Global wildfire dynamics in the Anthropocene: Statistical characterisation of burned-area distributions and its consequences for environmental risk assessment and Carbon budgets
Science of the Total Environment, cilt.1046, 2026 (SCI-Expanded, Scopus)
- Yayın Türü: Makale / Tam Makale
- Cilt numarası: 1046
- Basım Tarihi: 2026
- Doi Numarası: 10.1016/j.scitotenv.2026.181964
- Dergi Adı: Science of the Total Environment
- Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus, BIOSIS, Chemical Abstracts Core, Chimica, Compendex, EMBASE, Geobase, MEDLINE
- Anahtar Kelimeler: Burned area, Carbon budget, Extreme events, Heavy-tailed statistics, Lognormal distribution, MODIS, Power-law, Wildfire size distribution
- Hatay Mustafa Kemal Üniversitesi Adresli: Evet
Özet
Wildfires are dominant drivers of biosphere–atmosphere exchange and ecosystem turnover, yet the statistical frameworks underpinning fire-risk assessment commonly assume power-law (scale-invariant) distributions for fire sizes. Here we examine whether this assumption holds for fire cluster area—the contiguous burned-area extent of individual fire events, measured in MODIS 500-m pixels (1 pixel ≈ 21.5 ha) and serving as our primary metric of event magnitude. We fitted discrete power-law and truncated lognormal distributions to 23 years (2001–2023) of MODIS MCD64A1 Collection 6.1 burned-area data across four sentinel regions representing Mediterranean (California, Greece, Portugal) and Boreal (Central Siberia) biomes. Of 92 region–year combinations, 81 (88%) provided sufficient tail samples (n≥50 fire clusters above the lower fitting threshold xmin) for robust inference. In every valid case, the lognormal distribution significantly outperformed the power-law model: the mean log-likelihood ratio was R=12.8±8.2 (all Vuong p<0.001; mean ΔAIC=24.8±16.4). Despite this lognormal preference, fitted power-law exponents remained in the heavy-tailed regime in 98.8% of cases (α<2; mean α=1.73±0.11), confirming that catastrophically large fire clusters dominate total burned area. No significant temporal trend in tail exponents was detected over the study period (p=0.42), suggesting that the statistical mechanics of fire-cluster growth have remained stable even as overall fire activity has intensified. A counterfactual simulation using Central Siberian parameters quantified the “inflationary bias” introduced by unconstrained power-law extrapolation: at theoretical continental scales, power-law models overestimate cumulative burned area by a factor of ≈8.3 relative to a physically constrained lognormal model. These results call for the adoption of lognormal-based frameworks in environmental risk assessment and carbon-budget modelling for fire-prone biomes.