Advancing seismic inversion in practice

点击翻译按钮获取中文摘要

Seismic images are unrivaled in their ability to reveal the location and identity of economic resources in the earth's crust. No other technology can provide the volumetric coverage and the spatial resolution of a seismic image, and the information content of seismic images has increased dramatically over their 50 year lifespan. Today, the computer methods used to create such images are transitioning from a standard methodology (SM), which incorporates an evolved blend of physical theory and practical experience, to the more modern full-waveform inversion (FWI), which is more firmly rooted in modern mathematical physics. Widespread use of FWI is hindered due to a variety of difficulties including insufficient low-frequency content in seismic data, the unknown seismic waveform, incompletely understood physics, and the extreme computational effort required. However, these difficulties are not new and have been addressed in SM in a variety of practical ways. This research will address the current roadblocks to FWI by incorporating the practical aspects of SM where possible and by new developments where required. FWI seeks an earth model which, when used in a simulation, can replicate the observed seismic data. In contrast, SM seeks an earth model that matches well logs measured in available boreholes. FWI does not constrain the earth model to match well information, and SM does not try to match the observed seismic data. In SM well logs are essential to both provide the low-frequency information and to determine the seismic waveform. Incorporation of data matching into SM or putting well matching into FWI should converge to a common algorithm. This research program will develop specific methods to incorporate well log information into FWI while retaining other aspects of FWI such as matching the observed data. A second aspect of this research will investigate methods to record lower seismic frequencies and to successfully process them into a viable image. These efforts, if successful, could significantly increase the information that can be derived from seismic images. Ensuring that the information is consistent with both the observed seismic data and available well logs should also reduce uncertainty.