Predicting potential occurrence of pine wilt disease based on environmental factors in South Korea using machine learning algorithms

Pine wilt disease (PWD) is one of the most destructive diseases affecting pine trees, and South Korea is one of the most severely damaged countries in the world. Based on the information on PWD occurrence and their environmental characteristics (i.e., geographical, meteorological, and land-use) in South Korea, we evaluated the conditions most conducive for PWD occurrence and developed projection models using machine learning algorithms; random forest (RF) and maximum entropy (Maxent). Our results showed that PWD mainly occurred in areas like highly urbanized area; low elevations, at close proximity to roads. Also, both RF and Maxent models presented high prediction performance for PWD occurrence. Geographical factors (e.g., elevation and distance to roads) were major determinants of PWD occurrence and largely contributed to explaining variability and partial dependence plots of each model. We developed an ensemble model composed of the RF and Maxent models to predict a potential risk map for PWD occurrence on a national scale. In South Korea, most territory was included potential risk of PWD occurrence, and it was predicted to be expanded in the future according to the climate change. The study results showed a high utility for use in surveillance and monitoring of PWD occurrence by inferring the spread pathway or spread direction of PWD based on the potential occurrence map.

Automated summary

The study evaluated the conditions most conducive for pine wilt disease (PWD) occurrence in South Korea and developed prediction models using machine learning algorithms, showing that geographical factors were major determinants of PWD occurrence. A potential risk map for PWD occurrence on a national scale was predicted, with high utility for surveillance and monitoring based on the spread pathway or spread direction inferred from the potential occurrence map.