The Long Term Ecological Research Network adopted Ecological Metadata Language as its metadata standard in 2003, and has since generated over 5000 Ecological Metadata Language documents that describe diverse datasets collected at its 26 member research sites. The Long Term Ecological Research Network has accumulated vast experience using Ecological Metadata Language, and our survey of the implementation process across the Long Term Ecological Research Network gives us insights into the pragmatic aspects of Ecological Metadata Language use and a sense of what is useful in Ecological Metadata Language, what needs to be improved, what elements can be ignored or deprecated and what is lacking. In this paper, we describe a set of common practical working concepts for Ecological Metadata Language that will give new Ecological Metadata Language users and adopters a valuable guide to surviving the Ecological Metadata Language implementation process. In addition, our analysis of Long Term Ecological Research Network's archive of Ecological Metadata Language documents revealed that while overall metadata content is rich, elements such as the measurement type and geospatial domain frequently contain errors due to inconsistency or incorrect data entry. We discuss simple tools that can be used to quality assure the Ecological Metadata Language as it is generated, and we recommend more specific guidelines and practices for documenting data sets using the Ecological Metadata Language, such as providing spatial coordinates and typing attributes. We also make suggestions on the possible directions for the next generation of Ecological Metadata Language based on community experience and future data integration challenges. Finally, we show some scientific results of data synthesis driven by quality-controlled, content-rich, Ecological Metadata Language data sets generated with a semi-automated process synthesis framework that illustrate the power of EML to facilitate LTER research.