Food codes as an AI-based framework to investigate bird foraging ecology and interspecific interactions

Area of Science:

• Ornithology
• Ecology
• Artificial Intelligence in Wildlife Monitoring

Background:

• Bird feeders are effective tools for studying bird behavior and human-bird interactions.
• Controlled variables in feeder experiments (location, food type, timing) are crucial for data accuracy.

Purpose of the Study:

• To investigate bird feeding preferences and interspecific interactions using AI-powered image analysis of garden feeder data.
• To categorize bird species based on temporal feeding patterns and assess competition dynamics.

Main Methods:

• Utilized a time-lapse camera system to capture 2.8 million image frames of birds at a garden feeder over seven months.
• Employed supervised image processing with the Squeezebrains SDK, an AI tool for wildlife monitoring, to classify 1,232,456 frames.
• Conducted cluster analysis to group nine identified bird species by temporal feeding preferences.

Main Results:

• Great tit (65.59%) and blue tit (13.62%) were the most frequent visitors, with nine species identified in total.
• Cluster analysis revealed three distinct temporal feeding groups, with chaffinch unclustered.
• Highest interspecific competition observed among blue tit, great tit, and red-billed leiothrix; severe weather increased feeder visits.

Conclusions:

• AI-driven analysis of garden bird feeder data provides robust insights into bird community dynamics.
• Temporal feeding patterns and interspecific competition vary significantly among species.
• Findings offer valuable perspectives for ecological management and understanding bird behavior in human-influenced environments.