Detection and Delineation of Coal Mine Fire and Estimation of Greenhouse Gas Emissions Using Satellite Data

Abstract

Coal is a fossil fuel that tends to catch fire in in-situ conditions primarily due to spontaneous combustion. Many other coal-producing nations across the world, including China, India, the United States of America (USA), Australia, Indonesia, face significant risks due to subsurface and surface coal fires. In-situ coal fires lead to substantial loss of coal and pollute the environment through the emission of toxic gases due to coal burning. Thus, monitoring coal fire propagation is needed to minimize the environmental risk and extreme loss of energy resources. However, monitoring coal fires using the traditional survey method is challenging and incurs a higher cost and workforce. The present study aims to develop a methodology to detect and delineate the surface and subsurface coal fires using a remote sensing-based approach. The study also analyzed the dynamics of coal fires using time-series analysis of satellite data. The study also extended to estimate the role of coal burning on greenhouse gas emission using remote sensing-based approaches. The study region selected for conducting the study is Jharia coalfield (JCF). JCF is one of India's largest and oldest coalfields, affected by fire for the last few decades. The coal fire detection using remote sensing data was done using multiple approaches and validated using field data. The study used different techniques to detect the surface and sub-surface coal fires. The sub-surface fire detection was done using two different methods (radiative transfer method (RTM) and single-channel algorithms (SCA)), and surface fire detection was done using one method (index method). In the current study, the fire-affected regions from 1989 to 2019 at an interval of 5 years were analyzed to understand coal fire dynamics. The remote sensing data used to determine the surface and sub-surface fire are thermal infrared, near-infrared, and red band data of Landsat-5 and Landsat-8 sensors. In both the methods of sub-surface fire detection, the land surface temperature (LST) was estimated using Landsat data, and fire zones were detected based on the estimated threshold LST values. The threshold LSTs were determined using a statistical method. The study results indicated the land surface temperature (LST), derived from satellite data, was saturated within 55 °C in both the methods (RTM and SCA). Therefore, the surface fire locations with a much higher temperature than 55 °C cannot be distinguished using these methods. Thus, the surface fire estimation was done using the index-based method without estimating the LST. The study results indicated that the locations are situated within the subsurface fire zones in most of the cases and have smaller coverage areas. The LST estimated using Landsat data was validated with field LSTs, measured at 20 different locations using a thermal camera during the same day of satellite data acquisition. The locations of each station were tracked using a GPS device (GARMIN 72S). It was found that the predicted LSTs from satellite data using RTM and SCA method are highly correlated with the field data with a correlation coefficient of 0.91 and 0.94, respectively. The RMSE values between the predicted LST and observed LSTs were 1.29 and 1.63, respectively, for SCA and RTM. The predicted LSTs using two approaches exhibit nearly similar patterns, but the SCA approach offers relatively better accuracy. The results reveal that the coal fire zones estimated using both techniques have good agreement with the field LSTs, and thus either of the methods can be used for fire detection with greater preference to SCA. Thus, SCA was used to analyze the time-series satellite data from 1989 to 2019 at an interval of five years for understanding the coal-fire dynamics in JCF. The study used thermal band data of Landsat-5 and Landsat-8 along with the radiosonde data for estimating the land surface temperature (LST). The results of time-series analysis can be used to determine the extinguished fire zones, new fire zones, and active fire zones. All the maps were stacked in ArcGIS 10.8 software, and an overlay analysis was performed to estimate the propagation characteristics with time. The Landsat data for seven different years (1989, 1994, 1999, 2004, 2009, 2014, and 2019) were downloaded from USGS online portal. The spatial resolutions of the thermal band in Landsat-5 and Landsat-8 data are 120 m and 100 m, respectively but resampled to 30 m spatial resolution. The coal fire maps were generated for seven different years from 1989 to 2019 at five-year intervals. The new fire zones established at every five-year interval from 1989 to 2019 were mapped for understating the fire dynamics. The fire area coverages were found to be 2.026 km2, 3.009 km2, 3.159 km2, 3.991 km2, 4.664 km2, 8.656 km2 and 9.957 km2 respectively for the year 1989, 1994, 1999, 2004, 2009, 2014 and 2019. The study results revealed that the coverage area of coal fire had been continuously increased from 1989 to 2019. Moreover, the results of the temporal analysis of coal fire indicated that the fire propagation was not in any specific direction. That is, new fire zones have been developed pocket-wise due to multiple factors and not due to the continuous burning of in-situ coals. It has been observed that the production level in the mines during this period (1989 to 2019) has also been continuously increased, possibly due to the higher number of working mines. Thus, a higher number of coal blocks are exposed to the atmosphere, which leads to catches of fire due to spontaneous heating. Also, few coal fire pockets have been extinguished with time. The coal fire in a typical zone may be extinguished either due to entire coals being burned or extracted out. The change in fire area coverage of extinguished fire zones and the new fire zones indicate that the formations of new fire zones were always higher than the extinguished fire zones. Thus, the fire area coverage has been continuously increased. The current study also analyzed the spatio-temporal profiles of columnar density of three major greenhouse gases (carbon monoxide (CO), sulphur dioxide (SO2), and nitrogen dioxide (NO2)) over the coal-mining region (JCF, India) using hyperspectral TROPOspheric Monitoring Instrument (TROPOMI) sensor data for the year 2019. The columnar density of the gaseous pollutants in the mining region was also compared with the columnar density of the same over the rural, urban, and forest regions for identifications of the major emission inventories. The results indicated that coal fire is the major source of CO emission over the study region, as the columnar density of CO was high over the fire regions compared to that of the non-fire regions. But the primary source of NO2 is the traffic, as the columnar density of NO2 was high in the city area compared to other areas. The spatial distribution of columnar density of SO2 did not reveal any specific emission sources. However, the results indicated that approximately 7575.6 tonnes of CO and 40.641 tonnes of SO2 had been emitted only from six collieries (Kenuadih, Godhar, Kusunda, Alkusha Ena, and Dhansar) of JCF during 2009 – 2019 based on the assumptions that the entire coal is burned with complete combustion. Thus, the present study demonstrated the potential of satellite remote sensing approaches for monitoring and mapping coal fire dynamics. The TROPOMI onboard Sentinel-5P sensor gridded datasets also aided to analyzed the spatio-temporal pattern of greenhouse gaseous over the coal fire regions. The outcome of the study shall be decisive and key inputs for policy-making, planning, and implementing sustainable development programs over the mining region. The study further assists in identifying the coal fire emission inventories. Moreover, satellite-based earth observation offers information to understand and manage greenhouse gas emissions over a large area.