Posted in | News | Ecology | Ecosystems

Evaluating Machine Learning Approaches for Predicting Black Planthopper Pest Spread

The black planthopper, Ricania speculum, feeds on more than 60 plant species, including crops such as citrus, tea and coffee. Feeding on young branches can weaken plants, cause leaves to yellow and, in severe cases, lead to branch dieback. The insect is established across parts of East and Southeast Asia and has been recorded in northern Italy. Researchers use species distribution models to estimate where pests might find suitable conditions by comparing known sightings with local climate data. Different models can produce different maps, however, making it difficult to judge where surveillance will be most useful as the climate changes.

A study (DOI: 10.48130/jzser-0026-0003) published in Journal of Zoological Systematics and Evolutionary Research on 22 July 2026 by Wang-Hee Lee's team, Chungnam National University, reports that three machine-learning models identify broadly similar areas of current climate suitability and project a northward expansion under future climate scenarios.

The researchers compiled global occurrence records through June 2025 and cross-checked them against records from the European and Mediterranean Plant Protection Organization. They filtered out geographically clustered observations to reduce the influence of heavily sampled locations, leaving 437 records for analysis. They then selected nine of 19 temperature and precipitation variables, removing strongly correlated measures to limit redundancy. Using those records and variables, the team trained and evaluated three algorithms, MaxEnt, Random Forest and XGBoost. All three scored above 0.85 on the true skill statistic, which measures how well a model distinguishes occurrence locations from comparison locations.

XGBoost scored slightly higher than the other two and produced the smallest estimate of currently suitable land at 8.9%, compared with 9.3% for Random Forest and 11.1% for MaxEnt. To combine the maps, the researchers put the suitability scores on a common scale and weighted each model by its performance. The combined map classified 7.6% of the analyzed land grid, which excluded Antarctica, as moderately suitable. It highlighted parts of East and Southeast Asia, India, northern South America, the European Mediterranean coast and coastal Central Africa.

The researchers then projected suitability for three future periods using three global climate models and two emissions scenarios. For 2081-2100, the combined estimate reached 12.3% under the intermediate scenario and 18.0% under the high-emissions scenario. Under higher emissions, favorable conditions extended farther north in Europe, while projected suitability increased in South America from midcentury onward. The models differed in how much land they projected would become suitable, with Random Forest showing the greatest variation across scenarios. About 5.5% of the analyzed land grid was classified as moderately suitable under current conditions and remained so across the future projections.

The study identifies both areas where suitable climates could persist and areas where changing temperatures and rainfall could create new opportunities for the pest to establish. Actual establishment will also depend on whether the insect reaches those areas, finds host plants and survives local conditions. Combining the climate maps with information on host crops and trade routes could help surveillance teams focus field checks where introductions are most likely to lead to established populations.

Tell Us What You Think

Do you have a review, update or anything you would like to add to this news story?

Leave your feedback
Your comment type
Submit

While we only use edited and approved content for Azthena answers, it may on occasions provide incorrect responses. Please confirm any data provided with the related suppliers or authors. We do not provide medical advice, if you search for medical information you must always consult a medical professional before acting on any information provided.

Your questions, but not your email details will be shared with OpenAI and retained for 30 days in accordance with their privacy principles.

Please do not ask questions that use sensitive or confidential information.

Read the full Terms & Conditions.