A Novel Method to Monitor Coal Fires Based on Multi-Spectral Landsat Images

Coal fires pose a serious threat to the environment worldwide, and they are responsible for atmospheric pollution, water contamination, land subsidence and the safety of miners. The multi-spectral Landsat images offer the possibility of detecting and monitoring coal fires at large scales. In this study, the thermal infrared spectral is extracted, and a mono-window algorithm is used for retrieving coal fire temperature. However, the surface emissivity and atmospheric water vapor content play important roles in determining the temperature for this algorithm. The surface emissivity is particularly difficult to obtain with satellite overpasses because it is affected by a variety of factors. In general, an average emissivity value is assigned to represent all land cover categories, which leads to a big error for retrieving coal fire temperature. Meanwhile, atmospheric water vapor content is calculated by simulating atmospheric profile through standard atmosphere models. However, it is difficult to obtain real water vapor content and atmospheric profile is affected by many factors with each satellite pass. The lack of knowledge of the real atmospheric profile is a large constraint, and inaccurate simulation can introduce big errors. Aiming at overcoming drawbacks mentioned above and increasing the accuracy for this algorithm, the NDVI threshold method is applied to estimate surface emissivity. The NDVI threshold method separates different land cover categories, and different emissivity is assigned to different land cover classifications. Based on the ground meteorological parameters’ relationship between atmospheric water vapor and atmospheric water vapor pressure, an empirical relationship is found to estimate atmospheric water vapor content. For this method, the ground meteorological parameters are easily obtained from meteorological observation stations and it is convenient to estimate water vapor content. The mono-window algorithm is improved and coal fire temperature is retrieved. This methodology was applied to the Wuda coalfield, in China, and coal fire temperatures were retrieved and extracted from the background from 1988 to 2015 in the study area. A thorough inventory of coal fire areas and locations is annually presented, and area changes are qualitatively analyzed during the observation period. This method is considered as feasible and effective for retrieving and monitoring coal fires based on multi-spectral Landsat images in comparison to other techniques in the Wuda coalfield, China.