Cross-Domain Adversarial Learning for Sea Surface Temperature Super-Resolution

High-Resolution Sea Surface Temperature (SST) data can provide more accurate information on ocean conditions, which plays a significant role for environmental monitoring and climate research. However, most existing super-resolution methods fail to account for geographical differences across oceanic domains, which hinders their ability to accommodate data diversity and reduces the model's generalization capacity in reconstructing complex SST. To address this limitation, this paper proposes a novel Cross-Domain Adversarial Learning (CDAL) model that utilizes both microwave and infrared SST data for super-resolution task. Specifically, the multiscale information of SST data is extracted to capture local detail and global trend in SST, thereby enhancing feature representation capabilities. Additionally, we design the Asynchronous Frequency Network to further encode and represent the distinct characteristics of different frequency components, e.g., overall trend changes and local rapid variations, leading to improved performance in handling complex data patterns. Finally, a Random Adversarial Classifier is proposed to dynamically alter adversarial samples during training, enabling the model to focus on global properties and intrinsic patterns rather than specific regional characteristics, thus achieving more consistent and generalized performance across different areas. Comprehensive experiments validate the effectiveness of our method for SST super-resolution task.