Sensitivity of Sentinel-1 and Sentinel-2 features for detecting pine wilt disease under complex interference

Pine wilt disease (PWD) has caused severe ecological and economic losses worldwide, creating an urgent need for cost-effective and large-scale monitoring tools to support pest management. Satellite remote sensing complements UAV observations by enabling wide-area, repeated monitoring at low cost; however, its effectiveness is limited by background interference from red-yellow soil and seasonal broadleaf discoloration. This study evaluated the ability of multisource satellite time-series features to improve PWD detection under complex environmental conditions.

Multitemporal features derived from Sentinel-1 (S1) and Sentinel-2 (S2) data using Complementary Ensemble Empirical Mode Decomposition and Hilbert-Huang Transform showed strong discrimination ability. The vegetation index VIgreen and its IMF₃ temporal component were the most sensitive indicators. Red-edge, near-infrared and shortwave infrared features effectively reduced background interference. Although S1 features alone showed limited performance, incorporating temporal information improved their sensitivity. The integration of single-time, sparse-temporal and time-series features achieved an overall accuracy of 0.76.

Satellite time-series features significantly improve PWD detection under complex interference conditions by capturing disease-related temporal dynamics. This approach provides a reliable and scalable tool for operational forest pest monitoring and supports improved surveillance and management of PWD. © 2026 Society of Chemical Industry.