<p>Accurate retrieval of Land Surface Temperature (LST) is critical for assessing the cooling effects of urban green spaces (UGS) and blue spaces (UBS), which help mitigate the urban heat island effect and improve thermal comfort. This study introduces a novel methodology that integrates deep learning with domain expertise to predict LST, using real LST data, vegetation spectral indices, and spectral bands as inputs. A 1-Dimensional Convolutional Neural Network (1D-CNN) was developed, which outperformed conventional machine learning and alternative deep learning models, demonstrating high predictive accuracy and strong generalization ability. Spatial regression analyses were further employed to examine how UGS of varying sizes influence LST. Results revealed that larger UGS provide strong cooling benefits, while smaller patches contribute less. Remote sensing data from 1991 to 2022 confirmed the significant role of both green and blue spaces in mitigating urban heat, with notable cooling observed around wetlands, rivers, and urban parks. Importantly, combined green–blue configurations enhanced cooling more effectively than blue spaces alone, indicating synergistic benefits when vegetation is integrated with water bodies. Case analysis of afforestation in Daan Forest Park demonstrated how urban greening initiatives can substantially lower local temperatures. Similarly, UBS surrounded by adjacent vegetation exhibited greater temperature reductions compared to isolated water bodies. These findings underscore the need to preserve and expand large, continuous green areas while enhancing connections between green and blue infrastructures. The proposed framework provides robust evidence and practical guidance for urban planners and policymakers to design climate-resilient cities that maximize the co-benefits of UGS and UBS.</p>