Precise Urban Green Volume-enabled Building and Environment Simulation:
Sub-Meter Voxel Modeling of Airborne and Hand-held 3D Scans of Urban Trees
Qianyun Zhou 1*, Jiajia Wang 2, Bin Chen 3, Fan Xue 4
This is the authors’ version of the paper:
Zhou, Q., Wang, J., Chen, B. & Xue, F. (2022). Precise urban green volumeenabled building and environment simulation: Sub-meter voxel modeling of
airborne and hand-held 3D scans of urban trees. Proceedings of the 27th
International Symposium on Advancement of Construction Management and
Real Estate (CRIOCM2022), pp. 1651–1659, Springer, DOI: 10.1007/978981-99-3626-7_127.
This file is shared for personal and academic use only, under the license CC
BY-NC-ND 4.0 (Non-Commercial, No Derivatives, and with an Attributed
citation when you use). The final published version of this paper can be found at:
https://doi.org/10.1007/978-981-99-3626-7_127.

5

10

15

Abstract: High-rise high-density cities around the world suffer from severe urban heat
island effects. Greenery has the potential to heal urban microclimates, such as shading,
lowered air temperatures, and increased humidity, apart from other benefits to urban
health. Existing numerical simulation studies employing simplified, proxy greenery
models have validated the potential at a macro level; However, the human-centric threedimensional nature of greenery (e.g., tree crown volume, canopy density, and leaf area
index) was ignored, leading to inaccurate results for buildings and blocks, especially in
the high-rise high-density settings. This research proposes a precise voxel modeling of
urban green volume for building and environment simulations. First, two scans, i.e., the
airborne scan of tree canopies and hand-held LiDAR scan of lower parts, are registered
and merged into a voxel model at 0.5-m resolution. Then, simulations and analysis of
the voxelized green volume are conducted using Rhinoceros and Grasshopper. A
preliminary experiment of shading on a university campus in Hong Kong proved the
concept of this study. Future work will be directed to analysis of the voxel resolution
and more types of simulations such as winds and urban thermal exposure.
Keywords: Urban green volume; Urban heat island; Building and environment
simulation; LiDAR point cloud; Voxelization; Numerical simulation

Qianyun Zhou
* Corresponding author, Department of Real Estate & Construction, The university of Hong Kong, China
E-mail: qianyunz@hku.hk

1

Jiajia Wang
Department of Real Estate & Construction, The university of Hong Kong, China

2

Bin Chen
Division of Landscape Architecture, The University of Hong Kong, China

3

Fan Xue
Department of Real Estate & Construction, The university of Hong Kong, China

4

1

1 Introduction
20

25

30

35

40

45

50

The intensive urbanization process and anthropogenic activities have increased the
vulnerability of cities to climate change. The climate hazard is expected to trigger more
frequent natural disasters, such as severe and protracted heat waves, which further harm
economic, social, and environmental perspectives of the built environment [1]. For
intensive land use and high-density cities, the increasing and long-term air temperatures
are frequently associated with intense urban heat island (UHI) effects. UHI not only
leads to heat stress among citizens but also raises urban energy consumption (e.g., air
conditioning in buildings) and has a detrimental influence on the sustainability of the
urban environment [1]. Thus, United Nations’ Sustainable Development Goal-11 (SDG11) calls for actions to resolve sustainability issues for cities and communities.
Existing studies in the literature have confirmed urban greenery’s positive and
negative effects on mitigating UHI. For example, the crown shape and leaf features
have been shown to have cooling potential due to their influence on solar exposure, air
humidity, and wind environment [2]. Morakinyo, et al. [2] found that dense foliage tall
tree species play a significant role in cooling cities through field surveys and model
validation, while Chen, et al. [3] extracted and mapped greening coverages with satellite
remote sensing to analyze the multi-scale difference in greenspace exposure. Yet, Oshio,
et al. [4] showed that urban greenery can significantly reduce wind speeds in streets, thus
hindering the local microclimate. Therefore, holistic and accurate quantitative studies
are demanded to assess three-dimensional greenery.
In recent years, 3D scanning, including airborne drone photogrammetry and mobile
LiDAR, has been widely used for building and city information models [5]. The related
remote sensing technologies obtain the coordinates (X, Y, Z), sometimes with the color
(R, G, B) or laser intensity (I), from the points on the objects. Airborne 3D scan data
was used to assess urban greening and its effects on mitigating UHI [4]. However, the
airborne data that reflects the top parts of tree crowns cannot provide sufficient 3D
vegetation features holistic and accurate quantitative studies.
This study presents a sub-meter voxel modeling method for integrating airborne
and hand-held 3D scans of urban trees for building and environment simulation. The
main contribution of this study is twofold. Firstly, the findings plot a technological
pathway to integrate multi-source 3D scans of urban trees into computable voxels of
3D greenery; Secondly, it showed that the sub-meter green volume model could
improve the accuracy of building and environment simulation at an ignorable cost of
computational load.

2 Methodology
55

60

65

This study employs a point cloud-based scan registration for 3D greenery. First,
photogrammetry of airborne scans generates high-resolution 3D color points composed
of 18,431,241 points. Then, a total of 558,158 points of urban greenery can be collected
from semantic analysis of urban models based on machine learning [5, 6] or by color
filtering and segmentation, which separates target greenery from the surrounding
environment, as shown in Figure 1. Then, handheld LiDAR scanning collects more of
a human perspective on the three-dimensional greening features, with 18,146,233
colorful points. The collected information (e.g., the bottom of the tree crown, tree trunks,
and ground shrubs) aptly complements the airborne scans. Similarly, the points of
interest of 3D greenery can be collected using automatic or manual processing (e.g.,
functions in Cloud Compare). Figure 2 shows an example of pre-processed handheld
laser-scanned greenery data. A complete 3D greenery data registers the airborne and
2

70

handheld scans, as shown in Figure 3. The distance between the overlapping parts of
the two source point clouds is calculated by the distance function in Cloud Compare,
resulting in 0.3702m for average distance and 0.3104 for root-mean-square deviation
(RMSD). The registered 3D greenery data is saved as a point cloud in the LASzip
Compressed Lidar (.las) format, consisting of spatial coordinates (X, Y, and Z), source
of points, and related variables.

Figure 1. Pre-processing of airborne scanning data
75

Figure 2. Pre-processing of handheld scanning data

Figure 3. Registration results of airborne (in dark green) and handheld (in light green) 3D scans
80

85

90

Then, a voxelization modeling method for greening crown shapes is applied to
crown volumes for measurement and visualization. A voxel is a 3D cubic representation
of volume. Poux [7] presents an automation workflow for transforming point clouds
into 3D voxels with Python codes. The voxelization process in this research is based on
three Python (ver. 3.8) libraries: laspy (ver. 2.0.3), open3d (ver. 0.11.2), and NumPy
(ver. 1.21.2). The voxels are assembled from the dataset’s spatial bounding box, and
the voxel location is generated as a relative value to the initial bounding box. The voxel
model can be transformed into other 3D formats, such as .obj format, for further
computation and simulation. Such transformations can be done through the open3d
library. As a result, 3D greenery volume is calculated by voxel quantities, as shown
in Figure 4.

3

Figure 4. Transforming the point cloud of a tree to voxel model

95

100

105

Numerical simulations can finally be conducted using the voxel model. Numerical
simulation is often promising for assessing the influence of urban greening on the
environment. Previous research has utilized environmental models to study the
association between greening and urban air temperatures, local microclimate, and
outdoor thermal comfort [2, 4]. Although parametric simulation tools are becoming
increasingly well-established, the 3D features of greening in urban microclimate studies
were generally replaced by theoretical, empirical values, or oversimplified (e.g., a
conical-shaped tree crown). The complex characteristics of real greenery (e.g., the
density of tree canopy, tree crown volume, and leaf area index) were seldom addressed
in numerical simulations [8]. Therefore, it is essential to represent the urban environment
via a high-precision vegetation modeling and simulation. In addition, the voxel-based
approach allows for the numerical simulation of the urban thermal environment with
changeable resolutions.

3 Preliminary experiment
3.1 Study area and settings
110

115

The preliminary study was established on a university campus (22°16’58.17” N, 114°
8’18.25” E) in sub-tropical Hong Kong, as shown in Figure 5. The city suffers from the
severe UHI effect, while greenery significantly provides outdoor thermal comfort
through shading [2]. Both the airborne and handheld scans were collected around Mong
Kwok Ping Garden by the authors using a DJI Marvic 2 drone (with Pix4D Mapper)
and a Paracosm PX-80 scanner, respectively. The scanned 3D point clouds were preprocessed in CloudCompare (including de-noising, filtering, and segmentation) and
then the composite sources were registered and merged. The two types of point cloud
data were then exported for simulations and further processing.

Figure 5. Location of study area

120

The voxelization process was coded on the Spyder IDE environment, which reads
the full 3D greenery data (.las format). The selection of voxel size significantly impacts
4

125

130

135

140

145

150

the spatial structure and volume of the vegetation (especially tree canopies). In general,
larger voxels produce faster visualization results but only summarize the trees’
structural information. In contrast, smaller voxels can convert more comprehensive
vegetation information but require a longer time [8]. For this experiment, the
transformed voxel model was set up at 0.5m resolution to balance computational time
with the abundance of greenery information, as shown in Figure 6.
Lastly, simulations were based on Rhino/Grasshopper3D, with an open-source
environmental plugin named Ladybug V.1.5.0. The daylighting engine in Ladybug
V.1.5.0 follows ray-tracing algorithm from Radiance and combines imported weather
data files (.epw) to calculate readable plots as well as interactivity data. [9] In the raytracing algorithm for computing daylight, the following integral equation enables
recursive evaluation for testing points on each surface [10]:
2𝜋𝜋 𝜋𝜋
𝐿𝐿𝑟𝑟 �θ𝑟𝑟, ∅𝑟𝑟 � = 𝐿𝐿𝑒𝑒 �θ𝑟𝑟, ∅𝑟𝑟 � + ∫0 ∫0 𝐿𝐿𝑖𝑖 (θ𝑖𝑖 , ∅𝑖𝑖 )ρ𝑏𝑏𝑏𝑏 (θ𝑖𝑖 , ∅𝑖𝑖 ; θ𝑟𝑟 , ∅𝑟𝑟 )|cos θ𝑖𝑖 |sin θ𝑖𝑖 𝑑𝑑θ𝑖𝑖 𝑑𝑑∅𝑖𝑖
(1)
Where:
θ means the surface normal polar angle;
∅ means the surface normal azimuthal angle;
𝐿𝐿𝑒𝑒 �θ𝑟𝑟, ∅𝑟𝑟 � is the emitted radiance;
𝐿𝐿𝑟𝑟 �θ𝑟𝑟, ∅𝑟𝑟 � is the reflected radiance;
𝐿𝐿𝑖𝑖 (θ𝑖𝑖 , ∅𝑖𝑖 ) is the incident radiance;
ρ𝑏𝑏𝑏𝑏 (θ𝑖𝑖 , ∅𝑖𝑖 ; θ𝑟𝑟 , ∅𝑟𝑟 ) is the function for bidirectional reflectance-transmittance distribution.
The greening model was imported into Rhinoceros in .obj format, and the
surrounding
built
environment
was
obtained
from
OpenStreetMap
(https://www.openstreetmap.org) by blender. The daylight experiment started by
importing
both
geometries
and
local
meteorological
data
(https://www.ladybug.tools/epwmap/), and then simulated direct sunlight hours in
Hong Kong during the summer (June to August). It also measured the total sunlight
hours, single simulation time, and green volume under different greening model
accuracy.

3.2 Results
155

Figure 6. Voxelized 3D greenery model of the study area

This experiment compares and analyses the results of five digital modeling simulations:
5

160

165

the model of the built environment without greenery, the oversimplified proxy greenery
model, the airborne/handheld scanned voxelized greenery model, and the hybrid
voxelized greenery model. The numerical simulations were compared with
consistent .epw weather data, the target season (summer), and the surrounding built
environment. Table 1 depicts the simulation results for the five greening precision
models, as well as the greening volume, total sunshine duration, and simulation
completion time. It can be observed that the integrated greenery voxel model
demonstrates the lowest total solar exposure hours, which means that this model
provides the largest area of shade.
Table 1. Comparison of radiation simulation results with varying accuracy models
Type
Experimental Subjects
Green
Simulation Visualization
Volume/m3
1

No greenery model

2

Conventional oversimplified proxy greenery
model

3

4

5

170

The airborne scanned
greenery model
(no handheld, 0.5m
resolution)

The handheld scanned
greenery model
(no airborne, 0.5m
resolution)

The precise urban green
volume model
(This study, 0.5m
resolution)

Total
Radiation/h

Time/
s

-

596.63

40s

3200

579.71

70s

1841

561.46

75s

3022.25

576.24

90s

4863.25

560.38

100s

The experimental results indicated that registering two source point cloud data
increased the overall volume of greenery by 34.2% over the traditional simplified
version of the greening model. Moreover, with the increase of the projected area, the
total radiation duration is correspondingly reduced by 3.33%. This sub-meter voxel
6

175

model collects point cloud data from a multi-directional field of view, which
complements airborne photogrammetry with handheld scanning data, and makes the
numeric model more closely resemble real-world vegetation. This method boosts the
precision of the digital simulation, contributing to further urban microclimate analysis.
Furthermore, the voxelized greenery model offers an automated solution for
quantifying greening volumes.

4 Discussion and future work

180

185

190

195

200

Urban greenery is commonly convinced to relieve urban heat, due to its function of
providing shadows. [2] However, the quantity and density of urban forests can also
impede the urban wind environment (especially in intensive development areas),
causing an impact on the local microclimate [4]. The majority of computational
simulations used for validation employ hypothetical greening models. These digital
representations of vegetation are frequently replaced by empirical values or overly
simple geometries, making the simulation results that do not accurately reflect the real
urban environment [8]. This research proposes a technical solution for integrating
scanned point cloud data from multiple sources, acting as the semantic enrichment of
city information models. [5] Subsequently, a Python-based automated program
efficiently converts the point cloud model into a sub-meter quantifiable voxel model.
This voxel model can contribute to urban greening planning by raising the accuracy of
numerical simulations without increasing the computational load considerably.
Nevertheless, this study also has some limitations. Firstly, the voxelized greening
model was only presented at a resolution of 0.5m in the experiments, thus lacking
comparative analysis of the impact of other resolution values of voxel models on
environmental simulations. The optimal voxel resolution for specific urban
morphologies (e.g., high density, low density, enclosed areas, semi-open areas, open
areas) should be further investigated. Moreover, this study only employed point cloud
processing software to align multi-source scanned data. A more intelligent and
automated technique may be required for future experiments on complex urban
environments to extend the model's applicability. Finally, the initial simulation
validation in this test has focused on the shading effects of urban greenery, and future
work will pay more attention to urban wind simulation (CFD) and outdoor thermal
comfort studies.

5 Conclusion
205

210

215

Urban greenery has a positive impact on improving urban thermal comfort; however,
the volume and density of the tree crown may also affect the urban microclimate. In
recent years, numerical simulations have been frequently employed to validate and
quantify the influence of greenery on urban outdoor comfort and environment, but
lacking highly accurate vegetation models. In this study, the fusion of multi-source
scanned greening point clouds (detailing the three-dimensional attributes of the top and
middle crown layers, as well as the lower shrub vegetation) is proposed and automated
to generate a sub-meter resolution voxelized greening model. This semi-automated
technique enables the reconstruction of the urban forest from an actual physical
environment at a low cost, thereby effectively boosting the accuracy and credibility of
the environmental simulation. Future work includes optimal voxel resolution in urban
environments with different densities and comprehensive urban microclimate
simulations (e.g., CFD wind environment studies).
7

Acknowledgment
The work presented in this paper was supported by the Hong Kong Research Grants
Council (RGC) (Grant No. T22-504/21-R).
220

225

230

235

240

245

250

References
[1]C. Rosenzweig, W. D. Solecki, S. A. Hammer, and S. Mehrotra, Climate change and cities: First
assessment report of the urban climate change research network. Cambridge University Press,
2011.
[2]T. E. Morakinyo, K. K.-L. Lau, C. Ren, and E. Ng, "Performance of Hong Kong's common trees
species for outdoor temperature regulation, thermal comfort and energy saving," Building and
Environment,
vol.
137,
pp.
157-170,
2018/06/01/
2018,
doi:
https://doi.org/10.1016/j.buildenv.2018.04.012 .
[3]B. Chen et al., "Beyond green environments: Multi-scale difference in human exposure to
greenspace in China," Environment International, vol. 166, p. 107348, 2022/08/01/ 2022, doi:
https://doi.org/10.1016/j.envint.2022.107348 .
[4]H. Oshio, T. Kiyono, and T. Asawa, "Numerical simulation of the nocturnal cooling effect of
urban trees considering the leaf area density distribution," Urban Forestry & Urban Greening,
vol. 66, p. 127391, 2021/12/01/ 2021, doi: https://doi.org/10.1016/j.ufug.2021.127391 .
[5]F. Xue, L. Wu, and W. Lu, "Semantic enrichment of building and city information models: A tenyear review," Advanced Engineering Informatics, vol. 47, p. 101245, 2021/01/01/ 2021, doi:
https://doi.org/10.1016/j.aei.2020.101245 .
[6]M. Li, F. Xue, Y. Wu, and A. G. O. Yeh, "A room with a view: Automatic assessment of window
views for high-rise high-density areas using City Information Models and deep transfer
learning," Landscape and Urban Planning, vol. 226, p. 104505, 2022/10/01/ 2022, doi:
https://doi.org/10.1016/j.landurbplan.2022.104505 .
[7]F. Poux and R. Billen, "Voxel-based 3D Point Cloud Semantic Segmentation: Unsupervised
Geometric and Relationship Featuring vs Deep Learning Methods," ISPRS International
Journal of Geo-Information, vol. 8, no. 5, p. 213, 2019. [Online]. Available:
https://www.mdpi.com/2220-9964/8/5/213 .
[8]H. Xu, C. C. Wang, X. Shen, and S. Zlatanova, "3D tree reconstruction in support of urban
microclimate simulation: A comprehensive literature review," Buildings, vol. 11, no. 9, p. 417,
2021.
[9]M. Pak, A. Smith, and G. Gill, "Ladybug: A Parametric Environmental Plugin For Grasshopper
To Help Designers Create An Environmentally-conscious Design," Building Simulation
Conference Proceedings, 2013.
[10]G. J. Ward, "The RADIANCE lighting simulation and rendering system," in Proceedings of the
21st annual conference on Computer graphics and interactive techniques, 1994, pp. 459-472.

8