Deep learning (DL)-based sea\textendash land clutter classification for sky-wave over-the-horizon-radar (OTHR) has become a novel research topic. In engineering applications, real-time predictions of sea\textendash land clutter with existing distribution discrepancies are crucial. To solve this problem, this article proposes a novel Multisource Semisupervised Adversarial Domain Generalization Network (MSADGN) for cross-scene sea\textendash land clutter classification. MSADGN can extract domain-invariant and domain-specific features from one labeled source domain and multiple unlabeled source domains, and then generalize these features to an arbitrary unseen target domain for real-time prediction of sea\textendash land clutter. Specifically, MSADGN consists of three modules: domain-related pseudolabeling module, domain-invariant module, and domain-specific module. The first module introduces an improved pseudolabel method called domain-related pseudolabel, which is designed to generate reliable pseudolabels to fully exploit unlabeled source domains. The second module utilizes a generative adversarial network (GAN) with a multidiscriminator to extract domain-invariant features, to enhance the model's transferability in the target domain. The third module employs a parallel multiclassifier branch to extract domain-specific features, to enhance the model's discriminability in the target domain. The effectiveness of our method is validated in twelve domain generalizations (DG) scenarios. Meanwhile, we selected 10 state-of-the-art DG methods for comparison. The experimental results demonstrate the superiority of our method.