Physical and Mental Restorative Benefits and Spatial Intervention Mechanisms of Urban Green Spaces Under Low-Altitude 3D Noise Disturbance

Objective Low-altitude cargo drones are now flying over peri-urban green spaces, and their noise profile differs markedly from conventional traffic sound. Unlike road noise, the acoustic energy of multi-rotor UAVs arrives from above, creating a three-dimensional intrusion that existing park design standards have not addressed. Current health assessments of urban green infrastructure still rely primarily on two-dimensional visual metrics such as the Green View Index, tacitly assuming that visible greenery reduces stress in a roughly linear manner. This assumption becomes problematic when overhead mechanical noise enters the scene, because the interaction between visual and auditory stimuli in three-dimensional space is not well understood. Although prior research has separately documented the restorative effects of vegetation views and the annoyance caused by drone noise, few studies have examined their joint influence in real spatial settings. We therefore conducted an empirical study to investigate how micro-scale spatial form, visual greenness, and acoustic exposure together shape autonomic physiology and subjective states during drone overflight. Our broader purpose was to test whether conventional Nature-based Health Services (NBHS) planning remains valid when low-altitude noise is present.

Methods Fieldwork was carried out in Tanzhou, Zhongshan, a peri-urban corridor adjacent to Zhuhai’s low-altitude logistics pilot zone. Twelve micro-scale green spaces were selected to represent a realistic gradient of vegetation density and built-form enclosure, ranging from open lawn to dense canopy. A DJI FlyCart 30 logistics drone served as the noise source, hovering at 40 m, 30 m, and 20 m relative altitude above each measurement point; a no-drone baseline was also recorded. Spatial parameters including distance to the nearest building façade, distance to the nearest road, and building coverage ratio were derived from ArcGIS Pro and high-resolution satellite imagery. We later removed building coverage from the final statistical model because variance inflation factor diagnostics showed unacceptable collinearity with visual greenness. Visual data consisted of 1.6 m eye-level panoramic photographs processed through a ComfyUI workflow with the OneFormer semantic-segmentation model to extract pixel-level GVI. Acoustic metrics—L_Aeq and acoustic kurtosis—were recorded simultaneously with a Class-1 sound level meter at 48 kHz sampling. Thirty healthy volunteers were exposed to these audio-visual scenes under immersive laboratory conditions. Continuous ECG was monitored with a Polar system, and a retrospective time-window method aligned physiological data with specific scene exposures. RMSSD, mean heart rate, and Baevsky's stress index were extracted with the NeuroKit2 framework. Subjective ratings of interest, tension, focused attention, and fear were collected via post-scene questionnaires. All data were analyzed with multilevel linear modeling using Restricted Maximum Likelihood estimation.

Results The analysis revealed that environmental factors shape health outcomes in ways that are not always intuitive. Drone altitude mattered sharply: the initial appearance at 40 m already raised psychological load above the control level, and as the drone descended to 20 m, tension and fear reached their peak while focused attention and recreational interest steadily dropped. Spatial context produced a telling disconnect between what participants felt and what their bodies showed. Greater distance from urban roads lowered self-reported tension, yet it also raised mean heart rate and Stress Index while depressing RMSSD. This indicates that without the masking blanket of continuous traffic noise, impulsive drone intrusions become more acoustically salient and provoke stronger sympathetic activation. Distance from nearby buildings, by contrast, improved physiological relaxation. On the acoustic front, L_Aeq emerged as the dominant stressor, eclipsing the more transient role of acoustic kurtosis, which failed to register a significant main effect in the final regression. Higher L_Aeq consistently eroded recreational interest and attention while amplifying tension and fear. Vegetation provided tangible buffering: a higher GVI was linked to lower mean heart rate, confirming that visual greenness can partly offset physiological stress even under noise. Notably, dense forest canopies reduced subjective tension but produced autonomic co-activation—RMSSD rose alongside a peak Stress Index—and these enclosed spaces also diminished cognitive attention, probably because overhead noise in confined settings heightens a sense of oppressiveness. Open grassy areas with high GVI delivered the best overall physiological resilience.

Conclusion Relying on visual metrics alone falls short when low-altitude drone noise enters the picture. Our results show that perceived calmness and actual physiological relaxation do not necessarily align; quiet green pockets deprived of continuous background noise to mask sudden intrusions can harbour hidden physiological stress. Planning urban Nature-based Health Services (NBHS) sites therefore calls for integrating soundscape design with spatial layout. We argue for extending the classic prospect-refuge theory into three-dimensional settings: landscape configurations should secure visual openness through adequate building setbacks, giving people the capacity to visually register overhead threats without feeling enclosed, while tree canopies supply physical shelter yet retain enough vertical permeability to avoid a boxed-in sensation. Introducing positive natural masking sounds offers a practical way to make up for the loss of traffic-noise masking in deeper green areas and thus safeguard their health benefits. Follow-up work would benefit from testing additional drone models and incorporating further psychoacoustic metrics to refine these design guidelines.