GNGTS 2015 - Atti del 34° Convegno Nazionale

142 GNGTS 2015 S essione 3.3 original amplitude section. The algorithm uses this feature to track any event with lateral phase continuity, connecting signal phases with the same polarities and close arrival times. Given its independence from the reflection strength, the system can even track events characterized by lateral amplitude variations or changes in the shape of the reflected wavelet. The picking procedure is nevertheless sensitive to the presence of noise and interference in the recorded profile. For example, a hyperbolic diffraction can intersect an unrelated reflection and deform horizons originating from such reflection, while noise-related distortions of the signal phases can interrupt the tracking of an event. Therefore, the analyzed profiles may require a certain degree of processing before picking. The algorithm itself can reduce such effects by automatically connecting horizons which belong to the same coherent event but were separated by a few distorted phases. This is done by using close parallel horizons as patches across the gaps, and it has the effect of reducing the total number of horizons, while increasing their average length. Once all possible horizons have been constructed, they are superimposed on the recorded profile for visual interpretation. Optional constraints on the reflection strength can be set, in case the interpreter wants to selectively display events that reach specific energy thresholds. After picking, the algorithm automatically analyzes each horizon, searching for close sub- parallel events that can be considered part of the same reflection. The aim is to obtain single horizons representing each reflection, using either phase or energy-based constraints, in order to reduce the total number of displayed picked events, and therefore improve the interpretation. The phase-based method groups together sub-parallel horizons by reconstructing the reflected wavelet in each of them and then confronting their shapes. The wavelets are reconstructed by averaging the cosine phase along each horizon, which preserves the reflected signal while removing other unrelated events. If the two shapes are similar by means of cross-correlation, the two sub-parallel horizons are considered to be part of the same reflection. Under favorable conditions of high signal-to-noise ratio and absence of interfering events, this process (also called phase assessment) can be used to identify the initial phase (as well as later ones) in each reflection, and therefore its polarity, which allows to reconstruct the subsurface reflectivity. Specific phases can then be selected for further analysis and interpretation. The energy-based method is instead used when lateral signal variations, caused by either dispersion, noise, or interference, do not allow the accurate reconstruction and comparison of the reflected wavelets. In this case the recorded profile is separated into energy packages by defining the main peaks of the reflection strength. Sub-parallel horizons from the same package are then grouped into the same event using statistical analyses of their arrival times. Specific horizons can then be selected by using either energy or time-related constraints with respect to the peak reflection strength in each energy package. Examples of application. Stratigraphic interpretation (reflection seismics) . In this section, we apply the automated picking to a seismic profile acquired in 2010 in the west Mediterranean Sea by the Istituto Nazionale di Oceanografia e Geofisica Sperimentale (OGS), as part of the WS10 exploration project (Geletti et al. , 2014). Attribute analysis allows to clearly define the subsurface structures by disregarding the amplitude information, as it can be noticed from the cosine phase profile in Fig. 1A. In the analyzed pre-stack time migrated profile, three different domains can be identified, namely a sedimentary basin (S), the seismic basement (B) and a salt dome (D), all showing slightly different seismic signatures. As previously discussed, we applied the energy-based grouping method, due to the inherent variation of the reflected wavelet caused by seismic processing (e.g. stacking), which prevents its accurate reconstruction through cosine phase averaging. The horizon grouping results are shown in Fig. 1B, superimposed on the amplitude section. The algorithm divided the profile into energy packages and selected from each reflection those horizons with an average energy equal to at least 50% of the peak reflection strength in the respective package. This selection method can be sensitive to vertical resolution, which can cause close parallel reflections to be grouped into the same energy package so that, although accurately picked, the weaker one is not automatically displayed after grouping (e.g.