Changjiang Diluted Water (CDW), discharged into the East China Sea (ECS) during the summer rainy season, decreases sea surface salinity, strengthens stratification, and promotes sea surface warming. As a major driver of changes in the stratifying vertical structure of the ocean, CDW influences marine ecosystems and environmental variability and is closely associated with marine heatwaves. However, subsurface temperature and salinity observations in the ECS remain very limited except for in situ measurements, making it difficult to investigate the influence of CDW on subsurface water masses and to characterize its vertical propagation and spatial expansion. To address this limitation, this study employed a Vision Transformer (ViT)-based model to estimate subsurface temperature and salinity structures. Temperature and salinity observations from the World Ocean Database (WOD) were used as target variables, while five environmental variables were used as input to reconstruct temperature and salinity fields at depths of 10–50 m for the period 2015–2025. The reconstructed dataset was subsequently used to analyze the long-term variability of subsurface marine heatwaves and to investigate their relationship with CDW. By reconstructing subsurface temperature and salinity fields using an in-situ observation-based deep learning model, this study aims to improve our understanding of the mechanisms governing marine heatwaves and salinity variability in the ECS, thereby providing new insights into the oceanographic impacts of CDW on the marine environment.