The impact of the 3D effect on the regional-scale velocity model building using 2D full-waveform inversion

Regional-scale active seismic surveys are the methods of choice to probe the lithosphere. One of the key parameters of these surveys is the spatial sampling of the investigated area. Dense sources and receivers sampling coupled with broadband frequency of signals translates to fine-scale probing of the subsurface from a broad range of perspectives. In practice, however,  acquisition design of academic deep crustal seismic surveys typically assumes a compromise between the experiment logistic and the key parameters of the acquisition geometry. This compromise often leads to 2D surveys realised with few tens of receivers deployed along the profiles that can be up to few hundred kilometers long. As a result the high costs of 3D surveys are mitigated at the price of the quality of the resulting data, that cannot be fully exploited by advanced processing techniques - such as waveform-based inversion methods. In particular, the 2D data acquisition and subsequent imaging pose the inconsistency between the 3D wavepath traveled during the survey and the 2D wavepath forced during the 2D processing. This is because geological heterogeneities cause changes of direction of wave propagation, which is indicated by the three-dimensional wave vector spanned at a given point of subsurface. If the 3D-effect is strong due to the complexity of the underlying structure, then the 2D assumption of wavefield propagation during processing cannot honor the field conditions and must lead to errors in the reconstructed velocity model or migrated image. 

Recent years have shown a massive development of waveform inversion and migration methods. In terms of regional-scale seismic imaging, there were few documented onshore and offshore case studies that attempted to process 2D archival academic data with full-waveform inversion (FWI) and extract structural information beyond the resolution-limit of the traveltime tomography. However, the limitations originating from the legacy acquisition make it difficult to fully exploit the potential of FWI. In this study we evaluate the ability of regional-scale velocity model-building technics to handle out-of-plane propagation and investigate how this effect manifests itself in the data. Through the insight of  wave propagation within complex subsurface models we underline the problem and make first attempts to the broader investigation of the optimization of 3D academic regional surveys. We extract various 3D target models from the synthetic model of a subduction zone and we use those models to generate seismic data along 2D OBS lines. Subsequently we use 2D FWI to evaluate how the out-of-plane propagation affects the results of 2D velocity model-building from the data generated along the 2D OBS lines but using 3D modelling and 3D target models. We compare those results with the scenario where the 2D FWI is applied to the data from the same 2D OBS lines but generated using 2D velocity models and 2D modelling. We perform polarization analysis to demonstrate how the 3D effect manifest itself in the OBS gathers. Finally we also run 3D FWI with different OBS acquisition settings to investigate their impact on the final model reconstruction.