Extraction of Thermal Anomaly Intensity in Underground Coal Fire Areas and Analysis of Its Driving Factors Using Multi-temporal UAV Remote Sensing

Objectives: Underground coal fires are global hazards causing resource loss and environmental pollution. Establishing a multi-temporal UAV-based analytical framework is essential to overcome the limitations of single-temporal static observations and to quantitatively characterize the thermal anomaly intensity and spatio-temporal evolution of associated surface deformation and fractures. Methods: Taking the Miaoergou coal fire zone in Xinjiang as the study area, multi-temporal UAV thermal infrared, optical, and digital elevation model (DEM) data collected from 2019 to 2022 were utilized. An adaptive kernel density-based method was developed to extract variations in thermal anomaly intensity. Under the constraints of the coal seam's geological conditions, a comprehensive analysis was conducted on the intensity variations of thermal anomalies and their influencing factors, incorporating surface deformation and fracture development characteristics. Results: 1) The adaptive kernel density method significantly enhanced the spatial representation of thermal anomalies. Compared with the fixed bandwidth approach, the overall kurtosis of the thermal anomaly intensity distribution was markedly higher across different monitoring periods, rising from 24.94 to 29.22 in 2019 and from 11.65 to 15.36 in 2022. This method identified a pattern of "localized decline with leading-edge enhancement" in the West zone and continuous intensification in the East zone. 2) Multi-temporal sensing effectively characterized the coupled evolution of thermal anomalies, deformation, and fractures, supporting the analysis of ventilation and heat release mechanisms. Fracture distribution closely corresponded to enhanced thermal zones, while subsidence displayed synchronous or spatially lagged patterns. 3) Comprehensive analysis indicated that the monitoring results obtained from multi-temporal UAV remote sensing were highly consistent with the spatial distribution of the coal seam geology, enabling effective identification of coal fire spread pathways and their stage-wise evolution patterns. The coal fire in the study area generally propagated along the coal seam dip from deep to shallow, and extended from west to east under the control of the seam strike, with the observed evolution direction matching the northeastward horizontal expansion revealed by UAV monitoring. Conclusions: Multi-temporal UAV remote sensing could effectively reveal the dynamic variations of coal fire thermal anomaly intensity and their driving factors, providing a scientific basis for precise identification of combustion states and targeted mitigation of coal fires.