A high-resolution survey with a total of 2300 km 2 of 3D reflection seismic data was acquired over the Al Shaheen Field, Block 5, offshore Qatar, from October 2006 to April 2007. Reverberations from the hard shallow water bottom and other types of noise limited the value of the original processing. Re-processing using new technologies has produced a step change in quality resulting in cleaner images for interpretation and superior angle stacks for inversion.The multiples were removed by applying an enhanced processing workflow based on a predictive, data-driven algorithm. The workflow involved attenuating short-period water-layer related multiples -a process that is referred to as shallow water demultiple (SWD). The SWD method makes use of water-layer multiples in the data to reconstruct the missing water-bottom primary reflection, and then uses the reflection for predicting these shallow multiples. The method takes into account the spatial varying nature of the subsurface. Since the multiple model predicted by SWD has similar amplitude and phase as the input data, very short matching filters can be utilised in the adaptive subtraction process.These processing improvements have influenced a broad spectrum of interpretations such as better structural representation including fault mapping and improved understanding of facies. In conclusion, detailed and careful testing has resulted in new added value from this large high resolution 3D data set by applying technologies that were not available when the data was acquired and processed in 2006-2008.

No abstract

No abstract

Getting a good image by stacking after normal moveout (NMO) correction and/or after prestack migration requires reasonably flat gathers. Amplitude variation with offset (AVO) studies on prestack data require flat gathers after NMO. Otherwise slope and intercept attributes get contaminated. When the raypath of the seismic energy includes inhomogeneities, NMO algorithms based on flat layer assumptions, even prestack time migration, may fail to produce flat gathers ( Figure 1). Such gathers may require a robust, brute force, gather flattening method.The need for such an application was shown by Hinkley et al. (2004) and a solution which they called "dynamic gather flattening (DGF)" was presented. The method in this paper is similar in that it is a brute force (i.e., nonphysical) gather-flattening method. We will refer to it as FLATN for brevity. How can we correct gathers which have residual, even abnormal, moveout? First, such a correction should be done as a one-to-one mapping. That is, every output data sample should come only from one input data sample on the same trace:where x is offset (or trace number of the gather sorted from smallest to largest offset), t is time, a stands for after and b for before, and m(t,x) is the moveout function. To do flattening, we first estimate the moveout function by tracking the events (wavelet) across traces at each zero offset (t 0 ) time sample. Starting at the innermost traces and at time t 0 , we track the event times, t, of the wavelet across the gather to get the moveout function. This can be done by cross-correlation of neighboring traces (the "two-trace correlation method") and integrating the static shifts so calculated.In order to track an event, one needs to allow a correlation window of two traces to climb up and down as the event moves. So, cross-correlation of two nearby traces may not be centered at the starting time t 0 after one moves away from inner traces. Repeating the same process for other near-offset times t 0 completes the determination of the moveout function to be applied to the input gather at time t and offset x.Note that, while picking the static shift that gives the cross-correlation maximum, one should use the maximum of the absolute value so that static shift resulting with negative cross-correlation peaks is not eliminated from the picks. Also, sudden changes in pick times cause wavelet stretch and squeeze in the moveout function and this needs to be smoothed after some spatial consistency check. Afterwards, the moveout values can be output for QC displays or for further processing. Figure 2a shows a simple (and noise-free) gather of parabolic events with large residual moveout values that we used while developing FLATN. Here, 120-ms time windows centered around each t 0 time were used during cross-correlation of two consecutive traces; a maximum trace-to-trace static of 12 ms in inner offsets and 36 ms maximum at far offset was used. Figure 2b shows the moveout function obtained with the twotrace, event-tracking algorithm on this gather. Red ...

No abstract