Résumé
Sedimentary charcoal elongation is increasingly being used in paleoecology to distinguish herbaceous from woody fuel in past fires. However, the relationship between charcoal morphotypes and plant types has never been formally tested in tropical environments, despite its potential to improve understanding of fire regimes and deforestation, and to analyze long-term ecosystem resilience. Separation between herbaceous and woody charcoal commonly relies on a fixed elongation threshold, but this arbitrary division overlooks the morphological continuum of charcoal and may bias interpretations. To address this, we developed a Bayesian model to classify charcoal fragments into herbaceous and woody subpopulations based on their elongation distributions. We applied it to 19 sediment charcoal records spanning the last 30 years across a broad geographic and vegetation gradient in West and Central Africa. Compared with contemporary vegetation cover derived from remote sensing, the estimated proportion of woody charcoal was significantly positively correlated with tree cover, validating our approach and confirming charcoal elongation as a reliable proxy for past fire-related biomass sources. Finally, we applied the model to a paleosequence from Lake Sélé (Benin) to reconstruct temporal changes in biomass and compared results with fossil pollen data, further assessing the proxy relevance for reconstructing past fire regimes and advancing tropical paleoecology.