Urban heat stress increasingly threatens environmental sustainability in rapidly growing megacities. Despite their widespread promotion as nature-based solutions, the cooling performance of blue-green spaces (BGS) heavily relies on their spatial structure and scale. This study examines how the morphological characteristics of BGS influence urban thermal patterns in Delhi. Using satellite-derived land surface temperatures and land cover classifications, patch-level thermal behaviour was analysed across different urban surface types. Spectral indices and spatial metrics were integrated with machine-learning modelling to identify the dominant drivers of cooling and warming dynamics. The findings reveal that urban thermal behaviour is strongly governed by spatial configuration rather than mere green cover. Impervious and exposed surfaces intensify heat accumulation as their spatial dominance increases. However, cooling effects strengthen when blue-green spaces become larger, more continuous, and structurally connected. Water bodies and dense vegetation demonstrate the most stable cooling performance, while fragmented or small patches provide limited thermal relief. Machine-learning interpretation highlights moisture availability and vegetation condition as key regulators of cooling intensity. The results indicate that cooling benefits emerge only when blue-green spaces surpass critical spatial thresholds and maintain morphological integrity. The study demonstrates that sustainable heat mitigation in megacities requires a shift from area-based greening targets toward morphology-sensitive spatial planning. Embedding connected and scale-appropriate blue–green infrastructure within urban form is essential for enhancing long-term environmental sustainability.