Land, jet stream, and other atmospheric effects on burned area estimation during the South Asian heatwave of 2022

• DOI: 10.1016/j.jag.2024.103720
• Publication Date: 2024-02-24T05:29:56Z
• Authors: Amir Mustofa Irawan; Mercè Vall-llossera; Carlos López-Mart...; Adriano Camps; David Chaparro; Gerard Portal; Miriam Pablos; Alberto Alonso-Go...
• Keywords: Geopotential height; Gradient boosting

Understanding the key variables that characterise fire propagation is important for a better estimation of events and their impacts. This study uses machine learning combined with satellite remote sensing atmospheric modelled data to enhance estimations burned areas. It focuses on intense early summer weather patterns in South Asia during April May 2022 explores relationship between environmental factors spread. The employs various algorithms, including random forest, extra trees, extreme gradient boosting (XGBoost), regressor, support vector regressor neural networks. XGBoost proves be most accurate approach. An isolation forest algorithm used adjust outliers area estimations. comprehensive analysis conducted includes identification sensitivity tests incorporating changes up 25 % natural conditions assess model's consistency. results indicate integrating vegetation, atmospheric, human-related algorithm, outlier adjustments leads effective performance (R2 ≥ 0.7), jet stream enhancing accuracy by approximately 11.5 %. highlights notable impact increases value 300-hPa meridional circulation index flow (MCI300) high 500-hPa geopotential height anomaly (ΔZ500), indicating development strong blocking (upper tropospheric ridge). As compared other factors, e.g. land surface temperature, vapour pressure deficit, soil moisture vegetation optical depth, metrics (MCI300 ΔZ500) was more pronounced, greater sensitivity. These insights emphasise complexity spread, importance using estimate areas, particularly severe heatwaves.