Non-stationary Iterative Time-Domain Deconvolution for Enhancing the Resolution of Shallow Seismic Data

The resolution of near-surface seismic reflection data is often limited by attenuation and scattering in the shallow subsurface which reduces the high frequencies in the data. Compensating for attenuation and scattering, as well as removing the propagating source wavelet in a time-variant manner can be used to improve the resolution. Here we investigate continuous non-stationary iterative time-domain deconvolution (CNS-ITD), where the seismic wavelet is allowed to vary along the seismic trace. The propagating seismic wavelet is then a combination of the source wavelet and the effects of attenuation and scattering effects, and can be estimated in a data-driven manner by performing a Gabor decomposition of the data. For each Gabor window, the autocorrelation is estimated and windowed about zero lag to estimate the propagating wavelet. Using the matrix-vector equations, the estimated propagating wavelets are assigned to the related columns of a seismic wavelet matrix, and these are then interpolated to the time location where the maximum of the envelope of the trace occurs within the iterative time-domain deconvolution. Advantages of using this data-driven, time-varying approach include not requiring prior knowledge of the attenuation and scattering structure and allowing for the sparse estimation of the reflectivity within the iterative deconvolution. We first apply CNS-ITD to synthetic data with a time-varying attenuation, where the method successfully identified the reflectors and increased the resolution of the data. We then applied CNS-ITD to two observed shallow seismic reflection datasets where improved resolution was obtained.