Fungi are a major biosecurity concern because they cause a wide variety of plant diseases that threaten agricultural productivity, market access, and natural ecosystems. The surveillance of fungal pathogens is an essential part of Australia’s biosecurity efforts. However, airborne fungal spores are especially challenging to monitor due to their microscopic size, wind-driven dispersal over long distances, and their low concentration relative to the vast volume of air that can carry them.
A conventional spore trap is a device designed to capture spores from the air for subsequent analysis. When deployed in the field, a spore trap can be used to monitor the abundance and diversity of fungal pathogens in the surrounding environment. Existing research has developed a variety of designs for spore traps, ranging from low-cost passive devices to sophisticated high-volume automatic samplers.
Traditional spore traps are often labour-intensive and rely heavily on costly, time-consuming laboratory analysis. Conversely, modern real-time bioaerosol sensors can be expensive and operationally complex for widespread agricultural deployment. Currently, there is a critical need for an integrated system that is simultaneously accurate, scalable, and cost-effective to identify both known pathogens and unexpected incursions.
Aim: This project aims to design, develop, and field-validate an adaptive plant biosecurity surveillance framework. The system will combine a network of smart spore traps with machine learning (ML) to proactively detect airborne fungal pathogens. Specifically, the rotating arm-impaction sampler will be modified to increase spore capture efficiency through computational fluid dynamics.
On the other hand, this project will develop an ML-based system to complement the existing molecular diagnostics workflow and reduce operational costs. Rather than serving as the sole diagnostic tool, ML acts as a rapid triaging tool. By capturing images before molecular analysis, a decision is made as to whether a sample should proceed to molecular diagnostics, all while preserving an archival visual record of the sample alongside metadata like temperature, wind speed, and geolocation. By identifying known threats and dynamically flagging anomalies, the framework will optimise the deployment of limited laboratory analysis resources toward high-risk nodes within the network. Ultimately, this research will enable targeted interventions, reduce crop losses, and minimise pesticide use, thereby safeguarding agricultural productivity and environmental health.
Meet the PhD student for this project
Shiron Thalagala
“My name is Shiron Thalagala, and I hold a Bachelor’s degree in Production Engineering and a master’s degree in Electromechanical Engineering. I am passionate about developing mechatronic systems and applying machine learning for systems control. Currently, I am pursuing a PhD titled where I am developing an advanced fungal spore monitoring system. My research combines electronics, mechanical design, and computational fluid dynamics (CFD), while also integrating machine learning and computer vision to detect and analyse fungal spores. The ultimate goal of my work is to strengthen the Australian biosecurity industry by providing reliable surveillance tools against potential fungal pathogens.
Outside of research, I enjoy running, meditation, and yoga.”
Supervisors and advisors
Associate Professor Bronson Philippa and Professor Wenxian Lin, James Cook University.
Dr Rohan Kimber and Dr Nicole Thompson, South Australian Research and Development Institute (SARDI).
Approach
The research will be executed through the following integrated phases:
- Developing the hardware: The research involves engineering a novel spore trap, utilising a rotating arm-impaction style sampling approach. It includes two rods coated with adhesives that collect spores through impaction while rotating along a common axis. It will be aerodynamically optimised via computational fluid dynamics to maximise spore capture efficiency. This hardware will be combined with automated microscopy to capture images from the sampling rods and send them to the detection algorithm in the cloud.
- Curating a novel dataset: A dataset of fungal spore images will be developed and benchmarked. The images will be collected by the smart trap nodes and annotated to train subsequent ML models.
- Developing the computational intelligence: The system will utilise supervised deep-learning models for the automated identification of known fungal spores. Additionally, an adaptive decision-making algorithm will be created to detect anomalies (e.g., novel threats). Based on this detection, a risk probability will be calculated to execute the rapid triaging and systematically prioritise high-risk samples for further laboratory analysis.
- System validation: The physical hardware and ML intelligence will be integrated into a single operational biosecurity system. This framework will be validated in real-world field settings to assess its performance, including operational reliability, detection accuracy, and the timeliness of its alerts.
Want to know more? Email shiron.thalagala@my.jcu.edu.au.