Realistic Simulations of Geodetic Network Data: The Fakenet Package

TitleRealistic Simulations of Geodetic Network Data: The Fakenet Package
Publication TypeJournal Article
Year of Publication2013
AuthorsAgnew DC
JournalSeismological Research Letters
Volume84
Pagination426-432
Date Published2013/05
Type of ArticleArticle
ISBN Number0895-0695
Accession NumberWOS:000318328000003
Keywordscontinuous gps; coordinate time-series; deformation; random number generators; software
Abstract

The Southern California Earthquake Center (SCEC) Transient Detection intercomparison required that all participants apply their methods to the same datasets, each dataset containing some tectonic signal (or none), plus realistic noise. One way to produce these would be to add tectonic signals to actual data from the California continuous GPS network (CCGPSN), but there are two problems with this: using the same background series for all datasets makes it easy to isolate signals by differencing, and we would not know if there was a transient already present in the data. Instead, fully synthetic datasets were used, each of which had a different realization of the noise added to a different signal known only to the creator of the dataset. The Fakenet package was written to provide software for creating these datasets.

In developing Fakenet, the goals were to be able to:

  • Replicate, at least approximately, the observed statistical properties of Global Positioning Systems (GPS) data.

  • Create multiple realizations, with different non‐tectonic fluctuations. Testing a detection algorithm can then be done by examining its success or failure for datasets with the same signal but different versions of the noise.

  • Allow a wide range of source locations, types, and time histories.

  • Provide a log file to document the inputs for a particular simulation. This is useful for debugging, but more importantly it provides complete reproducibility (Mesirov, 2010; Lees, 2012): given the information in the log file, we can rerun the code to get the same results.

This paper describes the procedures used in this package, as well as outlining its structure. 

DOI10.1785/0220120185