacoustics models used to detect, classify, and monitor insects
Cite this dataset
Kohlberg, Anna; Myers, Christopher; Figueroa, Laura (2024). From buzzes to bytes: A systematic review of automated bioacoustics models used to detect, classify, and monitor insects [Dataset]. Dryad. https://doi.org/10.5061/dryad.hmgqnk9r1
Abstract
Insects play vital ecological roles; many provide essential ecosystem services while others are economically devastating pests and disease vectors. Concerns over insect population declines and expansion have generated a pressing need to effectively monitor insects across broad spatial and temporal scales. A promising new approach is bioacoustics, which uses sound to study ecological communities. Despite recent increases in machine learning technologies, the status of emerging automated bioacoustics methods for monitoring insects is not well known, limiting potential applications.
To address this gap, we systematically review the effectiveness of automated bioacoustics models over the past four decades, analyzing 176 studies that met our inclusion criteria. We describe their strengths and limitations compared to traditional methods and propose productive avenues forward.
We found automated bioacoustics models for 302 insect species distributed across nine Orders. Studies used intentional calls (e.g., grasshopper stridulation), by-products of flight (e.g., bee wingbeats), and indirectly produced sounds (e.g., grain movement) for identification. Pests were the most common study focus, driven largely by weevils and borers moving in dried food and wood. All disease vector studies focused on mosquitoes. A quarter of the studies compared multiple insect families.
Our review illustrates that machine learning, and deep learning in particular, are becoming the gold standard for bioacoustics automated modeling approaches. We identified models that could classify hundreds of insect species with over 90% accuracy. Bioacoustics models can be useful for reducing lethal sampling, monitoring phenological patterns within and across days, and working in locations or conditions where traditional methods are less effective (e.g., shady, shrubby, or remote areas). However, it is important to note that not all insect taxa emit easily detectable sounds, and that sound pollution may impede effective recordings in some environmental contexts.
Synthesis and applications: Automated bioacoustics methods can be a useful tool for monitoring insects and addressing pressing ecological and societal questions. Successful applications include assessing insect biodiversity, distribution, and behavior, as well as evaluating the effectiveness of restoration and pest control efforts. We recommend collaborations among ecologists and machine learning experts to increase model use by researchers and practitioners.
README: From buzzes to bytes: A systematic review of automated bioacoustics models used to detect, classify, and monitor insects
https://doi.org/10.5061/dryad.hmgqnk9r1
The following files correspond to the results from a systematic review of automated bioacoustics models used to detect, classify, and monitor insects.
Description of the data and file structure
This submission contains five CSV files required for analyses and one R script:
1) data_extraction_metadata, which provides metadata information for each study evaluated, such as title, author, journal, and language.
Details:
duplication_flagged (whether this was likely a duplicate article already present in the dataset),
article_ID (individual article),
inclusion (included yes/no data from this article),
exclusion_criteria (if excluded, what was the reason why),
exclusion_insect_clumped (if excluded because the aggregate of insects were clumped into a single category that was not relevant for the article),
concern_type (ACI_clump, ambiguous_results, data_reporting_descrepancy, methods_lack_detail, NA, no_error_reported, other, small_sample_size, some_data_unreported, unclear_automation, unclear_model_usage),
concerns_notes (additional notes regarding potential concerns about article),
notes (other types of notes),
title (article title),
author (article authors),
journal (journal where article published),
volume (volume in which article published),
issue (issue in which article published),
pages (pages in which article is published),
publish_year (year in which article published),
article_language (language in which article written),
tropics (whether article was conducted in the tropics),
article_country (country in which article conducted/published),
multiple_locations (whether article included multiple locations),
DOI (DOI associated with article),
full_text_link (link to full article).
2) data_extraction-methods, which provides information about the different automated bioacoustics models used in each study as well as other relevant results from the study level, including the resolution obtained from the model (e.g. species), the number of taxa classified, and the accuracy estimates obtained with each model.
Details:
study_ID (individual article)
resolution (taxonomic resolution used in this study, e.g. species, family, order, class)
number_taxa_classified (number of taxa that the model classified)
date_recording_data_collection (date of data collection recording)
ecosystem (ecosystem in which the data collection took place)
monitoring_type (the focus of the type of monitoring; beneficial_other, hive_monitoring, NA, parasite_monitoring, pest_monitoring_food, pest_monitoring_other, pest_monitoring_wood)
substrate_specific (the substrate on which the recording took place, if relevant; e.g. food, wood, etc.)
uses_raven (whether article uses the Raven software, from Cornell Lab of Ornithology)
conservation_mentioned (whether conservation was mentioned in the article)
community_science (whether article employed community/citizen science)
using_existing_database (whether article used existing audio database)
using_mulitple_databases (whether article used multiple audio database)
database_name (name of audio database, if relevant)
open_source_database (whether article used an open source audio database)
using_existing_code (whether article used existing code for the development of their automated models)
open_source_code (whether article provides open source code for the development of their automated models)
create_user_interface (whether article creates a user interface)
employ_user_interface (whether article uses a user interface)
data_reporting (the general state of the data reporting, namely ambiguous_partial_or_nodata, application or unambiguous_fulldata).
train_test_ratio_clean (ratio of training to test data in the article; cleaned for analysis)
train_test_ratio_raw (raw ratio of training to test data in the article if provided)
train_test_rounded (ratio of training to test data in the article; cleaned for analysis and rounded to nearest multiple of 5)
paper_type (type of article, namely whether 1) applied, i.e. testing questions using bioacoustics; 2) methods; i.e. development of bioacoustics methods but not applied; or 3) methods_applied, developed bioacoustics methods and applied them)
automation_type (automation type, namely DL_ML, ML, SSO, or unknown; DL = deep learning; ML = machine learning; SSO = Signals, statistics, and other).
one_taxa_model (whether the model only included one taxa)
one_taxa_name (what the taxa was if only one taxa evaluated)
ecoacoustics_model (whether an ecoacoustics model)
detection_only_model (whether a detection only model)
model_role (model role, namely behavior, classification, detection or NA)
feature_extraction (feature extraction method)
features_used (feature used)
ML_classification (type of ML classification used, broad category)
ML_spec (type of ML classification used, specific category)
covariate (covariate if present)
accuracy_percent (model accuracy result)
accuracy_error_type (model accuracy error type)
accuracy_error_value (value of accuracy value)
precision_percent (model precision result)
precision_error_type (model precision error type)
precision_error_value (precision error value)
recall_percent (model recall result)
recall_error_type (model recall error type)
recall_error_value (recall error value)
f1 (f1 = harmonic mean of recall and precision value)
f1_error_type (f1 error type)
f1_error_value (f1 error value)
other_metric (if any other model performance estimate was used)
other_metric_score (score of additional model)
other_metric2 (if a second additional model performance estimate was used)
other_metric2_score (score of second additional model)
3) data_extraction-taxa, which provides information on taxa-specific model outputs (for studies that specified output for specific taxa).
Details:
study_ID (individual article)
focal_class (taxonomic class of organism evaluated in model)
focal_order (taxonomic order of organism evaluated in model)
focal_superfamily (taxonomic superfamily of organism evaluated in model)
focal_family (taxonomic family of organism evaluated in model)
focal_genus (taxonomic genus of organism evaluated in model)
focal_species (taxonomic species of organism evaluated in model)
category_common_name (common name of organism evaluated in model)
substrate_specific (substrate on which data collected)
data_reporting (the general state of the data reporting, namely ambiguous_partial_or_nodata, application or unambiguous_fulldata).
automation_type (automation type, namely DL_ML, ML, SSO, or unknown; DL = deep learning; ML = machine learning; SSO = Signals, statistics, and other).
ecoacoustics_model (whether an ecoacoustics model)
detection_only_model (whether a detection only model)
model_role (model role, namely behavior, classification, detection or NA)
feature_extraction (feature extraction method)
features_used (feature used)
ML_classification (type of ML classification used, broad category)
ML_spec (type of ML classification used, specific category)
covariate (covariate if present)
accuracy_percent (model accuracy result)
accuracy_error_type (model accuracy error type)
accuracy_error_value (value of accuracy value)
precision_percent (model precision result)
precision_error_type (model precision error type)
precision_error_value (precision error value)
recall_percent (model recall result)
recall_error_type (model recall error type)
recall_error_value (recall error value)
f1 (f1 = harmonic mean of recall and precision value)
f1_error_type (f1 error type)
f1_error_value (f1 error value)
other_metric (if any other model performance estimate was used)
other_metric_score (score of additional model)
other_metric2 (if a second additional model performance estimate was used)
other_metric2_score (score of second additional model)
4) data_extraction-all-ecosystems, which provides information on the ecosystems studied for each of the studies.
Details:
study ID (individual article).
ecosystem (ecosystem in which the data collection took place).
5) data_extraction-key, which provides additional details on some of the terms used in the other .csv files.
6) Kohlberg_et_al_JAE_Final.R, R file with code for all analyses and figures presented in the main and supplementary material of this article. The R file has been annotated to provide additional details and explain logic of all decisions made. The R file calls all except the data_extraction-key .csv file.
Funding
National Science Foundation, Award: 2010615, Division of Biological Infrastructure (DBI)