Cross-scale soil moisture content monitoring of winter wheat by integrating UAV and sentinel-1/2 data
Xingjiao Yu, Qi Yin, Long Qian, Chaoyue Zhang, Lin Shao, Danjie Ran, Wenè Wang, Baozhong Zhang, Xiaotao Hu
Abstract
Accurate estimation of soil moisture content (SMC) is critical for agricultural irrigation, water resource management, and monitoring the ecological environment. The development of multi-sensor UAV platforms offers a novel approach to cross-scale SMC monitoring. This study presents an innovative framework that integrates ground, UAV, and satellite data to estimate SMC and generate county-scale spatial distribution maps of SMC in winter wheat fields. Firstly, UAV images were utilized at the subplot scale to extract winter wheat planting areas through supervised classification, and SMC was estimated by employing partial least squares regression (PLSR). Subsequently, the UAV SMC mapping results were upscaled and integrated with Sentinel-1 synthetic aperture radar (SAR) features and Sentinel-2 multispectral features to develop XGBoost-based satellite-scale SMC estimation model. The study demonstrated that at the plot scale, combining vegetation indices and texture features achieved the highest accuracy (0–20 cm: R 2 = 0.775, RMSE = 0.018 m 3 /m 3 ; 20–40 cm: R 2 = 0.723, RMSE = 0.021 m 3 /m 3 ). At the satellite scale, the XGBoost model also performed well (0–20 cm: R 2 = 0.901, RMSE = 0.0071 m 3 /m 3 ; 20–40 cm: R 2 = 0.884, RMSE = 0.011 m 3 /m 3 ). Furthermore, compared to traditional ground-satellite models, the integrated ground-UAV-satellite approach improved accuracy, with R 2 increasing by 9.53–10.52 %, RMSE decreasing by 11.11–1.25 %, and MAE reducing by 18.19–25.00 %. This cross-scale remote sensing framework enhances SMC monitoring efficiency and accuracy, offering a robust solution for large-scale applications. • Developed a novel framework integrating ground-UAV-satellite data for large-scale soil moisture monitoring. • Generated accurate 10 m-resolution county-level soil moisture maps using the XGBoost model. • Validated the model's reliability against traditional ground-UAV methods. • SHAP analysis revealed feature contributions, ensuring interpretable multi-source data fusion.