A new achievement of satellite-based gas flaring volume estimation: decision tree modeling

Gas flaring (GF) is a long-term issue in the oil/gas industries and has a critical effect on the environment. In the last decade, remote sensing technology has shown resounding capabilities to detect and characterize GF. Iran has many natural oil/gas processing plants and petrochemical companies that are located in the southern regions. The main goal of this research is estimation of the volume of GF for two years (2018–2019) by day/nighttime radiation and air pollutant data. We used Decision Tree modeling/Exhaustive CHAID (Chi-squared Automatic Interaction Detector) based on remote sensing data such as shortwave infrared (SWIR) and thermal infrared (TIR) of Landsat 8/ M10 of VIIRS (Visible Infrared Imaging Radiometer Suite) / air pollutants of TROPOMI (Tropospheric Monitoring Instrument) in three types of models. Results showed that R² values for model 1 (based on all variables/SWIR, TIR, Pollution products), model 2 (based on SWIR bands and pollution data), and model 3 (based on SWIR and TIR bands) is 0.52, 0.50, and 0.51, respectively. The results of sensitivity analysis showed that the shortwave infrared band for two sensors OLI (Operational Land Imager) /VIIRS (Visible Infrared Imaging Radiometer Suite) had the most important role in the estimation of gas flaring volume. The valuable findings of this research represent the important effect of the shortwave infrared bands of the sensors in estimating the GF volume at the local/global scale by hierarchical decision scheme modeling.