A comparison of change point detection methods for pest outbreak detection

Area of Science:

• Agricultural Science
• Data Science
• Biosecurity

Background:

• The Agricultural Quarantine Inspection Monitoring (AQIM) program faces challenges in detecting pest outbreaks.
• Existing methods often focus on sampling strategies rather than direct outbreak detection analysis.

Purpose of the Study:

• To numerically evaluate change point detection (CPD) algorithms for identifying significant shifts in pest arrival rates within AQIM data.
• To assess the reliability of CPD methods in detecting pest outbreaks when arrival rates exceed critical thresholds.

Main Methods:

• Computational experiments were conducted to compare the accuracy, sensitivity, and robustness of various CPD algorithms.
• Specific focus on cumulative sum (CUSUM) and exponential moving average (EMA) techniques.
• Evaluation under diverse outbreak scenarios and data conditions.

Main Results:

• CPD techniques, particularly CUSUM and EMA, demonstrated high detection rates for pest outbreaks.
• Low false alarm rates were observed, even for small-scale outbreaks.
• The methods proved robust across various challenging data conditions.

Conclusions:

• CPD algorithms are feasible for integration into AQIM operations for enhanced outbreak detection.
• These methods offer a scalable and practical approach to strengthening agricultural biosecurity.
• Effective pest outbreak detection is crucial given global trade and evolving pest risks.