Study Of Waveform-based Joint Inversion Method With Data Attribute Extraction

Due to the complexity between the seismic data and the geophysical parameters of the underground media,the full waveform inversion(FWI)suffers from the problem of data cycle-skipping and inversion nonlinearity.It may cause the waveform inversion hardly match all the information of the complex seismic data,especially in the field data applications.On the other hand,different wave components of the seismic data correspond to different performance on the geophysical parameter inversion.We utilize mathematical transforms to extract or extrapolate the particular information in the seismic waveform data,and invert for these information instead of the whole data.It will greatly help the waveform inversion to reduce the nonlinearity and improve the convergence.Therefore,we propose a waveform-based joint inversion strategy based on data extraction and extrapolation.According to the specific characteristics of the actual seismic data,different data extraction and extrapolation methods are applied to construct the objective functions with lower degree of nonlinearity.Additionally,we associate the different parts of objective functions together by utlizing a joint inversion strategy.It not only ensures the high linearity of the objective function,but also matches the different information in the seismic data simultaneously.In the linearized seismic full waveform inversion,it is well known that a good initial model is needed to avoid cycle skipping when the low-frequency components of data do not exist.To solve the problem,we develop a joint first-arrival traveltime and early-arrival envelope inversion method(JTE)to recover low-wavenumber model components and create an appropriate near-surface velocity model for FWI.The envelopes of waveforms introduce low-frequency components for linearized waveform inversion,while the traveltime inversion is nonlinear and stable.The combination of the two approaches could take advantages of the two methods and also compensate drawbacks in each method.In two synthetic experiments,we demonstrate that traveltime inversion constrains the top near-surface velocity structures tightly,while the envelope inversion recovers the low-wavenumber structures with low-velocity objects or layers.Using the results of JTE as starting models,we find that FWI can produce accurate solutions for complex numerical models.In a real data example,JTE followed by FWI resolves a near-surface velocity model that can help improve statics corrections for the subsurface stacking image.The first-arrival traveltimes constrain very shallow velocities,the waveform envelope presents low-frequency data,and the high-frequency waveform itself includes information regarding structural details.We propose a joint traveltime,waveform,and waveform envelope inversion method(JTWE)for inverting near-surface velocity structures.By inverting three types of data,we are able to recover both low-wavenumber and high-wavenumber structures and mitigate the cycle-skipping problem in waveform inversion.The calculation of traveltimes and raypaths is fast.Most of the computation effort is focused on dealing with the waveform and waveform envelope.JTWE backward propagates both the waveform residual and envelope residual simultaneously to calculate the model updating gradients.This simultaneous backward propagation strategy ensures that the computational cost of JTWE is similar to the cost of inverting waveform alone.In the synthetic experiment,we demonstrate that JTWE mitigates the cycle-skipping problem and recovers the near-surface structures without the need for additional low-frequency data.The final results of JTWE indicate that it delivers improved results with low-velocity inclusions compared to regular full waveform inversion(FWI).For field data from the Middle East,JTWE helps resolve a complex near-surface model with rugged topography and fit all three types of data.Elastic full waveform inversion(EFWI)is also an important topic.Most FWI schemes are based on acoustic wave equations.Solving an elastic FWI problem in 2D or 3D presents a great computational challenge.Therefore,we develop a pseudo 2D(and 3D)elastic full waveform inversion method in the CMP domain.It is fast because the forward elastic waveform modeling is in 1D,but inversion is performed in 2D(and 3D).We select the time windows of different wave events for the P-wave velocity inversion and the S-wave velocity inversion.Furthermore,we apply cross gradients between P-and S-wave slowness for simultaneous inversion of the two types of waveform data.The approach can be applied to resolve both P-and S-wave velocities where CMP assumption is valid.We demonstrate the method by applying to both synthetics and real data.