Numerical comparisons

Again, the Amoco 2.5-D dataset provides an excellent test dataset for comparing flavors of weighting function. Figure  shows the data-space weighting function for the gather shown in Figure  (a). Similarly, Figure  shows the weighting function for a common-offset section corresponding to a half-offset of 500 m.

 

weightgather18 Figure 8 Data-space weighting function for the gather shown in Figure  (a). Darker shades of grey correspond to higher amplitude.

 

weightsection1 Figure 9 Common offset-section corresponding to h=500 m (a), and its data-space weighting function (b). Darker shades of grey correspond to higher amplitude.

Unfortunately the data-space weights proved susceptible to coherent noise in the form of multiples not predicted by the modeling operator. While data-space weights did improve the signal in poorly illuminated areas, they also boosted up the noise level causing an increase in NSD. Figure  illustrates this with a plot to compare with Figure . While the reliability of the amplitudes are increases in some areas (notably around x=19 km), in other areas, especially x<13 km, there are obvious artifacts. Because of the large (non-Gaussian) outliers visible in Figure  the normalized standard deviation is no longer a meaningful measure of reliability.

 

eventampnd5 Figure 10 Normalized peak amplitude of 3.9 km reflector after migration of the raw data (solid line), and after migration of a dataset that has been normalized by an appropriate data-space weighting function (dashed-line). Compare with Figure .