Modelling taper weight and bark allocation along stems of Eucalyptus camaldulensis before and after fire

In forests populated by bark-shedding tree species, bark-to-wood allocation ratio along the stem influences litter deposition, fuel accumulation, and nutrient and carbon dynamics. Although biomass is a more relevant metric for these ecological processes, current taper models, often based on diameter at breast height (DBH), are predominantly volumetric and largely generic. Typically, these models do not directly estimate green weight taper and often fail to incorporate adjustments for bark allocation, which is widely assumed to be an adaptive response to previous fire events. Also, the DBH-dependency of the models limit their applicability in post-logging or post-fire contexts where stems are missing or damaged, and empirical DBH data are inaccessible. Because stumps are often the only intact remnants after such disturbances, their stability makes them a more dependable and feasible biometric alternative for reconstructing stem form and formulating volume or weight taper models. This study shows that stump metrics can account for fire-induced shift in bark allocation while explaining variations in the bark-to-wood weight ratio along the stem in baseline and post-fire recovery forests. Using data from destructively sampled trees in baseline and post-fire regrowth Eucalyptus camaldulensis stands in the Kenyan coastal savanna, we developed weight-explicit taper equations with stump diameter over and under bark (DSoB, DSuB) as predictors. Parsimonious generalized least‑squares (GLS) models and spline-based generalized linear mixed models (GLMMs) were calibrated and validated through cross-validation and analytical diagnostics. All GLS models achieved high predictive accuracy (>84), confirming stump diameter as a robust predictor of segment‑level disaggregated green weight of stem, bridging the methodological gap in forest biomass and carbon assessments where DBH data are unavailable. DSoB and DSuB performed equivalently, supporting their interchangeable use. While GLS models provided ease of predictive use, spline-based GLMMs better captured diameter-dependent site interactions and localized changes in taper shape as well as bark–wood partitioning after fire. Overall, the results demonstrate that stump‑based mass-taper modelling provides a practical and sensitive tool for detecting bark allocation shifts and can be applied in baseline and fire-free plantations to test bark acclimation across regions, improving our ability to quantify bark‑driven litter production, decomposition dynamics, nutrient cycling, and CO₂ exchange.