Objective  With the development of city information modeling, digital twin, smart city and other relevant technologies, traditional 2D-based planning progress are now being transformed into "panoramic" territorial spatial planning supported by 3D technology. Under the digitalization trend of landscape planning, design and management practice, digital information model has become a necessity for the characterization of 3D green space. Unlike the modeling of buildings or urban furniture, plant modeling can be hardly achieved by traditional modeling methods due to the complex branching structure and enormous details of plants and such limitations as low fidelity and low efficiency of the widely used polygon mesh model in plant simulation. As a generative algorithm, the L-system algorithm can be used to rapidly construct different vegetation models based on the branching characteristics of plants and the self-similarity of plant morphology. Based on the L-system algorithm, this research proposes a method to quickly construct detailed 3D plant models for visualization and quantitative analysis. Taking the Modern Agricultural Research Institute of Huazhong Agricultural University in Xiangyang, Hubei Province as the case study site (hereinafter referred to as "the Site"), the research, through testing the rules and attributes of the L-system algorithm, constructs parametric vegetation models based on the branching structures and morphological characteristics of 24 species of plants.

Methods  To visualize and simulate plants in the Site, the research adopts Houdini as the L-system modeling platform to generate a parametric branching model for these plants. The research derives such information as site terrain, building entities and plant geometry parameters from relevant site planting documents, and selects 4 basic plant branching structures and 12 kinds of parametric leaf models for different tree species in combination with their respective structures. After that, the research, based on the aforesaid site planting documents, sets 24 different L-system algorithm plant models to fill the planting area of the terrain model. By adjusting the iteration parameters of the L-system algorithm, the research simulates different growth stages of plants in the Site respectively at the initial operation and plant maturity stages thereof, with the parameters of relevant vegetation models at each stage being selected from the plant specifications in the plant construction maps and the i-tree plant database respectively. The voxel model is a geometric model designed to represent the spatial volume of a plant by voxels, which is well suitable for quantitative analysis. Then the research adopts parametric tools in Houdini to transform each vegetation model generated by the L-system algorithm into a point cloud, and transform the point cloud into a voxel vegetation model with the size of each voxel unit set to 0.2 m × 0.2 m × 0.2 m. The 3D green quantity is estimated by the method of volumetric accumulation of the voxel units.

Results  To describe and evaluate the 3D green space, the research proposes the following four different 3D green quantity evaluation indicators based on traditional 2D green space evaluation indicators: total 3D green quantity, 3D green space ratio, 3D green quantity per unit area and 3D green quantity per capita. Then the research respectively calculates the 2D and 3D indicators of the Site at the two stages for comparative analysis. The results show that at the initial operation stage, the total 3D green quantity of the Site is 13, 497.08 m³, and 3D green quantity per unit area is 12.25 m³/m². At the plant maturity stage, the total 3D green quantity can reach 215, 097.42 m³, and the 3D green quantity per unit area can reach 195.00 m³/m².

Conclusion  The vegetation models generated by the L-system algorithm are very effective and adaptable in the visualization of branching structure. The L-system algorithm can also be used to construct models for plants with different growth periods in combination with the plant growth curve. Compared with traditional models, the plant models generated by the L-system algorithm have a lower demand for local storage. L-system plant models can, through parameter adjustment, be set to different detail levels to suit the modeling needs of different scenarios. In the planning and design practice, the vegetation models based on the L-system algorithm can be used to solve such problems as inconsistency in plant modeling, complexity and confusion of model data, and lack of a universal method for expressing detail levels, and can also be used as a plant database for interpreting and constructing city information models, simulating spatial-ecological interactions, and evaluating urban ecosystem services, so as to support the workflow of digital planning and design. However, there are some limitations in the current L-system algorithm. For example, the model based on the L-system algorithm takes no account of real-world climate conditions, growth competition between plants, and other environmental influencing factors. Future research may integrate the above-mentioned factors into the L-system algorithm to construct better tree growth simulation models.