A Study on Automated Detection of Surface and Sub-Surface Coal Seam Fires Using Isolation Forest from Landsat 8 OLI/TIRS Images

Coal seam fire (CSF) affects various social, economical, and environmental aspects. Surface and sub-surface CSF are detected in the past by studying the anomalies of land surface temperature patterns. Most of them use manual observations and thresholds to separate CSF. Such techniques are also susceptible to other regions, which have high temperature and high mineral content. The proposed work presents a novel technique to detect surface and sub-surface CSF by treating them as outliers. It provides a multi-level anomaly detection technique using isolation forest, which is robust to the presence of other high temperature and high mineral content regions over the seasons.