Large‐Scale Seismic Inversion Framework

Abstract

Since its development and first applications in the late 1970s (e.g., Aki and Lee, 1976; Aki et al. , 1977; Dziewonski et al. , 1977), seismic tomography has developed into one of the most powerful tools to investigate the internal structure of the Earth from local to global scales. Tomographic Earth models have become increasingly detailed, thanks to the continuous densification of the global station network (e.g., Roult et al. , 2010; Gee and Leith, 2011), the installation of dedicated arrays (e.g., SKIPPY, van der Hilst et al. , 1994; USArray, www.usarray.org, last accessed November 2014; IberArray, Diaz et al. , 2009), and the deployment of ocean‐bottom seismometers (e.g., Shiobara et al. , 2009; Obayashi et al. , 2013). Furthermore, methodological developments have sharpened our picture of the Earth. Depending on the nature of the data, the scientific question, and the available resources, seismic tomographers can choose from a rich variety of techniques, including ray tomography (e.g., Kissling, 1988; Spakman, 1991; Grand et al. , 1997; Rawlinson and Sambridge, 2003), various finite‐frequency methods (e.g., Yomogida, 1992; Dahlen et al. , 2000; Friederich, 2003; Yoshizawa and Kennett, 2004, 2005), or full‐waveform inversion based on numerical solutions of the wave equation (e.g., Tarantola, 1988; Chen et al. , 2007; Fichtner et al. , 2009; Zhu et al. , 2012; Fichtner et al. , 2013; Afanasiev et al. , 2014). Improvements of data coverage and inversion technology give rise to new challenges that need to be addressed to ensure continued progress. These challenges include the following: (1) Exponentially growing amounts of data and metadata must be retrieved, organized, quality controlled, and updated. (2) Data and metadata are available in many different, often purpose‐tailored formats and with variable pieces of information, which makes the handling of large datasets unnecessarily cumbersome. (3) The …