Although supervised deep learning (DL) offers a potent solution for removing noise from seismic records, challenges are encountered owing to the scarcity of noise-free labels. The innovative Noise2Noise method eliminates the need for clean training targets and extends the applicability of deep learning to seismic data denoising. In this study, we introduce the Noise2Noise Enhancement (N2NE) framework, which improves upon the conventional noise reduction methods used in seismic processing. The applicability of this framework was quantitatively examined using actual field noise under two scenarios: with and without repeated shots. In scenarios with repeated shots, the N2NE framework enhances the conventional stacking method. In addition, the substack strategy, which employs smaller substacks for preliminary noise suppression before DL training, boosts noise suppression. In scenarios without repeated shots, the N2NE framework refines conventional denoising methods (F-X deconvolution) by utilizing information from the common-shot and receiver domains. The N2NE framework lays a foundation for future research on N2N-based seismic denoising methods and contributes to improving the quality of seismic records and the efficiency of data acquisition.

Over the last decade, the evolution of survey acquisition design composed of numerous receivers and shot points has exponentially increased the seismic data volume. As a result, the demand for efficient algorithms for detecting seismic first arrivals has been growing. Machine learning methods, especially convolutional neural networks, have shown great potential in previous studies. However, most existing methods lack generalization ability or be sensitive to noise. This paper proposes a new picking strategy based on the single-to multi-domain selftrained network to achieve automated picking for the first arrivals. Our self-trained network learns from a single domain to multiple domains while rejecting the wrong predicted picking and assigning the pseudo-first-arrivals-times to the unlabeled dataset. The model inherits the advantage from both single-and multi-domain picking strategies; it has high generalizability and noise robustness. In addition, the model is designed for a limited amount of manually annotated data to reduce human labor and total seismic processing time. In experiments on an open hard rock seismic multi-survey dataset, our picking method outperforms the benchmark on most evaluation metrics. Remarkably, our approach achieves approximately 10 % accuracy improvement against Lalor site data with a sampling rate that differs from training data. The experimental results demonstrate our method’s high generalizability and robustness for unseen datasets.