Noise separation in field data

Four shot-gathers are used to explore the efficacy of this method to remove linear events from shot-gathers. The paradigm is that linear events due to direct arrivals and ground-roll are noise and hyperbolic events due to subsurface reflectors should be maintained. The simple approach is to forward model only the linear model-space which can will then be directly subtracted from the data.

Figure shows the data examples that will be used. They are shots 08, 14 and 25 from the Yilmaz shot-gathers. The three examples contain a nice compilation of common land data artifacts to be attacked through the removal of linear events. Direct arrivals, refractions, and ground-roll are all experienced to some degree both at early time and obscuring primaries late in the sections. A strong linear wavetrain dominates the early time of shot 08, Panel (a). Shot 14, Panel (b), has faint linear events at late time and several slopes at early time. The non-hyperbolic moveout at 5.2 seconds is a nice marker for judging the quality of the various inversion schemes. The obvious ground-roll in shot 25, Panel (c), is the main goal to eliminate.

 

d
Figure 3 Shots from the Yilmaz data collection. Panel (a) is shot 08, Panel (b) is shot 14, Panel (c) is shot 25.

 

• Radon domain
• Data residuals
• Noise subtraction