Global Wildfire Dynamics in the Anthropocene: Beyond the Power-Law Illusion and Its Consequences for Environmental Risk Assessment and Carbon Budgets

Description / Abstract: This repository contains the source code, processed datasets, and supplementary materials required to replicate the analysis and figures presented in the manuscript "Heavy Tailed Lognormal Dynamics in Global Wildfires:Reevaluating the Power Law Assumption and Its Implications for Extreme‑Event Risk" submitted to Remote Sensing of Environment. Overview: This study analyzes 23 years (2001–2023) of global wildfire size distributions using MODIS MCD64A1 burned area data. The codebase implements rigorous statistical frameworks to compare Power-law and Lognormal distributions across Mediterranean (California, Greece, Portugal) and Boreal (Central Siberia) fire regimes. Contents: Source Code: Python scripts for fire cluster extraction, discrete Maximum Likelihood Estimation (MLE) of distribution parameters, Vuong’s likelihood ratio tests, and information-theoretic model selection (AIC/BIC). Processed Data: Extracted fire cluster sizes (in pixels and hectares) for all 92 region-year combinations analyzed in the study. Visualization: Scripts to generate the CCDF plots, sensitivity analysis figures, and decision matrices used in the manuscript. Methodological Notes: The analysis relies on the discrete (Zipf) power-law form and truncated lognormal distributions. Parameter estimation uses numerical optimization with the mpmath library for high-precision evaluation of the Hurwitz zeta function. Original MODIS MCD64A1 data were accessed via the Microsoft Planetary Computer STAC API. Due to size constraints, raw raster files are not included; however, the extraction scripts provided allow for reconstruction of the dataset from the original source. Keywords: Wildfire, Fire size distribution, Power-law, Lognormal, MODIS, Burned area, Python, Statistical inference, Remote Sensing. License: Creative Commons Attribution 4.0 International (for data) MIT License (for code)