Search USGS⌕ Search
Home/Archive/USGS

USGS · 70196566

Evaluating autonomous acoustic surveying techniques for rails in tidal marshes

Lydia L. Stiffler·James T. Anderson·Todd E. Katzner
View original source ↗Export metadata ↓

Abstract

There is a growing interest toward the use of autonomous recording units (ARUs) for acoustic surveying of secretive marsh bird populations. However, there is little information on how ARUs compare to human surveyors or how best to use ARU data that can be collected continuously throughout the day. We used ARUs to conduct 2 acoustic surveys for king ( Rallus elegans ) and clapper rails ( R. crepitans ) within a tidal marsh complex along the Pamunkey River, Virginia, USA, during May–July 2015. To determine the effectiveness of an ARU in replacing human personnel, we compared results of callback point‐count surveys with concurrent acoustic recordings and calculated estimates of detection probability for both rail species combined. The success of ARUs at detecting rails that human observers recorded decreased with distance ( P ≤ 0.001), such that at <25 m, 90.3% of human‐recorded rails also were detected by the ARU, but at >75 m, only 34.0% of human‐detected rails were detected by the ARU. To determine a subsampling scheme for continuous ARU data that allows for effective surveying of presence and call rates of rails, we used ARUs to conduct 15 continuous 48‐hr passive surveys, generating 720 hr of recordings. We established 5 subsampling periods of 5, 10, 15, 30, and 45 min to evaluate ARU‐based presence and vocalization detections of rails compared with each of the full 60‐min sampling of ARU‐based detection of rails. All subsampling periods resulted in different ( P ≤ 0.001) detection rates and unstandardized vocalization rates compared with the hourly sampling period. However, standardized vocalization counts from the 30‐min subsampling period were not different from vocalization counts of the full hourly sampling period. When surveying rail species in estuarine environments, species‐, habitat‐, and ARU‐specific limitations to ARU sampling should be considered when making inferences about abundances and distributions from ARU data.

Explore related subjects

Wildlife Society BulletinArticle

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Lydia L. Stiffler, James T. Anderson, Todd E. Katzner. 2018-03-08. Evaluating autonomous acoustic surveying techniques for rails in tidal marshes. https://doi.org/10.1002/wsb.860

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related USGS reports

USGS2026

Developing a low-cost drone-based method for deploying and retrieving autonomous recording units in inaccessible areas

Autonomous Recording Units (ARUs) are increasingly used in research to support passive acoustic monitoring, but deployment in remote or inaccessible locations remains logistically challenging. Traditional manual placement is labor-intensive, time-consuming, potentially hazardous, and often limited to habitat edges, creating sampling biases and restricting spatial coverage. To address placement limitations in wetland systems, we developed and field-tested a low-cost, lightweight floating platform paired with a drone-based deployment and retrieval system. The drone-deployable ARU platform, constructed from inexpensive, off-the-shelf materials (~US $21 per unit), weighed ~560 g and was mounted with an ARU (AudioMoth) at ~1.2 m above water. The platform was designed for stability, portability, and compliance with Federal Aviation Administration Part 107 regulations. Field trials were conducted across 50 operations in various types of historical rice-field impoundments along the South Carolina coast, which provide critical habitat for secretive marsh birds such as rails (e.g ., Rallus, Laterallus , and Porzana ) and bitterns (e.g ., Botaurus and Ixobrychus ). All 50 deployments and retrievals were successful, demonstrating the robustness of the method under diverse hydrologic and vegetative conditions. Deployment times were significantly shorter than retrieval times (median 6 vs. 9 min), with retrieval requiring greater precision for hook engagement. Operation times scaled predictably with distance: deployment increased by 0.93 min/100 m (R 2 = 0.94) and retrieval by 1.17 min/100 m (R 2 = 0.95). No platform damage or displacement was observed, as emergent vegetation was likely providing natural anchoring. Our approach offers an affordable and effective solution for expanding ARU coverage in large, inaccessible wetlands, reducing sampling bias and enhancing biodiversity monitoring. With continued advances in drone payload capacity, battery endurance, and beyond-visual-line-of-sight flight regulations, our workflow could be adapted for deploying multi-sensor platforms (e.g., ARUs, eDNA samplers, water and air quality sensors, plant sample collectors, insect traps, and trail cameras) to support integrated biodiversity monitoring in challenging landscapes worldwide.

South Carollina↗
USGS2026

Comparison of data handling techniques for modeling bat acoustic activity

With the proliferation of acoustic sampling to investigate bat distribution and ecology, researchers have implemented a myriad of statistical modeling approaches to interpret findings. Bats are taxa of high conservation concern; therefore, ensuring the accuracy of species-level habitat association models is critical for informing management. We sought to determine prediction differences among statistical approaches to modeling counts of acoustic detections, using generalized linear mixed models with 8 acoustic data-handling techniques. We applied each approach or combination of approaches to a rare species, the northern long-eared bat ( Myotis septentrionalis ), and a common species, the eastern red bat ( Lasiurus borealis ), from summer survey results on a landscape in south-central Pennsylvania, USA, 2024. We evaluated the accuracy of habitat association models of bat acoustic activity at the species level using cross-validation and compared resulting predictions of models using spatial correlations. We determined that filtering data by automated identification software (Kaleidoscope Pro), the maximum likelihood estimate (MLE) P -value thresholds reduced relative mean absolute error (rMAE) in cross-validation of northern long-eared bat models. Using the MLE-retained data produced the most accurate predictions over using raw data or the overly conservative match ratio data. However, for the eastern red bat, the results from the most conservative approach of only retaining data with at least a 90% match ratio from software development training sets had the lowest rMAE. We have provided evidence that the current standard of filtering data by nightly MLE can result in more accurate and informative habitat-use acoustic activity models for rare bat species.

Pennsylvania↗
USGS2026

Abundance, trends, and challenges facing mountain goats throughout their North American distribution

Recent declines among some mountain goat ( Oreamnos americanus ) populations have heightened concern about their current status and ability to cope with future challenges. We conducted a range-wide assessment of the status of mountain goats across their distribution to understand the extent and patterns of change in recent years. We queried states, provinces, National Parks, and Indigenous governments with territories and reserves containing local wild populations, requesting updates on estimated mountain goat abundance and trends over time, as well as qualitative assessments of conservation challenges. We supplemented the questionnaires with existing literature (both published and available from agency websites). We subjected a subset of mountain goat population time−series to exploratory meta-analyses, examining covariates associated with rates of change. Respondents identified 203 units within which estimates were documented, a majority of which were raw counts from aircraft; fewer jurisdictions used site-specific sightability models, mark-recapture statistics, or other approaches to account for imperfect detection. Jurisdictions updated unit-specific estimates on average every 4.0 years (SD = 2.9, Min–Max = 1–24 years). Among 152 qualitative estimates of trends of native populations, 3% were reported as having increased substantially, 10% as having increased slightly, 24% as having declined slightly, and 14% as having declined substantially (49% reported either no detectable trend or insufficient data to identify a trend). Among 133 estimates of trends of introduced or pioneering populations, 5% increased substantially, 24% increased slightly, 20% declined slightly, and 11% declined substantially (40% reported no detectable trend). Jurisdiction-wide abundance estimates were lower than those previously published in all jurisdictions except Colorado and Oregon, and in Alaska, where trends were unclear. Intrinsic rates of increase ( r ) were positively associated with the population being introduced or pioneering, negatively associated with heavy snow and, to a lesser extent, with drought, but were not associated with abundance or presence of hunter harvest. Existing evidence suggested that many local populations, especially native populations, have declined, but understanding of specific causes has been constrained by limited monitoring capacity. Greater spatiotemporal monitoring effort and detailed studies would inform appropriate strategies to address threats and future challenges to mountain goats.

Wildlife Society Bulletin↗