Extreme heat reshapes urban green space vitality: Dynamic response typologies, nonlinear built-environment associations, and configuration patterns in Nanjing

Source abstract

Urban green spaces (UGS) are critical for climate adaptation, yet it remains unclear how different UGS change their vitality under extreme heat relative to optimal-temperature conditions and how such changes are associated with nonlinear built-environment relationships. Using Nanjing as a case study, this study combined PCA–GMM clustering and XGBoost–SHAP explainable machine learning to identify dynamic response typologies from EH–OT 48-hour vitality-difference curves and compare the importance, nonlinear thresholds, marginal effects, and configuration patterns of built-environment features across scenario–day-type subsets and dynamic response typologies. Compared with the OT scenario, overall UGS vitality under the EH scenario was 20.38% lower, but the relative changes varied substantially across UGS. Two dynamic response typologies were identified: Heat-Enhanced Dominant and Heat-Inhibited Dominant UGS. Area showed the highest importance in most scenarios, with its importance ranging from 11.4% to 47.1%. Feature-importance reconfiguration was more pronounced in Heat-Enhanced Dominant UGS under the EH scenario, whereas Heat-Inhibited Dominant UGS showed a more stable importance structure. Single-feature nonlinear trends were generally consistent, whereas typological differences were clearer in multi-feature configurations. For example, Heat-Enhanced Dominant UGS were more often associated with high NDVI (0.47–0.70) and high population density (3.53–26.12), whereas Heat-Inhibited Dominant UGS were more often linked to low population density (0–2.12) and high SVF (0.61–0.68). These findings suggest that changes in UGS vitality under extreme heat should be understood through EH–OT dynamic response typologies, nonlinear threshold shifts, and built-environment configuration patterns. This study provides empirical evidence for type-specific and climate-adaptive UGS planning under increasingly frequent extreme heat.