SIAlex: Species identification and monitoring based on bird sound features

Ecological Informatics, Год журнала: 2024, Номер 81, С. 102637 - 102637

Опубликована: Май 13, 2024

The combination of deep learning and bird sound recognition is widely employed in species conservation monitoring. A complex network structure not conducive for deploying devices, resulting problems such as long inference time low efficiency. Using AlexNet the backbone model, we explore potential shallow straightforward models without connection techniques or attention mechanisms, named SIAlex, to recognise classify 20 datasets, which are simultaneously validated on a 10 class UrbanSound8k dataset. structural re-parameterization method, number model layers reduced, computational efficiency improved, significantly achieving decoupling training structure. To increase nonlinearity cascaded approach utilised activation functions, thereby improving generalisation performance model. Simultaneously, classifier section, convolutional layer replaces original fully connected layer, reducing increasing feature extraction ability accuracy, effectively recognising speech. experimental data show that SIAlex Birdsdata dataset improves accuracy 93.66%, piece only 2.466 ms. reaches 96.04%, 3.031 large comparisons have shown method proposed this paper achieves good results bringing breakthroughs application shallow, simple models.

Язык: Английский

---

A citizen science platform to sample beehive sounds for monitoring ANSP

Journal of Environmental Management, Год журнала: 2025, Номер 375, С. 124247 - 124247

Опубликована: Янв. 25, 2025

Язык: Английский

---

The potential of soundscapes as an ecosystem monitoring tool for urban biodiversity

Journal of Urban Ecology, Год журнала: 2025, Номер 11(1)

Опубликована: Янв. 1, 2025

Abstract As urbanization and densification often lead to significant biodiversity loss, understanding monitoring urban patterns is crucial. Traditional methods are costly, time-consuming, require specialized expertise. Passive acoustic soundscape ecology have emerged as promising, non-invasive techniques for ecosystem monitoring. This review aims provide an overview of approaches utilized in discuss their limitations. We highlight exemplary studies that focus on demonstrate recordings can be partially used predict cities, especially avian species. To realize the potential conservation, current challenges must addressed. includes data processing, security, missing standardized collection methods. call further research combines innovative technologies transdisciplinary develop effective conservation applications cities.

Язык: Английский

---

Continental-scale behavioral response of birds to a total solar eclipse

Scientific Reports, Год журнала: 2025, Номер 15(1)

Опубликована: Апрель 8, 2025

Язык: Английский

---

Landfill as a Food Source for the Herring Gull – What Can We Find in Pellets?

Опубликована: Янв. 1, 2025

Язык: Английский

---

Using photographs and deep neural networks to understand flowering phenology and diversity in mountain meadows

Remote Sensing in Ecology and Conservation, Год журнала: 2024, Номер 10(4), С. 480 - 499

Опубликована: Фев. 13, 2024

Abstract Mountain meadows are an essential part of the alpine–subalpine ecosystem; they provide ecosystem services like pollination and home to diverse plant communities. Changes in climate affect meadow ecology on multiple levels, for example, by altering growing season dynamics. Tracking effects change diversity through impacts individual species overall dynamics is critical conservation efforts. Here, we explore how combine crowd‐sourced camera images with machine learning quantify flowering richness across a range elevations alpine located Mt. Rainier National Park, Washington, USA. We employed three machine‐learning techniques (Mask R‐CNN, RetinaNet YOLOv5) detect wildflower taken during two seasons. demonstrate that deep can species, providing information photographed meadows. The results indicate higher just above tree line most which comparable patterns found using field studies. two‐stage detector Mask R‐CNN was more accurate than single‐stage detectors YOLO, network performing best mean average precision (mAP) 0.67 followed (0.5) YOLO (0.4). methods anchor box variations multiples 16 led enhanced accuracy. also show detection possible even when pictures interspersed complex backgrounds not focus. differential rates depending abundance, additional challenges related similarity flower characteristics, labeling errors occlusion issues. Despite these potential biases limitations capturing abundance location‐specific quantification, accuracy notable considering complexity types picture angles this dataset. We, therefore, expect approach be used address many ecological questions benefit from automated detection, including studies phenology floral resources, can, complement wide approaches (e.g., observations, experiments, community science, etc.). In all, our study suggests metrics efficiently monitored combining easily accessible publicly curated datasets Flickr, iNaturalist).

Язык: Английский

---

The bioacoustic soundscape of a pandemic: Continuous annual monitoring using a deep learning system in Agmon Hula Lake Park

Ecological Informatics, Год журнала: 2024, Номер 80, С. 102528 - 102528

Опубликована: Фев. 17, 2024

Continuous bioacoustic monitoring is an emerging opportunity as well a challenge, allowing detection of cryptic species' activity while producing high computational demands. In this paper, we present automated framework that allows the large number bird species by their vocalizations over extended periods. The relies on BirdNET-Analyzer deep learning model. We applied to >80 species; 20 with highest recall scores were selected for further analysis. used analyze acoustic signals recorded continuously two years using autonomous recorders at various locations in Agmon Hula Lake Park, Israel. During period there was acute outbreak avian influenza area. analyzed differences occupancy between consecutive (November 2020 October 2022). examined between-year population trends 17 species, both migratory and resident, found significant decline vocal 10 species. assume related outbreak, suggesting impact pandemic may be more widespread affected greater local than previously realized. This highlights power effectiveness detecting but dramatic dynamics.

Язык: Английский

---

Soundscape components inform acoustic index patterns and refine estimates of bird species richness

Frontiers in Remote Sensing, Год журнала: 2023, Номер 4

Опубликована: Май 15, 2023

Ecoacoustic monitoring has proliferated as autonomous recording units (ARU) have become more accessible. ARUs provide a non-invasive, passive method to assess ecosystem dynamics related vocalizing animal behavior and human activity. With the ever-increasing volume of acoustic data, field grappled with summarizing ecologically meaningful patterns in recordings. Almost 70 indices been developed that offer summarized measurements bioacoustic activity conditions. However, their systematic relationships varying sonic conditions are inconsistent lead non-trivial interpretations. We used an dataset over 725,000 min recordings across 1,195 sites Sonoma County, California, evaluate relationship between 15 established using five soundscape components classified convolutional neural network: anthropophony (anthropogenic sounds), biophony (biotic geophony (wind rain), quiet (lack emergent sound), interference (ARU feedback). generalized additive models ecoacoustic indicators avian diversity. Models included explained degrees performance (avg. adj-R 2 = 0.61 ± 0.16; n 1,195). For example, we found normalized difference index was most sensitive while being less influenced by ambient sound. all were affected non-biotic sound sources degrees. combined highly predictive modeling bird species richness (deviance 65.8%; RMSE 3.9 species; 1,185 sites) for targeted, morning-only periods. Our analyses demonstrate confounding effects on indices, recommend applications be based anticipated environments. instance, presence extensive rain wind, suggest minimally geophony. Furthermore, evidence measure biodiversity (bird richness) is aggregate biotic (biophony). This adds recent work identifies reliable generalizable biodiversity.

Язык: Английский

---

Automatic detection for bioacoustic research: a practical guide from and for biologists and computer scientists

Biological reviews/Biological reviews of the Cambridge Philosophical Society, Год журнала: 2024, Номер unknown

Опубликована: Окт. 17, 2024

ABSTRACT Recent years have seen a dramatic rise in the use of passive acoustic monitoring (PAM) for biological and ecological applications, corresponding increase volume data generated. However, sets are often becoming so sizable that analysing them manually is increasingly burdensome unrealistic. Fortunately, we also computing power capability machine learning algorithms, which offer possibility performing some analysis required PAM automatically. Nonetheless, field automatic detection events still its infancy biology ecology. In this review, examine trends bioacoustic their implications burgeoning amount needs to be analysed. We explore different methods other tools scanning, analysing, extracting automatically from large volumes recordings. then provide step‐by‐step practical guide using bioacoustics. One biggest challenges greater bioacoustics there gulf expertise between sciences computer science. Therefore, review first presents an overview requirements bioacoustics, intended familiarise those science background with community, followed by introduction key elements artificial intelligence biologist understand incorporate into research. building pipeline data, conclude discussion possible future directions field.

Язык: Английский

---

Soundscape mapping: understanding regional spatial and temporal patterns of soundscapes incorporating remotely-sensed predictors and wildfire disturbance

Environmental Research Ecology, Год журнала: 2024, Номер 3(2), С. 025002 - 025002

Опубликована: Май 15, 2024

Abstract Increased environmental threats require proper monitoring of animal communities to understand where and when changes occur. Ecoacoustic tools that quantify natural acoustic environments use a combination biophony (animal sound) geophony (wind, rain, other phenomena) represent the soundscape and, in comparison anthropophony (technological human can highlight valuable landscapes both communities. However, recording these sounds requires intensive deployment devices storage interpretation large amounts data, resulting data gaps across landscape periods which recordings are absent. Interpolating ecoacoustic metrics like biophony, geophony, anthropophony, indices bridge observations provide insight larger spatial extents during interest. Here, we seven acoustically-derived bird species richness heterogeneous composed densely urbanized, suburban, rural, protected, recently burned lands Sonoma County, California, U.S.A., explore spatiotemporal patterns measurements. Predictive models driven by land-use/land-cover, remotely-sensed vegetation structure, anthropogenic impact, climate, geomorphology, phenology variables capture daily differences with varying performance (avg. R 2 = 0.38 ± 0.11) depending on metric period-of-day interpretable sound related activity, weather phenomena, activity. We also offer case study data-driven prediction soniferous activity before (1–2 years prior) after post) wildfires our area find may depict reorganization following wildfires. This is demonstrated an upward trend 1–2 post-wildfire, particularly more severely areas. Overall, evidence importance spaceborne-lidar-derived forest phenological time series characteristics modeling upscale site map biodiversity areas without prior collection. Resulting maps identify attention occur at edge disturbances.

Язык: Английский

---