Computational server for seismic processing and inversion

点击翻译按钮获取中文摘要

Analysis of seismic data from active source surveys, which comprises the bulk of my research group's activity, has become increasingly computationally intensive over the last decade as processing algorithms have become more sophisticated and 3-D surveys have increased the size of the datasets recorded. A key element of our work has been the processing of seismic reflection surveys using a state-of-the-art, industry-standard software package (ProMAX). Some algorithms in this software, such as those used to attenuate long-period multiples and reveal underlying crustal reflections, are not otherwise available, and are critical to my work in the Mariana and Cascadia subduction zones where the seafloor depth varies dramatically. However, this software will no longer be supported on our existing proprietary Sun/Solaris platform. In this proposal, I am therefore requesting funds to purchase a high performance computational server that will run under the supported open-source Linux operating system and be equipped with around 1 Tb of disk to store the large datasets with which my group will be working. Tomographic inversion of first arrival travel times, which has also formed a significant part of my work over the last eight years, is very time-consuming on our existing Sun/Solaris computer system when applied to 3-D datasets such as the recent crustal study of the Mariana island arc. Inclusion in these inversions of other arrivals such as wide-angle reflections, which is the next logical step towards improving the resolution of these crustal models, requires a much greater degree of computation, and is not practical with our present computer hardware. Using this significant increase in our computational facilities, my group will be able to apply the latest data processing algorithms to problems such as the seismic mapping of intrusions within and beneath the crust of the Mariana arc, and the identification of regions of possible lower crustal foundering.