[Purpose/significance] Scientific data publishing is the main mode of academic communication for data-intensive scientific discovery, which is of great significance for data reuse and scientific verification. [Method/process] Earth sciences has undergone a great change from the data sharing model to the current data publishing. Current practices of data publishing can be divided into three modes: data journal publishing, data warehousing publishing, data and paper joint publishing. The author made statistics and comparison on the practice methods and key elements of each mode, and emphatically analyzed the advantages and disadvantages of the three modes, peer data review issues and the importance of hierarchical metadata in geoscience data publishing. [Result/conclusion] Through research, the author proposes that warehouse-based publishing facilitates integration into scientific data management process and facilitates data reuse. However, such publishing mode usually lacks peer review. Peer review of data should be different from academic papers and should focus on the reusability of data in the process of participating in scientific research and producing regenerative results. The hierarchical description of metadata is of great significance to the preservation and reuse of geoscience big data.