Height and crown area predict biomass differently across urban tree-growth contexts

Accurate estimation of tree aboveground biomass (AGB) is important for urban forest management, green-infrastructure planning, and carbon assessment. We tested 27 height- and crown-area-based candidate equations separately for 45 trees in an urban forest remnant and 50 isolated urban trees, using AGB from the pantropical equation of Chave et al. (2014) as a model-based reference. Candidate models were compared using leave-one-out cross-validation (LOOCV) on the original AGB scale, numerical and biological diagnostics, and a nested LOOCV sensitivity analysis of the selection procedure. The selected models were log1 (M9) for the forest remnant and log4 (M12) for isolated trees. Fixed-model LOOCV yielded RMSE = 0.234 t and predictive R² = 0.689 for the remnant, and RMSE = 0.311 t and predictive R² = 0.932 for isolated trees, without negative predictions. Nested model-selection validation was less stable, particularly when the largest isolated tree was omitted. The equations are therefore context-specific calibration models that require independent validation before spatial extrapolation. For urban forest management, they can support rapid biomass screening where UAV-derived height and crown area are available, but they should not replace locally validated inventories. Supporting data and analysis code: https://doi.org/10.5281/zenodo.22817375