Abstract

A vast amount of analogue seismograms recorded between the end of the nineteenth century and late twentieth century are often inaccessible for seismological research since they are not available as digital time series. This manuscript describes freely available software, DigitSeis, that takes a digital image of an analogue seismic record and returns waveforms either as a function of their x-y position on the image or as time–amplitude information. The overall structure and approach of the software are provided along with how they have evolved over different versions. The effectiveness of the software is demonstrated with three examples. The first example is a long-period east–west seismogram recorded at the Harvard Seismographic Observatory on photographic paper in May of 1938, which contains signals associated with a magnitude 7.7 earthquake that occurred off the coast of Northern Ibaraki, Japan. The second example is an analysis of a 35-mm microfilm copy of the short-period vertical seismogram recorded at Tucson, Arizona, on July 16, 1945, that shows blast signals from the first nuclear bomb detonation. The final example uses a 70 mm microfiche image of a long-period north–south seismogram recorded at College, Alaska in December of 1966, which shows a pair of earthquakes with nearly identical waveforms. The software is, by no means, perfect, and discussion of its limitations such as the compatibility with pen- and Develocorder-type seismograms is included, as well as comments about challenges of incorporating machine learning into the digitization process.