Earth System Science (ESS) has developed into a new historic stage. Because it takes the whole earth system and its interactive actions among spheres, its development needs lots of multi-disciplinary, multi-sources, multi-type, integrated geosciences data resources. According to this requirement, Data Sharing Network of Earth System Science (DSNESS) was established at the beginning of China Scientific Data Sharing Program (SDSP) launched in 2003. It has been one of the long-term data sharing platforms in National Science & Technology Infrastructure (NSTI) since 2005. Data sharing in DSNESS needs standards and specifications environment. Based on the ESS data characteristics’ analysis, 3 main problems needed to be studied at present, i.e., how to establish the ESS data sharing mechanism? How to integrate and share multi-disciplinary data? How to make data easily find and access for use? Basic data sharing concept models are studied firstly for these questions. According to four principles designed, ESS data sharing standards and specifications system and its relationship among related international and domestic standards systems are studied. Designed standards and specifications system includes four main classes, i.e., mechanism and rules class, data management standards and specification class, platform development specification class and data service specification class. A total of 18 standards, rules and specifications have been drawn up. Faced to the problems mentioned above, there are four key standards and specifications in the architecture, i.e., ESS data sharing union constitution, core metadata standard, data quality control measurement specification and data classification system. Through almost 6 years research and application, nowadays all of these 18 standards and specifications have been used successfully in the distributed data sharing network of ESS which include one general center and 13 sub-centers. In the near future, standards and specification environment of DSNESS will be further developed towards two directions, one is basic concept reference model, and the other is data fusion and assimilation standards and specifications.