Sentinel‐2 and UAV Hyperspectral Image Data Fusion Used in Natural Disasters Cause Soil Pollution of the Distribution Map

Abstract

Natural disasters, particularly, those occurring near mining areas, can damage the soil and lead to pollution, resulting in significant harm to the ecosystem. Timely and large‐scale investigation of soil composition in areas affected by geological disasters is crucial for monitoring pollution. With the traditional method, soil samples were mainly collected manually for the survey, which was extremely constrained in terms of efficiency and accuracy due to the complex topography of the areas where the geohazards had occurred. Using an unmanned aerial vehicle (UAV) with a hyperspectral camera to acquire hyperspectral images of soil, combined with machine learning algorithms for soil composition estimation, has the advantages of high efficiency and broad coverage. Meanwhile, the fusion of satellite imagery and UAV imagery has the potential to improve the accuracy of soil estimation models. In this study, we selected an area near a mine in Guilin, Guangxi, China, that had been affected by a mudslide caused by a geological hazard as the study area. Firstly, 30 soil samples were collected to test the total Zinc (Zn), total Lead (Pb), and total Cadmium (Cd) content in the soil. Second, the UAV hyperspectral images and Sentinel‐2 satellite images of the study area were obtained and preprocessed. Thirdly, the soil estimation model was generated using UAV images, Sentinel‐2 satellite images, and their fused data combined with deep learning algorithms, respectively. Finally, the model with the best accuracy was selected to generate a soil distribution map. The results show that after the fusion of UAV images and Sentinel‐2 images, the soil estimation model reached its highest accuracy, which is significantly improved compared with using UAV images and Sentinel‐2 image data alone. This shows that the fusion data of Sentinel‐2 images and UAV images, combined with deep learning algorithms, has advantages and can improve the ability of soil pollution monitoring.