Wind is a significant natural disturbance factor affecting many forested regions globally, with tree stability during storms being a critical issue in urban forested areas. Urban forests, integral to urban ecosystems, are particularly vulnerable to frequent coastal storms, which induced substantial economic losses and threaten ecological balance and human safety. Therefore, developing storm-resistant urban forests necessitates a deeper understanding of tree aerodynamics to mitigate risks and ensure sustainable urban development. Previous research has primarily focused on analyzing the aerodynamic characteristics of single tree/tree species or establishing empirical models based on limited experimental data, which often lack generalizability. Even within the same tree species, the aerodynamics of different trees can vary significantly due to differences in tree morphology. Therefore, a comprehensive analysis of the impact of morphology on the aerodynamic characteristics of trees is essential to properly plan urban greening to build a resilient forest. Based on this analysis, constructing tree reconfiguration models can also aid researchers in understanding the ability of trees to withstand wind storms. Furthermore, developing wind-induced load models can facilitate the rapid prediction of tree failure, providing valuable insights for early warning assessments of urban forests before suffering from wind storms. This research investigated the effect of tree morphology on the aerodynamics of sympodial trees from macroscopic and microscopic ways through wind tunnel tests, which laid the foundation on the developed tree reconfiguration model and deep learning model for predicting the drag coefficient of urban trees. First, the effect of different crown configuration on the reconfiguration ability of sympodial trees is macroscopically investigated through wind tunnel tests using natural trees. Then, a modelling technique in simulating the aerodynamics of natural tree is proposed and validated. Subsequently, the effect of microscopic tree parameters, i.e. branch angle and branch orders, on the aerodynamics of sympodial trees is accessed based on the experimental data of artificial sympodial trees. Meanwhile, the vibration characteristics of sympodial tree is deeply explored based on the experimental data of natural trees. Finally, a reconfiguration model and a deep learning model predicting the drag coefficient of sympodial trees are proposed based artificial intelligence techniques. The wind-induced reconfiguration performance of 9 sympodial trees with three different pruning techniques is examined by wind tunnel tests of natural trees, macroscopically and comprehensively revealing the effects of crown configurations on tree reconfiguration. The area reduction effect decreases with an increase in LAI
1 (the ratio of leave area to ground area) and LAI
2 (the ratio of leave area to frontal area). Streamlining effect is gradually enhanced with an increase in λct (the ratio of crown height to stem height), but shows a trend of first increasing and then decreasing with the increase of LAI
1 and LAI
2. To investigate the effect of microscopic tree parameters, i.e. branch angle and branch orders, on the aerodynamics of sympodial trees in controlled conditions, artificial sympodial tree model is established using 3D printing technique and artificial leaves. The drag coefficient and overturning moment coefficient of artificial tree model match well with those of natural trees. Meanwhile, the reconfiguration ability of artificial tree model can also provide reasonable variations compared with natural trees. Finally, the key points in modelling the aerodynamics of the natural tree are pointed out. Based on the modelling technique of artificial tree model, 1-order and 3-order sympodial trees with differing branch angle are established to investigate the influence of microscopic tree parameters on tree aerodynamics. Tree vibration experiences unstable oscillations induced by irregular leaf motion with an increase of wind speed, resulting in a rise in drag coefficient. Lower branch angles enhance the resistance for stem breakage, while higher branch angles increase resistance to tree uprooting. Meanwhile, despite higher drag force observed in trees with higher branch angles at wind speeds below 20 m/s, they exhibit superior reconfiguration capabilities, enabling them to withstand stronger winds effectively. Subsequently, higher branch orders have been confirmed to be more advantageous for the wind resistance of trees. The effect of extreme gust wind on tree aerodynamics is also investigated, resulting in a 30% increase of wind-induced loads on trees. Finally, a reconfiguration process for sympodial trees is proposed. The research on the vibration characteristics of sympodial tree, which exhibits extremely complex structures, is limited. Wind tunnel experimental data of natural trees is adopted to systematically investigate vibration characteristics of natural trees with different crown configurations. Results show that as the new Cauchy number increases, the base overturning moment coefficient (CM) gradually reduces, yet its reduction rate decreases. Subsequently, a higher spectral peak and narrower bandwidth of CM is observed with the extent of pruning severity. Finally, an energy coefficient (CE) is developed to depict the wind energy absorption ability of trees. It is revealed that CE displays a declining trend with an increase in LAI
1 and LAI
2 . Furthermore, it is found that trees characterized by robust and squat stems along with sparse leaves possess an enhanced capacity for wind energy absorption. Trees possess an inherent reconfiguration capability, enabling them to adjust their morphology to reduce the wind loads. To better understand the ability to adapt to wind storms of sympodial trees, a cumulative distribution function of Cauchy distribution combined with particle swarm optimization (PSO) is employed to develop a reconfiguration model of sympodial trees. A scaled Cauchy number characterizing the deformation ability of sympodial tree model is also proposed. Finally, a deep learning model is established to satisfactorily predict the drag coefficient based on the visual captures acquired by cameras, which can be combined with real-time meteorological data to enable early warning assessment of disasters in urban forests.
| Date of Award | 2024 |
|---|
| Original language | English |
|---|
| Awarding Institution | - The Hong Kong University of Science and Technology
|
|---|
| Supervisor | Anthony LEUNG (Supervisor) & Tim K.t. Tse (Supervisor) |
|---|