Data from: Free-ranging livestock grazing shifts the acoustic community in a Northeast Asian temperate forest
Data files
May 04, 2026 version files 51.91 MB
-
Acoustic_indices.csv
51.67 MB
-
Acoustic_SEM_Analysis.R
2.94 KB
-
Integrated_Acoustic_and_Vegetation_Data.csv
226.46 KB
-
README.md
6.94 KB
Abstract
Understanding livestock‒wildlife interactions, especially in forest ecosystems, is critical for biodiversity conservation and sustainable land management. However, the long-term and cascading impacts of livestock grazing on forest structure and community bioacoustics are important yet largely neglected areas of research. Here, we used acoustic indices and a sound event detection (SED) model to evaluate the effects of continuous cattle grazing on seasonal soundscapes in Northeast China. We collected and analyzed over 18,785 h of recordings from 10 cattle-grazed forest plots and 10 ungrazed forest plots in Northeast China. We identified sound events in each recording via deep learning and calculated six acoustic indices, as well as extracted vegetation characteristics using LiDAR point cloud data. Our results revealed that grazing activities significantly changed seasonal soundscape dynamics, with biophony being highest in grazed forests and lowest in ungrazed forests in winter. Livestock shifted the forest soundscape composition by increasing the audibility of birds and insects while decreasing the vocalizations of sika deer (Cervus nippon) and crows, resulting in reduced sound diversity and complexity in grazed forests. We also found that grazing can reduce the leaf area index, herbaceous plants and canopy density, which can influence these effects indirectly. Interestingly, cowbells noticeably altered the dawn chorus of birds; during spring and summer grazing periods, the chorus was characterized by an increased bird calling rate and greater vocal complexity (elevated Acoustic Complexity Index), patterns consistent with a behavioral adjustment to acoustic masking. This study highlights how livestock modify forest acoustic communities. To preserve natural soundscapes, we suggest mitigating cowbell noise through silent trackers (e.g., GPS) or reduced bell density in priority zones. Sustainable practices, including rotational grazing and buffer zones, are also vital to maintain forest structure and acoustic diversity. We suggest that integrating SED models with acoustic indices provides a robust framework for monitoring such anthropogenic disturbances.
Dataset DOI: 10.5061/dryad.0zpc867cg
Description of the data and file structure
This dataset was collected as part of a year-long (July 2022–July 2023) ecological study conducted in the Northeast Tiger and Leopard National Park, China. The research aimed to evaluate the impacts of free-ranging cattle grazing on temperate forest soundscapes and vegetation structure. We employed a comparative experimental design across 10 grazed and 10 ungrazed forest plots, integrating passive acoustic monitoring with high-resolution LiDAR scanning.
Files and variables
File: Acoustic_indices.csv
Description: Values of six traditional acoustic indices (ACI, BIO, ADI, AEI, H, NDSI) calculated for all 5-minute recordings, along with their principal component scores (PC1, PC2).
Variables
- Line: Forest type or treatment category. GZ indicates grazed forest plots; BD indicate ungrazed forest plots if used in the full dataset.
- FILENAME: Original audio file name for each 5-minute recording.
- Samplepoint: Unique sampling point/site identifier where the recorder was deployed.
- ndate: Recording date in numeric
YYYYMMDDformat. - Date: Recording date in calendar format.
- Month: Month of recording, coded from 1 to 12.
- Week: Week number of the year when the recording was collected.
- Hour: Recording time in decimal hour format. For example,
9.5= 09:30 and10.5= 10:30. - Longitude: Longitude of the sampling point in decimal degrees.
- Latitude: Latitude of the sampling point in decimal degrees.
- ACI: Acoustic Complexity Index. Measures temporal variation in sound intensity across frequency bins and is used as an indicator of acoustic complexity or vocal activity.
- BIO: Bioacoustic Index. Quantifies acoustic energy in the biological frequency band, calculated here over 2–11 kHz.
- ADI: Acoustic Diversity Index. Measures the distribution of acoustic energy across frequency bands and represents soundscape diversity.
- AEI: Acoustic Evenness Index. Measures evenness of acoustic energy across frequency bands; higher/lower values reflect differences in dominance among frequency bands.
- H: Acoustic Entropy. Represents overall acoustic complexity by combining temporal and spectral entropy.
- NDSI: Normalised Difference Soundscape Index. Indicates the relative dominance of biological sounds compared with anthropogenic sounds; calculated using 1–2 kHz as the anthropogenic band and 2–11 kHz as the biological band.
- Suntimerad: Solar time angle in radians, used to represent diel/solar timing of each recording.
- Part: Diel period category of the recording, for example
daytime. - PC1_org: Original first principal component score derived from PCA of acoustic indices before rescaling/normalisation.
- PC2_org: Original second principal component score derived from PCA of acoustic indices before rescaling/normalisation.
- Season: Season of recording.
- PC1: Rescaled or transformed first principal component score derived from acoustic indices. In the manuscript, PC1 mainly represents sound diversity and acoustic complexity.
- PC2: Rescaled or transformed second principal component score derived from acoustic indices. In the manuscript, PC2 mainly represents biophony prominence in the soundscape.
File: Integrated_Acoustic_and_Vegetation_Data.csv
Description: The master dataset used for statistical modeling. It integrates weekly acoustic audibility scores, NMDS ordination axis scores, and site-level LiDAR-derived vegetation metrics (e.g., LAI, canopy density).
Variables
- Samplepoint: Unique sampling point/site identifier.
- Week: Week number of the year.
- Line: Forest type or treatment category, e.g., grazed or ungrazed forest.
- Season: Season of recording: spring, summer, autumn, or winter.
- aud_Axis3_audibility: NMDS axis 3 score derived from weekly audibility data of sound events.
- indices_Axis3_psd: NMDS axis 3 score derived from weekly acoustic-index or power spectral density data.
- PC1: First principal component score derived from acoustic indices; mainly represents sound diversity and acoustic complexity.
- PC2: Second principal component score derived from acoustic indices; mainly represents biophony prominence in the soundscape.
- Human_aud: Weekly audibility of human sounds, calculated as the proportion of recordings in which human sounds were detected.
- Vehicle_aud: Weekly audibility of vehicle sounds.
- Cattle_aud: Weekly audibility of cattle vocalisations or cattle-related sounds.
- Cattle_bells_aud: Weekly audibility of cowbell sounds associated with free-ranging cattle.
- Anthrophony_aud: Weekly audibility of anthropogenic sounds, including human, vehicle, cattle, and cowbell-related sounds.
- Biophony_aud: Weekly audibility of biological sounds, including birds, insects, amphibians, mammals, and other animal sounds.
- Bird_aud: Weekly audibility of bird sounds, excluding crow sounds if crows were treated separately.
- Insect_aud: Weekly audibility of insect sounds, excluding fly sounds if flies were treated separately.
- Frog_aud: Weekly audibility of amphibian/frog sounds.
- Sika.deer_aud: Weekly audibility of sika deer vocalisations.
- Roe.deer_aud: Weekly audibility of roe deer vocalisations.
- Moving_aud: Weekly audibility of animal movement sounds.
- Fly_aud: Weekly audibility of fly sounds.
- Crow_aud: Weekly audibility of crow/corvid sounds.
- Wind_aud: Weekly audibility of wind sounds.
- Rain_aud: Weekly audibility of rain sounds.
- Quiet_aud: Weekly proportion of recordings classified as quiet environment, with no detectable biological, anthropogenic, or geophysical sound events.
- Tree_num: Number of trees within the 20 m × 20 m vegetation plot centred on the acoustic recorder.
- Tree_height: Mean or representative tree height within the vegetation plot, derived from LiDAR point cloud data.
- DBH: Mean diameter at breast height of trees within the vegetation plot.
- LAI: Leaf Area Index, representing total leaf surface area per unit ground area.
- Viewshed: Visibility or openness metric derived from vegetation structure around the sampling point.
- rumple_index: Canopy rugosity or structural complexity index; higher values indicate more complex canopy surface structure.
- Density05: LiDAR point-cloud density of the 0.5–2 m shrub layer.
- Density25: LiDAR point-cloud density of the 2–25 m canopy layer.
Code/software
For statistical analysis and data reproduction, the open-source software R (version 4.2.0 or later) is required.
The included R script (Acoustic_SEM_Analysis.R) was used to perform the piecewise structural equation modeling (pSEM). The following R packages must be installed and loaded:
piecewiseSEM
lme4
nlme
glmmTMB
