Scale-dependent shifts in urban park cooling performance, spatial associations, and area thresholds

Urban parks are important for mitigating daytime surface heat, yet the conventional treatment of park area as a single predictor can obscure whether cooling performance, spatial association structures, and marginal area responses differ among park scale classes. To address this gap, this study examined 53 urban parks in Shanghai and developed a scale-differentiated framework integrating multidimensional daytime surface-cooling performance, scale-specific spatial associations, and Conditional Planning Reference Thresholds (CPRTs). Based on Landsat 8–9 OLI/TIRS land surface temperature (LST) data under four representative seasonal thermal backgrounds, LST–distance curves were fitted using 70 m concentric buffers extending 700 m outward from park boundaries. Four indicators were calculated: park cooling area (PCA), park cooling efficiency (PCE), park cooling intensity (PCI), and park cooling gradient (PCG). The results revealed scale-dependent and indicator-specific patterns in daytime surface-cooling performance. PCA tended to increase with park scale, whereas PCE was higher in small parks, indicating a divergence between cooling extent and unit-area efficiency. PCI showed no stable monotonic scale response, while PCG varied more markedly among representative seasonal thermal backgrounds. Exploratory scale-specific regression models suggested differences in the relative prominence of cooling-related spatial associations among park scale classes. Small parks exhibited a pronounced park–surrounding coupling association pattern, with buffer-zone thermal background and three-dimensional built-environment conditions showing greater relative prominence; medium parks were associated mainly with park-interior landscape composition, particularly green-space proportion; and large parks were associated more strongly with peripheral spatial structure related to the outward extension and attenuation of park cooling effects. Estimated CPRTs were mainly concentrated within 4.7–10.0 ha for small parks and 20.8–29.1 ha for medium parks, whereas large parks yielded higher but less constrained estimates. Together, these findings add a scale- and cooling-objective-sensitive empirical perspective to urban park cooling and support locally calibrated, scale-differentiated planning assessment in Shanghai rather than the uniform application of a “larger is better” assumption.