Batch seismic inversion using the iterative ensemble Kalman smoother

Abstract An ensemble-based method for seismic inversion to estimate elastic attributes is considered, namely the iterative ensemble Kalman smoother. The main focus of this work challenge associated with waveform data. amount data large and, depending on size, it cannot be processed in a single batch. Instead solution strategy partitioning recordings time windows and processing these sequentially suggested. This demonstrates how can done adaptively, reliable efficient estimation. adaptivity relies an analysis update direction used procedure, interpretation contributions from prior likelihood update. idea that must balance; if dominates, estimation process inefficient while likely overfit diverge dominates. Two approaches meet balance are formulated evaluated. One based eigenvalue distributions enters affects weighting contributions. other balancing norm magnitude vector components Only latter found sufficiently regularize window. Although no guarantees avoiding divergence provided paper, results adaptive procedure indicate robust performance achieved