Integrating active and passive remote sensing data for quantifying wheat stem sawfly infestation in Montana

Wheat stem sawfly (Cephus cinctus Norton; WSS) remains one of the most damaging pests of cereal crops in the Northern Great Plains, causing estimated annual losses of US $44–80 million in Montana alone. Effective management remains difficult, as existing strategies, solid-stem cultivars, crop rotation, and swathing often face economic or logistical limitations. Conventional detection methods rely on manual stem dissections to identify larvae and injury, a labor-intensive process unsuitable for large-scale monitoring. This study presents a scalable approach that integrates Synthetic Aperture Radar (SAR) imagery with LiDAR-derived Canopy Height Models (CHMs) to quantify wheat lodging, a key indicator of WSS infestation. LiDAR CHMs provided high-resolution crop height measurements for model validation, while SAR data supported predictive modeling. Lodging patterns identified from LiDAR were used to train the SAR-based model. Field validation across two commercial spring wheat fields in Montana included systematic sweep net sampling and postharvest stem dissections to confirm larval presence. Machine learning models, specifically Random Forest classifiers, were trained using SAR-derived features and achieved 82.8% overall accuracy based on 107,106,816 training and 26,776,704 test samples. Preliminary results demonstrate that SAR data can enable large-scale, spatially explicit assessment of WSS-induced lodging. This approach supports predictive modeling and data-driven management decisions, including cultivar selection, optimized swathing timing, and targeted crop rotation. Overall, the integration of SAR and LiDAR provides a robust framework for rapid, scalable monitoring of pest-induced crop damage in cereal production systems.