Monitoring avian functional diversity in vineyards using autonomous recording units and deep learning

Summary

Objective

This research project aims to provide insight into management best practice for promotion of avian functional diversity, as well as provide tools for monitoring bird populations and assemblages on a long-term basis.

Background

This project will use passive acoustic monitoring and deep learning to facilitate avian functional diversity monitoring in vineyards. This will involve an investigation into how landscape factors and management methodologies influence avian assemblages and cross-habitat spill over. This will be accomplished by deploying autonomous recording units in vineyards and adjacent vegetation in conjunction with on-ground biodiversity surveys and landscape analysis. Deep learning classifiers will be developed for key species to facilitate efficient long-term monitoring. The focus of the project will be ‘functional biodiversity’ and therefore species which provide benefits to the grower (such as insectivorous species) as well as pest species will be prioritised (see Barbaro et al., 2017).

Research approach

The research methodology will involve placing paired Audiomoths (Hill et. al., 2018) in vineyards throughout South-east Queensland, particularly in the Granite Belt region. These recorders will be deployed on a long-term basis to capture avian variation throughout the entire growing season. The deployments will be accompanied by on-ground field surveys to observe avian movements between the vines and adjacent vegetation, as well as conduct flora and condition surveys. Landscape factors such as heterogeneity and phylogeny will also be assessed using remote sensing data. The recordings will be processed manually and tagged to form training datasets for deep learning classifiers, which will be used for automated species monitoring.

Sector benefits

This project primarily relates to Strategy 4: Grow sustainable environments. Birds are known to be indicators of environmental change (Furness & Greenwood, 2013). This research aims to develop automated tools which will contribute to monitoring sustainability practices on a long-term basis. These tools can be used to monitor bird diversity and will also have applications for pest-suppression in the form of monitoring desirable insectivores as well as avian pests.