Bioacoustic Bird Monitoring in Mandla Agricultural Landscapes

Comprising acoustic monitoring data from 15 locations in the Mandla district of Madhya Pradesh, India, collected from August 2024 to November 2025. The data was processed using machine learning tools to identify bird species from audio recordings, with identifications assigned a confidence rating. The study aims to understand the long-term impacts of landscape variables and farming practices on bird diversity.

Use Cases

• Analyze patterns in bird species identification from acoustic data to assess biodiversity across 15 monitoring locations.
• Investigate the relationship between landscape variables described in the study and the acoustic detection of bird species.
• Use the confidence ratings assigned to species identifications to filter or weight observations for ecological trend analysis.
• Study the impact of varied sampling effort, caused by data gaps from rain and inaccessibility, on long-term monitoring conclusions.

Strengths

• Data collection spanned over a year, from August 2024 to November 2025, providing a temporal perspective.
• Recordings were made at 15 distinct locations within a heterogeneous landscape of agriculture, forests, and waterbodies.
• Species identifications from audio were processed using a machine learning tool with a defined confidence rating threshold of 85%.

Limitations

• The description notes huge data gaps at certain locations due to rains, inaccessibility, and lost memory cards, leading to varied and incomplete sampling effort.
• The specific data format, column structure, and total volume of recordings or identified events are unknown, limiting immediate analytical utility.
• The geographic scope is limited to one district in central India, which may restrict generalizability to other regions.

Provenance

Source

The Alliance of Bioversity International and CIAT Dataverse, authored by Krishnan, Smitha.

Collection Method

Acoustic data collected using Song meter Micro devices at 15 locations, set to record 1-minute clips every 10 minutes at 44100Hz. Bird identifications were processed using a machine learning tool from Wildersensing.

Time Range

August 2024 to November 2025.

Freshness

Data collection concluded in November 2025, and the metadata was last updated in February 2026.

Geography

Mandla district, Madhya Pradesh, India, characterized by agricultural land, forests, rangelands, waterbodies, and human settlements.