Fusion of ocean data from multiple sources using deep learning: Utilizing sea temperature as an example

For investigating ocean activities and comprehending the role of oceans in global climate change, it is essential to gather high-quality data. However, existing observation data have deficiencies such as inconsistent spatial temporal distribution, severe fragmentation, restricted depth layers. Data assimilation computationally intensive, other conventional fusion techniques offer poor precision. This research proposes a novel multi-source network (ODF-Net) based on deep learning solution for these issues. The ODF-Net comprises number one-dimensional residual blocks that can rapidly fuse observations, satellite three-dimensional model output reanalysis utilizes vertical profile target constraints, integrating physics-based prior knowledge improve precision fusion. structure contains channel attention mechanisms guide model’s most crucial features, hence enhancing performance interpretability. Comparing multiple sea temperature datasets reveals achieves highest accuracy correlation with observations. To evaluate feasibility proposed method, monthly dataset resolution 0.25°×0.25° produced by fusing from sources 1994 2017. rationality tests show reliable various sources.