UPCommons. Portal del coneixement obert de la UPC
Publicación
Contribución de congreso (ConferenceOutput)
Unraveling fire dynamics with EO data: Causal modeling of burned areas in Siberia and Southern-Middle Africa
Understanding the biosphere-atmosphere interactions across multiple scales is critical for Understanding global burned area processes. This requires the interconnection of global factors, such as ocean climate indices (OCIs) and upper tropospheric conditions, with local variables, including soil moisture, surface dryness, and vegetation indices. This study identifies causal relationships among all variables in various Earth Observation (EO) datasets using causal networks constructed with the Peter-Clark algorithm and Momentary Conditional Independence (PCMCI). These methods effectively address nonstationary processes, including periodic variations and heteroscedasticity. This technique has been applied to boreal and tropical forests in two key regions: Siberia (Sib) and Southern-Middle Africa (SMAf). The results demonstrate that anomalies in soil moisture (△SM), vapor pressure deficit (△VPD), and vegetation indices (△ NDVI) are significant contributors to burned areas in both locations. In addition, cascading influences from global drivers, such as OCIs, and upper tropospheric factors, including ΔZ500, v300, and the jet core, have varying impacts on local variables and burned area outcomes across Sib and SMAf. Notably, a distinct jet stream pattern emerges on days when fire spread events occur in Sib, but it does not significantly affect the SMAf area. Composite maps reveal the formation of a jet stream trough that triggers heat dome entrapment over the Sib burned areas, with the dissipation occurring within 8 days after the fire events. By emphasizing meaningful causal relationships, this approach surpasses traditional correlation-based methods, improving terrestrial ecosystem modeling and enhancing wildfire forecasting systems., The project that gave rise to these results received the support of a fellowship from the ”la Caixa” Foundation (ID 100010434), with fellowship code LCF/BQ/DI21/11860028. This work was also supported by Project PID2023- 149659OB-C22 funded by MCIU /AEI /10.13039/501100011033 / FEDER, UE., Peer Reviewed