DATA MANAGEMENT SYSTEMS FOR WELL DEFINED AND REPEATABLE DATA SETS: PROVIDING THE INFORMATION NEEDED TO GO FROM IMAGINATION TO REALITY

摘要

There are both scientific and legal reasons for providing the information needed to be able to show exactly how a data set was made and to be able to repeat its production with reasonable expectation of getting the same answer. Current Earth observation data management systems often do not support or even allow provision of all the information necessary to do this. We will describe a system and operational procedures that automate collection of much of the information necessary to answer these questions and provide tools for accessing and using this information. The system automatically captures full provenance for all of the elements used in the data set production including algorithm version and source code, input data set version and source, and hardware and processing system software configuration. This is an ongoing development and we will be presenting first results of our system as well as the programmatic implications of maintaining this information on scientists, funding agencies, and archival institutions.