Optimizing Allometric Equations for Estimating Carbon Storage of Urban Shrubs: A Morphology-Driven Machine Learning Approach and Development of an Intelligent Decision-Support System

With the acceleration of global urbanization, neighborhood green spaces have emerged as important carbon sinks. However, current urban carbon inventories frequently neglect the understory shrub layer owing to morphological heterogeneity and a lack of standardized allometric models. To address this limitation, we analyzed 13 major shrub species (n = 665) through whole-plant excavation. Hierarchical cluster analysis and linear discriminant analysis classified the 13 species into three functional morphological groups based on intrinsic morphological traits (basal stem density, root-to-shoot biomass allocation, and secondary radial growth capacity) (p < 0.001): small shrubs (Type I), large shrubs with high root-to-shoot allocation (Type II), and multi-stemmed sprouting shrubs (Type III). Standard models accurately estimated biomass for Type I species, whereas symbolic regression improved the prediction of the complex non-linear biomass allocation of Type II species. For Type III species, characterized by multi-stemmed growth and anthropogenic management, robust regression provided stable biomass estimates. Gompertz growth models predicted carbon sequestration trajectories, indicating that urban shrubs function as rapid carbon sinks during the early establishment stage. To facilitate practical application, we developed the Urban Forest Carbon Storage Calculator, which integrates Monte Carlo simulation and bootstrapping to generate 95% confidence intervals for species-specific biomass estimation. This study quantifies the overlooked carbon value of the urban shrub layer and provides a morphology-driven methodological approach and a practical tool for sustainable urban forest management.