Beyond the horizon: Unveiling the impact of 3D urban landscapes on residents’ perceptions through machine learning

• Moderate 3D volume and patch density maximize urban vitality. • Human-scaled bulk and extreme fragmentation regimes enhance security. • Landmark heights and low interface complexity drive aesthetic appeal. • 3D metrics far outperform 2D indices in predicting perceptions. • Functional diversity with moderate fragmentation fosters dynamic, engaging environments. With burgeoning urban populations and mounting land constraints, vertical urbanization has become indispensable for sustainable city growth. Yet the perceptual implications of three‐dimensional (3D) form remain underexplored. In this study, we combine ResNet50‐based semantic segmentation of street‐view imagery with an XGBoost regression framework interpreted via SHAP values to quantify how 3D landscape morphology shapes residents’ perceptions of vitality, security, and aesthetics. Our model achieves R2 values above 0.85 for all perceptual dimensions, substantially outperforming analogous 2D‐only approaches. We find that urban vibrancy is maximized when total 3D volume (All_3DArea) falls between 1.0 and 3.0 × 10⁶ m 3 , patch density (PD_All_3D) ranges from 5.0 × 10 −5 to 2.0 × 10 −4 , and vertical relief (RMSofHeight) remains under 10 m. Security perception peaks in human‐scaled contexts (All_3DArea < 5.0 × 10 5 ), at very low or very high fragmentation (PD_All_3D < 5.0 × 10 −5 or > 2.5 × 10 −4 ), and with moderate summit clustering (Dsummit ≈ 0.005–0.01) and gentle curvature (SummitCur < 1.0). Aesthetic appeal corresponds most strongly to landmark elevations (TenptHeight ≈ 350–400 and near 500), subtle interface complexity (S_dr < 0.05), intermediate fragmentation (PD_All_3D ≈ 1.0–1.2 × 10 −4 ), and summit density (Dsummit ≈ 0.005–0.008). By uncovering these nonlinear “sweet spots,” our findings provide precise, human‐centered guidelines for 3D urban design—informing form‐based regulations, mixed‐use zoning, and topographic modulation to foster cohesive, vibrant, and beautiful cityscapes.

Authors

Institutions

Publication Details

Journal
Ecological Indicators
Published
2025-06-13
DOI
https://doi.org/10.1016/j.ecolind.2025.113736
Citations
5
Primary Topic
Urban Green Space and Health
Type
article
Field-Weighted Citation Impact
6.00

Funders

Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
article

Beyond the horizon: Unveiling the impact of 3D urban landscapes on residents’ perceptions through machine learning

Zhiyong Ding, Muwei Li, Jianping Gu, Qing Lu
5 citations
Ecological Indicators
Urban Green Space and Health
6.00
article

Beyond the horizon: Unveiling the impact of 3D urban landscapes on residents’ perceptions through machine learning

Zhiyong Ding, Muwei Li, Jianping Gu, Qing Lu
article en
5 citations

Abstract

• Moderate 3D volume and patch density maximize urban vitality. • Human-scaled bulk and extreme fragmentation regimes enhance security. • Landmark heights and low interface complexity drive aesthetic appeal. • 3D metrics far outperform 2D indices in predicting perceptions. • Functional diversity with moderate fragmentation fosters dynamic, engaging environments. With burgeoning urban populations and mounting land constraints, vertical urbanization has become indispensable for sustainable city growth. Yet the perceptual implications of three‐dimensional (3D) form remain underexplored. In this study, we combine ResNet50‐based semantic segmentation of street‐view imagery with an XGBoost regression framework interpreted via SHAP values to quantify how 3D landscape morphology shapes residents’ perceptions of vitality, security, and aesthetics. Our model achieves R2 values above 0.85 for all perceptual dimensions, substantially outperforming analogous 2D‐only approaches. We find that urban vibrancy is maximized when total 3D volume (All_3DArea) falls between 1.0 and 3.0 × 10⁶ m 3 , patch density (PD_All_3D) ranges from 5.0 × 10 −5 to 2.0 × 10 −4 , and vertical relief (RMSofHeight) remains under 10 m. Security perception peaks in human‐scaled contexts (All_3DArea < 5.0 × 10 5 ), at very low or very high fragmentation (PD_All_3D < 5.0 × 10 −5 or > 2.5 × 10 −4 ), and with moderate summit clustering (Dsummit ≈ 0.005–0.01) and gentle curvature (SummitCur < 1.0). Aesthetic appeal corresponds most strongly to landmark elevations (TenptHeight ≈ 350–400 and near 500), subtle interface complexity (S_dr < 0.05), intermediate fragmentation (PD_All_3D ≈ 1.0–1.2 × 10 −4 ), and summit density (Dsummit ≈ 0.005–0.008). By uncovering these nonlinear “sweet spots,” our findings provide precise, human‐centered guidelines for 3D urban design—informing form‐based regulations, mixed‐use zoning, and topographic modulation to foster cohesive, vibrant, and beautiful cityscapes.

Ecological IndicatorsVol. 177
Chongqing University (CN), Chinese Research Academy of Environmental Sciences (CN)
National Natural Science Foundation of China, Chongqing Social Science Planning Fund
Sustainable cities and communities
Openalex Percentile: Top 4%
Urban Green Space and Health
6.00
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.