An effective data processing workflow for broadband single-sensor single-source land seismic data

Abstract

Examples of raw and processed broadband single-sensor single-source land seismic data acquired in the Middle East region have been found to be significantly noisy, and very low-frequency signal has been either missing or unrecoverable. In response, an effective and pragmatic processing workflow has been developed that substantially improves the quality of the final processed data, to the extent where we can say that original survey objectives can be met. The new workflow includes early deterministic deconvolution for a number of filtering effects in the recorded signal wavelet, with the aim of flattening the signal wavelet amplitude spectrum over the vibroseis sweep frequencies and zeroing the wavelet phase. This includes the key innovative step of converting the recorded particle motion to that of the vibroseis far-field signal, those respectively being particle acceleration and particle displacement. This significantly boosts low-frequency amplitudes relative to higher frequencies such that it becomes possible to deterministically compensate for earth's absorption using a large gain limit with less concern for overamplifying high-frequency noise. An application of a source designature compensates for the nonflat design of the pilot sweep, further increasing signal amplitudes over the low-frequency ramp-up portion of the sweep. With the flattened signal spectrum, it is possible to better assess trace noise characteristics across the full bandwidth and perform better QC for its removal. Subsequent statistical deconvolution becomes more of a correction for residual effects on the signal wavelet, and the use of trace supergrouping further mitigates the effect of noise on statistical deconvolution and other data-adaptive processes.