Developing an automated and digital tool for integrated bird pest management in fresh produce fields

Birds carry foodborne pathogens that can contaminate fresh produce in fields. Farmers try various methods to keep birds away from the field, but these methods become ineffective as birds quickly get habituated. Also, different bird species respond differently to these methods, and not all birds carry pathogen spread risk. The current techniques cannot identify bird species and aim to keep all birds away, including beneficial pests controlling birds. Therefore, this proposal aims to develop a digital tool using sensors and models combined with various deterrent methods to deter birds from fields. The tool will perform digital sound surveillance to identify bird presence and species, which will activate multiple deterrents automatically, improving efficiency and effectiveness. The goals include developing a machine learning model for bird identification, integrating various deterrent methods, and evaluating the toolbox in produce fields. The digital tools, combined with surveillance and multiple deterrent methods, would protect crops and allow for species-level identification to target high-risk birds while maintaining ecological balance. In the future, this toolbox could be used on an autonomous robot for pest bird management. The primary benefits are preventing crop damage, reducing contamination risk, and offering a cost-effective bird deterrent solution.

Birds carry foodborne pathogens and pose a significant risk to food safety in fresh produce fields. Farmers are concerned and employ various bird dispersal techniques to minimize crop contact or damage. However, bird habituation is a known problem to the currently employed dispersal techniques since birds acclimate quickly to fixed frequency and uniform hazing techniques. Additionally, different bird species respond differently to various dispersal methods, and not all bird species carry an equal risk of pathogen spillover. However, current dispersal practices lack bird species-level identification and are targeted towards removing all bird habitats from produce fields or orchards, which may solve the problem but harm the beneficial, pest-eating birds, further disturbing the ecosystem balance. This proposal specifically aims to develop a digital tool to deter birds from fields through digital surveillance and an automated trigger system by integrating sensors, electro-mechanical systems, and machine-learning models. The developed, novel tools will perform digital surveillance (bird presence and species identification) and autonomous trigger systems, significantly improving the efficiency, effectiveness, and longevity of the current fixed frequency-based methods. In this proof-of-concept proposal, we plan to pursue the following objectives: 

(1) Develop a machine learning model to identify bird presence and species (i.e., digital surveillance) through sound in the produce field/orchard. 

(2) Integrate multiple bird dispersal methods (e.g., visual and auditory) to develop a bird deterrent toolbox. 

(3) Evaluate the digital toolbox in fresh produce fields for effective bird deterrence and management. 

The digital tools combined with multiple dispersal methods and digital surveillance systems will safeguard the produce field from birds. Additionally, it will allow bird species level identification to target the high-risk species with automated triggers, allowing beneficial birds to maintain the ecological balance through diversity and conversation efforts. The physical outcomes of this project would lead to the development of a prototype digital toolbox for safeguarding the produce field/orchard. In the future, the toolbox will be deployed on a solar-powered, autonomous robot platform for bird pest management while performing other tasks such as fresh produce field scouting for decision support, food safety inspection, and so on. The digital tools will primarily benefit fresh produce, horticultural, and specialty crop growers. The primary benefits/impacts include avoiding costly produce damage and reducing food safety contamination risk by birds with a cost-saving, practical, and effective bird deterrent method. The immediate outcome will be a simple, low-cost bird deterrent toolbox with digital surveillance and trigger capabilities for bird deterrence and quantifying avian prevalence and pathogen spillover risk. The long-term outcome would be harnessing an emerging technology to provide a practical solution to restrict the potential pathways of avian foodborne pathogen spillover to mitigate the produce contamination risk without jeopardizing the ecosystem benefits or conservation.

Objective 1: Develop a machine learning model to identify bird presence and species (i.e., digital surveillance) through sound in the produce field.

Objective 2: Integrate multiple bird dispersal methods (e.g., visual and auditory) to develop a bird deterrent toolbox. 

Objective 3: Evaluate the digital tool or platform in fresh produce fields for effective bird deterrence and management.

This project successfully developed and tested an automated robotic bird-deterrence system to address food-safety risks and economic losses associated with bird activity in fresh produce fields. The developed robotic system demonstrated reliable real-time acoustic detection of bird presence and triggered multiple deterrent modules. Integration of audio-visual deterrents with feedback-based activation reduced unnecessary firing compared with fixed-frequency approaches.

Preliminary field trials confirmed the operational feasibility of the robotic system under outdoor conditions, including stable system integration, adequate battery life for extended deployment, and coordinated triggering between surveillance and deterrent components. Immediate dispersal behavior was observed following deterrent activation.

However, bird activity during winter testing was low and irregular, limiting the ability to conduct statistically robust efficacy comparisons across treatments. Additionally, reliance on sound-based surveillance resulted in missed detections when birds were present but not vocalizing. These findings highlight both the promise of adaptive, artificial intelligence (AI)-driven deterrence systems and the importance of multi-season validation.

Recommendations

• Conduct multi-season field trials during peak bird activity to generate statistically robust deterrence performance data.
• Integrate complementary vision-based detection to address silent bird presence and improve overall detection reliability.
• Evaluate long-term habituation patterns and treatment sustainability over extended deployment periods.
• Conduct economic analyses to assess cost-effectiveness and return on investment for grower adoption.
• Explore semi-autonomous or fully autonomous navigation to enable scalable, hands-off deployment in commercial produce systems.

Overall, the project established technical feasibility and demonstrated the potential of AI-enabled adaptive deterrence systems as a proactive tool for bird pest management. Further validation and system refinement will support practical field implementation and grower adoption.