A computational method for full waveform inversion of crosswell seismic data using automatic differentiation

Full waveform inversion (FWI) is an effective and efficient data-fitting technique that has been widely used to produce accurate estimation of model parameters in Geophysics. The efficiency and accuracy of FWI are determined by the three main components: numerical solution for forward problem, gradient calculation and model update which usually involves the optimization method. The success of the adjoint-based field data inversion relies on the choice of the methods for solving these time consuming components. In this paper, we introduce the automatic differentiation (AD) tool TAPENADE to compute the gradient of the objective function, hence the FWI workflow is simplified so we can focus mainly on the forward modeling and the model updating. Based on the result from forward modeling, and the observational data, the objective function is calculated as the misfit function, which is then minimized by utilizing an optimization algorithm. A variety of methods can be used to minimize the misfit function, whereas in this work we choose the limited memory BFGS (L-BFGS) method due to the low requirement on memory usage. Numerical test for a 2D model shows that the combination of the AD tool and L-BFGS method is effective and efficient to solve the full waveform inversion problem.