Conferences  >  Physics  >  Metrology and Instrumentation  >  United Arab Emirates

While Full Waveform Inversion (FWI) has made giant strides recently for deep water subsurface imaging, the technology is lagging significantly for land and shallow water applications. Noise — whether ambient or source related — is clearly an issue, especially for the low frequencies that are so critical for successful implementation. Perhaps more problematic is the need to construct an accurate high-resolution model of the near surface. Depending on the terrain, the near surface can have large velocity contrasts, high anisotropy, high attenuation, significant heterogeneities leading to scattering, and strong multiple generators. A proper model of the near surface needs to be possibly viscoelastic and requires a fine computational grid to handle the ultra-low elastic velocities we encounter in the near surface, which may be a challenge for today’s computer capacity. To make matters worse, the near surface is often ill-sampled by cost effective 3D acquisition designs. Despite all these hurdles we are starting to see some beneficial FWI applications to land and shallow water surveys. How important are data conditioning, wavelet estimation, and the choice of the initial model? Do we have to move to elastic FWI? Can ground roll, mud roll, and multiples help rather than hinder the process? Do we need to design our surveys differently? Can we use other complementary data? What do we do with all the scattering noise? What is the role of machine learning (ML)? Is there room for improvement in acquisition? The goal of this workshop is to review the state-of-the art and investigate the next key steps to address some of these questions. Even though FWI is a potentially “complete model” of complex wave propagation, is the problem currently too difficult and too expensive that it needs help in the form of data conditioning and model conditioning? If so, what type of data conditioning and modeling conditioning?

Geophysics and Geology,    Scientific Computing and Data Science