Rough seas and statistical deconvolution
Ed Kragh and Robert Laws
Journal name: Geophysical Prospecting
Issue: Vol 54, No 4, July 2006 pp. 475 - 485
Info: Article, PDF ( 711.58Kb )
The rough-sea reflection-response varies (1) along the streamer (2) from shot to shot and (3) with time along the seismic trace. The resulting error in seismic data can be important for time-lapse imaging. One potential way of reducing the rough-sea receiver error is to use conventional statistical deconvolution, but special care is needed in the choice of the design and application windows.
The well-known deconvolution problem associated with the non-whiteness of the reflection series is exacerbated by the requirement of an unusually short design window – a requirement that is imposed by the non-stationary nature of the rough-sea receiver wavelet. For a synthetic rough-sea data set, with a white 1D reflection series, the design window needs to be about 1000 ms long, with an application window about 400 ms long, centred within the design window. Although such a short design window allows the deconvolution operator to follow the time-variation of the rough-sea wavelet, it is likely to be too short to prevent the non-whiteness of the geology from corrupting the operator when it is used on real data.
If finely spatial-sampled traces are available from the streamer, the design window can be extended to neighbouring traces, making use of the spatial correlations of the rough-sea wavelet. For this ‘wave-following’ approach to be fruitful, the wind (and hence the dominant wave direction) needs to be roughly along the line of the streamer.