Digitization Procedures of Analogue Seismograms from the Adam Dziewonski Observatory (HRV) at Harvard, MA

This project explores methods of digitization of analogue seismic recordings for better preservation and to facilitate data distribution to the community. Different techniques are investigated using seismograms from one particular station, the Adam Dziewonski Observatory (HRV) at Harvard, Massachusetts. This seismological station, still in operation as a part of the Global Seismographic Network today, is one of the oldest stations in the United States. The station was built in 1933, and since its installation, the station has produced approximately 16,000 analogue seismograms. The majority of these recordings were taken between 1933 and 1953, with some intermittent recordings between 1962 and 1998 after digital seismometers had become a standard. These analogue seismograms have the potential of expanding the database for seismological research such as identification of events previously not catalogued. Due to poor storage environment at the station, some of the records, especially those on regular type of paper, are damaged beyond repair. Nevertheless, many of the records on photographic paper are in better condition, and we have focused on a subset of these recordings that are least damaged. Even these seismograms require cleaning and, in consultation with the Weissman Preservation Center of Harvard Library, preparation techniques for the photographic records are examined. After the seismograms are cleaned and flattened, three different equipments are investigated for digitization, i.e., a copy machine, scanner, and camera. These instruments allow different imaging resolutions, ranging from 200 dots per inch (dpi) to 800 dpi. The image resolution and the bit depth have a wide range of implications that are closely linked to the digitization program one chooses to convert the image to time series. We explore three different software for this conversion, SeisDig (Bromirski and Chuang, 2003), Teseo2 (Pintore and Quintiliani, 2008), and NeuraLog (www.neuralog.com), and determine advantages and disadvantages associated with each software. One of the important features of the software is the automatic tracing algorithms. The success of the automatic tracing depends upon many factors, and this is examined using examples from long and short period recordings with high amplitude (thin and fading lines), and long and short period recordings with low amplitude (well-defined lines). Automatically traced data are also compared to manually traced samples. Based upon these results, we propose a set of procedures and recommendations for cleaning, imaging scheme including resolution and bit depth, and digitization software. Ultimately, we would like to outline a robust procedure for mass seismogram digitization and process all the Harvard station recordings and make them available to the community through the IRIS Data Management Center.