Automated Surveillance of Lepidopteran Pests

Switching from manual inspections to automated detection could be the future of biosecurity pest control in New Zealand.

A recent study into the potential of using smart sensor traps in surveillance programs highlights the potential in IoT monitoring systems.

As an example, in New Zealand, Tephritidae fruit flies and gypsy moth (Lymantria dispar) – potential invasive pests of significant economic importance – are the subject of a national surveillance program. The program uses of 9000 pheromone traps nationally. These need to be checked every two weeks over summer at a cost of about NZ$1 million. Around 95% of the time the traps are empty. Automating detection would not only significantly reduce costs but eliminate the lag between pest appearance and detection that occurs with manual inspection – which can make the difference between eradication and a disaster.

Researchers trialled a delta-shape sticky trap with pheromone attractant and optoelectronic sensor in a series of laboratory and field trials (pictured above). Not only could the sensor detect entries into the trap, it could also detect the frequency of the wing beat, which could be used in identification. (Most insects have a unique frequency wing beat.) For the trial they needed to use pests present in New Zealand, so focused on three moth pests common in apple orchards.

Whilst further optimisation and trials are necessary, the prototype produced encouraging results. Firstly, the number of visual detections was always significantly higher than the number of insects caught on the sticky trap itself. Part of this overestimate was due to insects entering the trap several times before getting caught. Whilst this may be detrimental in situations where an accurate pest count is important, for use situations such as biosecurity surveillance, it is desirable to have a sensitive system that picks up on target pest activity.

Secondly, around 70% of detections were identifiable at the species level i.e. if pests flew in too quickly or walked into the trap, their activity did not provide a wing beat frequency pattern that allowed identification. The authors believe despite this, such as system would provide significant benefits over a manual surveillance system. As with any IoT system it can be set up for automated alerts, providing additional analysis for users. Futhermore, machine learning can be used on the data set to improve accuracy.


Further reading: Welsh, T.J. et al. (2022). Automated Surveillance of Lepidopteran Pests with Smart Optoelectronic Sensor Traps. Sustainability 14(15): 9577. DOI:10.3390/su14159577