Rescuing Legacy Seismic Data FAIR’ly

The Earth’s interconnected and dynamical systems operate on a spectrum of time scales from millions of years to fractions of a second. Although the evolution of the Earth and its deep interior are beyond the time span of human observations, understanding of many natural phenomena operating on human time scales has benefited from direct scientific observation. Continuous processes and those that are repeated over time shape the environment we live in. As we are faced with unprecedented changes to climate, understanding the deeper patterns and trends in natural systems through time has taken on new importance (Research Data Alliance, 2019). The call to reuse data is driven not only by economics but also by the recognition of their uniqueness (observations of natural systems are not repeatable) and scientific value in enhancing current understandings as well as potential new discoveries especially in the era of big data. These data are part of the historical record and our scientific heritage (American Geophysical Union, 2019) not only in explicitly recording earth observations but implicitly recording, and thus providing the evidence that addresses the manner in which science was conducted. Recorded observations of ground motion began with the advent of the instrumental era of seismology in the late 1800s. Data began to be systematically collected in the early part of the twentieth century, evolving into today’s system of regional and worldwide networks. The digital and Internet revolution in the late part of the twentieth and early twenty-first century ushered in the current era of continuously recorded, digital data. Federated repositories quickly followed, easing the barrier of access to high-quality, digital time series data. With these data, the scientific community has developed new techniques and continues to discover new phenomena that reveal themselves as seismic moment, enhancing our understanding of the relationships among diverse Earth systems. However, digital data only capture a fraction of the observations in the instrumental era (Ishii, 2018). A better picture of phenomena such as the earthquake cycle, magmatic systems, and the solid earth’s response to climate change can be drawn by incorporating more observations from the historic record. The emerging use of machine learning in the geosciences shows future promise in finding signal and patterns—extracting previously unrecognized features in big data (Kong et al., 2018). Analog seismograms are an untapped resource. To be available to modern research techniques, the analog data must be accessible in digital form. These data are often kept in remote locations and not under archival condition leading to the physical deterioration and, hence, loss of some records. Adding to the sense of urgency is the lack of institutional will to maintain these resources and the retirement of a generation of scientists who used the data from these instruments. This knowledge represents significant institutional memory of how to properly interpret and analyze these data, as well as about the body of the resource itself that must be captured. Preservation efforts are underway at a limited number of institutions worldwide. Digital preservation efforts of historical Italian and European seismograms through the Istituto Nazionale di Geofisica e Vulcanologia SISMOS and EuroSeismos (Ferrari and Pino, 2003) projects, Japan (Murotani et al., 2020), and HRV (Ishii et al., 2015) provide leadership in best practices in conservation, imaging, and vectorization (see Data and Resources for links to these data). These efforts, however, have proceeded with little community discussion on standards and how the collections can explicitly meet Findable, Accessible, Interoperable, and Reusuable (FAIR) data principles. (Wilkinson et al., 2016). That these data be FAIR provides guidance for data management and stewardship in the modern digital ecosystem. Guidelines must also be supplemented with domain requirements and leverage existing standards and infrastructure. Seismology has a rich history of sharing and has established good data management practices in the federation of its natively digital data. However, it is generally acknowledged