A low-density tree digital twin model for refined urban greening management: a case study of tree wind disaster risk management

Objective Traditional idealized tree models struggle to accurately represent the real shape of trees. This study proposes a method to generate a refined tree digital twin model, providing an efficient and accurate model support for the application of smart city technologies in urban greening management.

Method This study constructed a refined digital twin model for 16 Chinese scholar trees (Sophora japonica) located beside a university building in Beijing. First, the single tree point cloud was segmented by cyclically applying the DBSCAN algorithm. Second, the α-shape algorithm of 2D point clouds, combined with the idea of dimensionality reduction and augmentation, was used to extract the outer contour of 3D tree point clouds. Thirdly, the 3D mesh of real tree was generated using up-sampling nearest interpolation method, and then the model smoothing was carried out by Grasshopper. Finally, model parameters were added on the Revit platform to complete the creation of digital twin model. The model quality was assessed in terms of tree morphology parameter reproducibility and point cloud fit using CloudCompare software. Additionally, the performance of model in CFD wind simulation was compared with that of an idealized tree digital model using Phoenics software.

Result The average error between digital twins of 16 diverse Chinese scholar trees and the point clouds was −3.06 cm, indicating that the model can accurately preserve the unique morphological characteristics of trees. Under strong wind conditions, the wind speed simulation results of the two models showed significant differences, with a maximum discrepancy of 97.23%.

Conclusion The model provides an effective decision-making tool for the refined management of urban greening, and helps to expand the application scenarios of smart city technologies in urban greening management. However, due to the limitations of DBSCAN point cloud segmentation algorithm, the method in this study is mainly applicable to modeling low-density tree communities. For high-density plant communities with a high degree of canopy overlap, new modeling methods need to be explored to further optimize the applicability of the model.