Subsurface and surface coalfires are a serious problem in many coal-producing countries. Combustion can occur within the coal seams (underground or surface), in piles of stored coal, or in spoil dumps at the surface. While consuming a non renewable energy source, coalfires promote several environmental problems. Among all GHGs that are emitted from coalfires, CO2 is the most significant because of its high quantity. In connection to this environmental problem, the core aim of the present research is to develop a hyperspectral remote sensing and radiative transfer based model that is able to estimate CO2 concentration (ppmv) from coalfires. Since 1960s remote sensing is being used as a tool to detect and monitoring coalfires. With time, remote sensing has proven a reliable tool to identify and monitor coalfires. In the present study multi-temporal, multi-sensor and multi-spectral thermal remote sensing data are being used to detect and monitor coalfires. Unlike the earlier studies, the present study explores the possibilities of satellite derived emissivity to detect and monitor coalfires. Two methods of emissivity extraction from satellite data were tested, namely NDVI derived and TES (Temperature emissivity separation) in two study areas situated in India and China and it was observed that the satellite derived emissivity offers a better kinetic surface temperature of the surface to understand the spread and extent of the coalfires more effectively. In order to reduce coalfire related GHG emissions and to achieve more effective fire fighting plans it is crucial to understand the dynamics of coalfire. Multitemporal spaceborne remote sensing data can be used to study the migration and expresses the results as vectors, indicating direction and speed of migration. The present study proposes a 2D model that recognizes an initiation point of coalfire from thermal remote sensing data and considers local geological settings to predict the speed and future location of coalfires. It was observed that the model can predict the future location of coalfires with a predefined time period. However, few uncertainties (e.g. abrupt climatic change) can not be taken account in this model. To explore the sensitivity of present hyperspectral sensors with different atmospheric CO2 concentrations, additive and multiplicative noise were introduced in FASCOD simulated spectra and evaluated. A comparison among the present available hyperspectral sensors was made to find out the most suitable remote sensing sensor for CO2 quantification. To achieve the core research objective, firstly, a band ratioing method was used for column atmospheric retrieval of CO2 and secondly atmospheric models were simulated in FASCOD to understand the local radiation transport and then the model was implemented with the inputs from hyperspectral remote sensing data. Both methods (band ratioing and radiative transfer based) were tested in a coalfire affected area in northern China. It was observed that retrieval of columnar abundance of CO2 with the band ratioing method is faster as less simulation is required in FASCOD. Alternatively, the inversion model could retrieve CO2 concentration from a (certain) source because it excludes the uncertainties in the higher altitude.
