Processing of 3D seismic reflection land data characterized by low signal-to-noise ratio and acquired in a complex geology area.

The goal of this thesis work is the processing of 3D seismic reflection land data acquired in a complex geology context and characterized by a poor signal-to-noise ratio to obtain a final 3D stack volume. The studied area extends for approximately 17 km2 and the acquisition is characterized by sources randomly distributed throughout the survey area while the receiver grid remains fixed. The sources used are Vibroseis and a total of 1887 shot gathers are acquired with 480 active receivers each. The recording length is 5000 ms. The biggest challenge of this data is represented by the occurrence of noise that affects, sometimes obliterating completely, the useful signal (i.e. reflections). Several types of noises, coherent and not coherent, contaminate the data and the most dominant and severe is the ground roll noise. In order to attenuate it, three filtering methods were tested, compared and then applied; that are Band-Pass filter (BP), Surface Wave Noise Attenuation (SWA) and Ground Roll Removal (GRR). The first is a standard procedure and operates in the frequency domain attenuating a predefined frequency range in a time and offset manner; SWA operates in the frequency-space domain attenuating surface wave by forming low-frequency arrays and the GRR is a Singular Value Decomposition based algorithm that after an initial estimate of the GR adaptively subtracts the ground roll. The results obtained were analyzed using the Signal to Noise Adaptive Processing (SNAP) tool combined with Wavelet Transform. This approach allowed a quality control on the removed signal by considering different frequency bandwidths allowing to assess the effectiveness of the methods. After the performed analysis, it turned out that GRR method is the most effective in GR attenuation. Moreover editing methods, such as kill traces and noisy trace editing (despiking and noise burst removal), were also used especially to attenuate strong noisy components distributed randomly throughout the data. Amplitude corrections were applied using a surface consistent approach (SCAC) to restore and correct amplitudes, which can give anomalous values due to instrumental causes. Subsequently, it was necessary to apply static corrections, which proved crucial for correcting those temporal shifts in arrival times due to differences in source and receiver elevation and different velocities of the weathering layer. The effectiveness of the static corrections was determined by computing constant velocity stack sections, in which a significant improvement in the coherence of the main reflectors was noted. After the denoising and static corrections steps, the Velocity Analysis was performed. However, due to the complex geology of the studied area and therefore the presence of dipping layers, it was necessary to apply the DMO (Dip Moveout) correction, which allowed to update the velocity functions in some parts of the data volume, especially in the shallow portions where the coherence of the main reflectors was improved by the DMO correction. Currently, due to the lack of an accurate velocity model and considering the low signal-to-noise ratio that affects the data, DMO provides more robust and effective results than preliminary test on pre stack migration. In order to eliminate residual noise, a predictive deconvolution was applied, after recovering the amplitudes with the geometrical spreading corrections. At the end of the processing a stacking velocity model was obtained; this was used to apply the Normal Move Out (NMO) correction and thus to obtain the final stack volume. The 3D stack volume estimated shows clear and strong reflectors mainly located in a time window between 1500 ms and 2000 ms. They exhibit a good but not constant lateral continuity and irregular shape revealing a complex geometry of the reflectors and inhomogeneous reflectivity properties according to the complex geology that characterize the studied area