Identifying Issues for Long-Term Data Understandability

In the past decade, we have seen a serious effort to capture and make useful a record of Earth observations. While the EOSDIS effort has had some success, the longer term prospects for preserving our knowledge of how the Earth has changed must still overcome a number of difficult challenges. The first, and most important, may be moving from the expectation that we can produce a single system that will solve all of the difficulties. Rather, we need to begin to view what we do from the perspective of a continuously evolving digital library. There are several key issues we need to deal with in trying to bring about this shift in perception: {a.}{We need a stable data referencing approach - like a URI, but one that allows us to reference individual data elements and subsets, not just files or data products.} {b.}{We need more adaptive metadata - including metadata that allows us to tie together data sets maintained in different places by scientific tribes that are linguistically in different communities of discourse.} {c.}{We need more automated approaches to ensuring the documentation of data provenance and of the web of scientific discourse that goes on mostly in scientific journals.} {d.}{We need to find ways of preserving both the data and its web of meaning even while we have to transfer it from medium to medium and system to system. The easy part of the problem is transferring the data - and that's the part that's been discussed so far. The hard part is carrying the scholarly apparatus that allows us to accumulate knowledge about what's in the data.} {e.}{We need additional approaches to visualizing the connections between various data sets and of recording objects that experts find for deeper examination.}