Detection error varies across space, time, and species in passive acoustic monitoring of co-occurring invasive and at-risk anurans

Passive acoustic monitoring (PAM) is increasingly used to detect species of conservation and management concern. However, optimizing detection methods to meet management needs remains a challenge. Spatiotemporal variation in detection error is rarely quantified, yet error has profound consequences for management actions informed by PAM. We examined variation in automated acoustic detection error rates across sites and day of year using a daily-resolution PAM dataset of the native Arizona treefrog (Hyla wrightorum) and the invasive American bullfrog (Lithobates catesbeianus) from eleven ponds in 2021 in southeastern Arizona, USA. We compared a manually verified daily presence-absence dataset with automated classifier output to calculate Type I and II error rates, precision, and recall. Adjusting classifier sensitivity revealed species-specific thresholds to minimize total error; Arizona treefrogs required higher sensitivity than American bullfrogs to achieve the lowest total error rates. After adjustment, the lowest total error rates were 21.2% for Arizona treefrogs and 22.5% for American bullfrogs. For both species, error varied 4- to 10-fold among sites, and none of the site characteristics we measured explained this variation. However, American bullfrog detection error followed a U-shaped seasonal pattern, with higher error at the beginning and end of the monsoon season. Our detection error rates fell within the range reported across taxa but were slightly higher than those of other anuran studies, likely reflecting the larger spatiotemporal extent of our study. No single set of classifier parameters minimized error for both species, so optimal detection settings depend on management goals. Invasive-species managers, for example, may prioritize minimizing missed detections, whereas managers of at-risk species must weigh missed occupied sites against false detections that divert limited resources. Manually verifying even a subset of recordings spanning sites and seasons allows these tradeoffs to be quantified and the workflow tailored to species-specific error tolerances.

Authors

Institutions

Publication Details

Journal
Authorea
Published
2026-11-19
DOI
https://doi.org/10.22541/authorea.15012464/v1
Primary Topic
Amphibian and Reptile Biology
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Detection error varies across space, time, and species in passive acoustic monitoring of co-occurring invasive and at-risk anurans

Meryl C. Mims, Joseph C. Drake, John D. Kraft, Grace O’Malley et al.
Authorea
Amphibian and Reptile Biology
preprint

Detection error varies across space, time, and species in passive acoustic monitoring of co-occurring invasive and at-risk anurans

Meryl C. Mims, Joseph C. Drake, John D. Kraft, Grace O’Malley, Charlotte Tury
preprint en

Abstract

Passive acoustic monitoring (PAM) is increasingly used to detect species of conservation and management concern. However, optimizing detection methods to meet management needs remains a challenge. Spatiotemporal variation in detection error is rarely quantified, yet error has profound consequences for management actions informed by PAM. We examined variation in automated acoustic detection error rates across sites and day of year using a daily-resolution PAM dataset of the native Arizona treefrog (Hyla wrightorum) and the invasive American bullfrog (Lithobates catesbeianus) from eleven ponds in 2021 in southeastern Arizona, USA. We compared a manually verified daily presence-absence dataset with automated classifier output to calculate Type I and II error rates, precision, and recall. Adjusting classifier sensitivity revealed species-specific thresholds to minimize total error; Arizona treefrogs required higher sensitivity than American bullfrogs to achieve the lowest total error rates. After adjustment, the lowest total error rates were 21.2% for Arizona treefrogs and 22.5% for American bullfrogs. For both species, error varied 4- to 10-fold among sites, and none of the site characteristics we measured explained this variation. However, American bullfrog detection error followed a U-shaped seasonal pattern, with higher error at the beginning and end of the monsoon season. Our detection error rates fell within the range reported across taxa but were slightly higher than those of other anuran studies, likely reflecting the larger spatiotemporal extent of our study. No single set of classifier parameters minimized error for both species, so optimal detection settings depend on management goals. Invasive-species managers, for example, may prioritize minimizing missed detections, whereas managers of at-risk species must weigh missed occupied sites against false detections that divert limited resources. Manually verifying even a subset of recordings spanning sites and seasons allows these tradeoffs to be quantified and the workflow tailored to species-specific error tolerances.

Authorea
US Forest Service (US), University of Massachusetts Amherst (US), Virginia Tech (US)
Amphibian and Reptile Biology
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon