.png){kind=icon}
.png){kind=icon}
Return
Target-oriented full waveform inversion based on optimal transport theory
.png){kind=icon}
PRE
AI
DOI：10.1016/j.jappgeo.2026.106120.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Target-oriented full waveform inversion (TOFWI) provides an efficient strategy for high-resolution imaging in local regions of interest, but its effectiveness is often limited by two main challenges: the need for redatuming of the acquisition system and the sensitivity of conventional L2-norm-based TOFWI (L2-TOFWI) to initial models and data quality. In this study, we propose an optimal transport (OT)-based TOFWI framework to overcome these limitations. First, we apply the Marchenko redatuming method to retrieve virtual reflection responses that isolate the target-zone wavefield. Then, we incorporate OT-based misfit functions, including graph-space OT (GSOT) and Kantorovich-relaxed OT (KROT), to enhance the convexity of the inversion landscape and improve robustness against noise and velocity-model inaccuracies. KROT introduces entropy regularization via the Sinkhorn algorithm, leading to smooth transport plans and improved numerical stability, whereas GSOT relies on a discrete assignment formulation that yields sparse transport plans and higher sensitivity to local waveform variations. Numerical experiments demonstrate that the proposed OT-TOFWI framework delivers more accurate and stable reconstructions than conventional L2-TOFWI, particularly under conditions of significant initial-model errors and low signal-to-noise ratio. Furthermore, comparisons with global full waveform inversion highlight that OT-TOFWI achieves better resolution in deeper structures with lower computational cost. These results confirm that integrating Marchenko redatuming with OT-based misfit functions provides a promising pathway for reliable target-oriented seismic imaging in complex geological settings.
Keywords:
Full-waveform inversion
Target-oriented
Marchenko redatuming
Optimal transport distance
Graph-space
Sinkhorn algorithm
Journal
J
IF:
2.1
Papers:
247
Citations:
8.3K