THE LOST WORLD: PRESERVING ANALYSES FOR FUTURE WORKERS

A motto in science is that we stand on the shoulders of others to do our research. In some cases, earlier insights were so complete that the final conclusions from older work served as the basis for newer studies. But frequently we have built upon raw observations and intermediate analyses, a capability that is increasingly lost to us. Papers from 50 and more years ago listed not only raw data but also key derivations from that data before making inferences. Later workers could base new studies from nearly any point in the analysis. More recently, papers present no more than a small fraction of the full analysis and often present it without a capability for recovering it for further work (e.g., a color slice of a tomogram is usually worthless for estimating a new seismic velocity structure). Although great strides are being made in preserving raw data, analyses built upon those data are generated once and then lost.

Two examples from my own experience serve to illustrate this. In the first, I used data from a 1955 GSA Bulletin paper to reexamine uplift of the Sierra Nevada. That 1955 paper had as original observations the locations of Eocene river gravels, which in turn were connected with likely up- and downstream equivalents, defining local grades and flow directions, and finally integrated into a paleoriver system. I reanalyzed the inferred grade and flow directions in the intermediate step to address questions of Sierran uplift; this reanalysis was easily completed from the published paper but would be difficult with the raw data or final result. In contrast, recent tomographic experiments students and I have published lack the winnowed datasets, complete set of arrival picks, and the full 3-D inversion results. Others seeking to build upon this would either have to rebuild from archived datasets or locate my students or me and hope we could provide this information. Thus one of the great challenges facing earth science is to find a way of archiving many kinds of intermediate results in a manner that is both relatively painless for the originator and useful for any future researcher. Without success in this, our successors will speak of standing on our sneakertops instead of on our shoulders.