An intelligent and dynamic monitoring method for coal fire hazards based on GEE and multi-source remote sensing data fusion

ZHANG Yuan; CUI Liu; YANG Hui; FENG Jian; SHEN Bingjian; WANG Yong; MA Guoqiang

doi:10.13247/j.cnki.jcumt.20250442

Abstract：Coal fires are a major geological hazard that severely constrain coal resource security and ecological environmental protection in China.To address the limitations of traditional remote sens-ing methods for coal fire monitoring,specifically in spatial resolution and anomaly extraction accu-racy,this study takes the Laojunmiao coalfield in the eastern Junggar Basin of Xinjiang as the re-search area and develops an intelligent coal fire monitoring framework based on the Google Earth Engine(GEE)platform by integrating multi-source remote sensing data.First,high-resolution multispectral features from Sentinel-2 and thermal infrared data from Landsat-8 are fused to con-struct a multi-dimensional input feature set composed of spectral bands and indices.An improved U-Net network is then employed to achieve 10 m spatial downscaling reconstruction of land surface temperature(LST),significantly enhancing the spatial accuracy of thermal anomaly information.Second,an unsupervised anomaly detection method based on autoencoders is introduced to auto-matically identify coal fire regions without the need for manual labeling,enabling precise boundary extraction and dynamic tracking of distribution changes.Finally,by combining multi-temporal coal fire identification results,a coal fire thermal anomaly index(CFTA)is proposed to quantitatively analyze the spatiotemporal evolution of coal fire activities and their mitigation effects from 2016 to 2025.The results demonstrate that the proposed intelligent monitoring framework can effectively capture the dynamic changes during coal fire development and remediation processes,offering high spatial resolution and sensitivity.This provides a reliable technical pathway for accurate monitoring and prevention of coal fire hazards.