SBAS-InSAR analysis of regional ground deformation accompanying coal fires in Jharia Coalfield, India

The present study explores a holistic approach toward better assimilation of the contextual relationship between coal fire-induced land subsidence in the Jharia Coalfield (JCF), India. For the process, 31 consecutive Sentinel-1A and Landsat 8 scenes of 2018 were processed to estimate mean Line-of-Sight displacement and Land Surface Temperature (LST) in JCF, respectively. The results indicated that the displacement rate in JCF significantly varies at active mine benches and overburden dump, and high degree of displacement owing to the additive compression inducted along with the volume reduction at the subsurface. The estimated displacement accounts were then spatially correlated with the thermally anomalous pixels to determine the categories of subsidence. Further, the contextual relationship between the displacements estimates (dependent variable) with a set of explanatory variables, i.e. pixel integrated LST was tested using Binary Logistic Regression. The performance of the model was cross-validated using statistical parameters derived from the confusion matrix.

Automated summary

The present study investigates the contextual relationship between coal fire-induced land subsidence in the Jharia Coalfield (JCF), India, using a comprehensive approach. By processing 31 consecutive Sentinel-1A and Landsat 8 scenes of 2018, the mean Line-of-Sight displacement and Land Surface Temperature (LST) in JCF were estimated. The results show significant variation in displacement rate at active mine benches and overburden dump, with high displacement due to additive compression induced by volume reduction at the subsurface. The estimated displacement accounts were spatially correlated with thermally anomalous pixels to determine subsidence categories. Binary Logistic Regression was used to test the contextual relationship between displacement estimates (dependent variable) and a set of explanatory variables, specifically pixel integrated LST. The model's performance was cross-validated using statistical parameters derived from the confusion matrix.