The paper discusses the need for better application of descriptive science methods in building integrated databases required by global change research. The concept of environmental 'characterization' is cited as an appropriate paradigm for defining this need. The role of characterization databases with respect to modeling and synthesis is discussed. An existing five year interagency project to develop a public 'global change' database is described. Preliminary lessons from this project imply a great need to bridge the gap between voluminous yet largely untested 'off the shelf' data on the one hand, and more focused quantitative information needed for coordinated global research on the other. Furthermore, it is apparent that more rigorous publication traditions for data (amalogous to those for literature), including open scientific review and testing of datasets, are necessary to improve the scientific and technical quality of descriptive information.