Effects of urban green-space landscape patterns on cooling services across four Chinese case cities

Source abstract

Urban green spaces (UGS) are critical for mitigating heat islands, and their cooling performance varies across climatic conditions and spatial contexts. However, how UGS landscape patterns shape different cooling dimensions across climatic contexts remains insufficiently understood. Here, we construct a multidimensional assessment framework integrating cooling efficiency (CE), cooling capacity (CC) and cooling benefit (CB) to evaluate UGS cooling effects across four Chinese cities representing distinct climatic zones (Changchun, Beijing, Hangzhou, and Haikou). We employed an interpretable machine-learning framework (XGBoost with SHAP) to quantify the influence of landscape configuration and environmental factors on these cooling dimensions. Our models showed robust predictive performance, particularly for cooling capacity (CC), with R² reaching 0.881 in Beijing and 0.789 in Hangzhou. Cooling efficiency (CE) exhibited a clear latitudinal gradient, whereas CC varied substantially among cities, from 0.8–4.1°C in Changchun to 2.2–9.2°C in Beijing. Landscape configuration effects were strongly nonlinear: edge contrast (ECON) emerged as a key predictor of CC, and lower ECON generally favored stronger city-scale cooling. However, a trade-off emerged for cooling benefit (CB): while lower ECON enhanced overall cooling capacity, localized high-ECON patches acted as “cool oases” in densely populated areas. Moreover, anthropogenic factors, particularly atmospheric CO₂ concentration, contributed substantially to CB variation, highlighting a pronounced spatial mismatch between biophysical cooling supply and population demand. These findings challenge uniform greening policies and suggest that citywide cooling capacity and localized cooling benefit may require different, climate-sensitive landscape strategies.