DATA BASE - Visualization and categorization of ecological acoustic events based on discriminant features
This repository presents audio files and root labels used in Hilasaca et al. (2021), published on Ecological Indicators DOI 10.1016/j.ecolind.2020.107316.
The region under study hosts the Long-Term Ecological Research project in the Ecological Corridor of Cantareira-Mantiqueira (LTER-CCM or Pesquisa Ecológica de Longa Duração-PELD CCM in Portuguese), localized in the transition between northeastern São Paulo state and south Minas Gerais state - Brazil. The region is considered as a conservation priority because it connects two important highly forested blocks, the Cantareira and the Mantiqueira. Landscapes are composed by forest in different succession stages, agriculture, forestry plantation (mainly Eucalyptus spp.), swamps, water, roads, villages and larger urban areas. The region is part of Atlantic Forest biodiversity hotspot, where habitat loss and fragmentation reduced forest cover to about 16%, with remnants reduced in size (84% is < 50 ha), highly isolated (average isolation c.a. 1,440 m) and high edge effect (50%) of forests are less than 100 m from any edge.
We have collected sound data in 22 landscapes, which varied in forest cover (16% to 86%) and spatial heterogeneity levels. In each landscape, we set up autonomous audio recorders Song Meter Digital Field Recorders (SM3) (Wildlife Acoustics, Inc., Massachusetts), which were fixed on trees at 1.5 m above ground. Three sites with different environment types were sampled within each landscape: forest, pasture and swamps. The equipment used two omni-directional microphones (frequency between 20 Hz and 20 kHz), and sounds were recorded at 44.1 kHz with 16 bits, in ”mono” mode.
The dataset comprised 2,277 sound files of one-minute each, divided into three classes: 615 for frogs, 822 for birds and 840 for insects. However, for the experiments, a total of four new data partitions were created: (a) DS1 (frogs, birds and insect), with 2,277 instances (minutes-files); (b) DS2 (frogs and birds), with 1,437 instances (minutes-files); (c) DS3 (frogs and insects), with 1,455 instances (minutes-files); and (d) DS4 (birds and insects), with 1,662 instances.
We also included a spreadsheet with extracted tags and characteristics and R.script used analysis. The minutes were labeled considering only the presence of acoustic signals of these three groups of species. For birds and anurans, every minute with presence of acoustic signal has been labeled as ”bird occurrence” or ”frog occurrence”. However, for insects, only minutes with predominance of insect vocalizations were labeled as “insect occurrence”. The management of the audio recordings was performed in the free programming environment R (R Core Team, 2018); labeling was done using the acoustic analysis software Raven Pro 1.5.
Hilasaca, L. H., Gaspar, L. P., Ribeiro, M. C. & Minghim, R. (2021). Soundscape_CCM1_exp01 - A acoustic dataset and root labels to visualization and categorization of ecological acoustic events.
- Liz Maribel Huancapaza Hilasa [email protected]
- Lucas Pacciullio Gaspar [email protected]
- Milton Cezar Ribeiro [email protected]
- Rosane Minghim [email protected]