# Procesamiento sísmico

File:Aplicaciones Geofisicas de los filtros digitales-Mario Caicedo-Milagrosa Aldana.pdf -- It's been downloaded from this link [Mario Caicedo & Milagrosa Aldana]

This is a good explanation about the Relations and differences between time-series analysis and statistical signal processing? https://stats.stackexchange.com/questions/52270/relations-and-differences-between-time-series-analysis-and-statistical-signal-pr#:~:text=2%20Answers&text=As%20a%20signal%20is%20by,significant%20overlap%20between%20the%20two.

## Convetional Seismic data processing

### Yilmaz - 2001

 1 Pre-processing Demultiplexing Orders seismic data by seismic trace instead of being ordered by time. Multiplexed recording is still found in older data or in current data shots with old recording instruments. Démultiplexage: Reformatting (SEG-Y/SEG-D) Puts the data in a format that is understandable by computers (e.g. SEG-Y format, SEG-D format, etc.). Reformatage: Seismic data edition Searchin for noisy, monofrequency and incorrect polarities traces. Examines seismic data for bad, noisy, and\or monofrequency traces, and deletes them, and also spots any traces with incorrect polarities and corrects them. Édition: Geometrical spreading correction It is also called amplitude correction. Because the signal becomes weaker as it moves away from its source (i.e. its amplitude decreases with time), amplitude correction is needed to count for this lose, and to make amplitudes stronger. Correction de la dispersion géométrique: Set-up of field geometry - Geometry QC Incorporates field geometry with seismic data processing. Coordinates, shot\ receiver locations and spacing must be entered to the system carefully and precisely because processes like CMP sorting, for example, highly depend on this. Géométrie de terrain: Application of field statics corrections This is needed for land data collected on non-flat areas. It reduces the travel times into a common datum level. The datum could be the sea level or any other local datum. Corrections statiques: 2 Deconvolution and trace balancing Deconvolution makes the signal look better by increasing the temporal resolution and removing echoes: The seismic wave can be mathematically represented by what is called the Convolutional model for the seismic trace. This is a mathematical model that states that the Seismic trace is the result of the convolution between the Seismic Wavelet (The initial wave produced by the seismic source ${\displaystyle W(t)}$) and the Reflectivity function. The reverse calculation, which is the process of calculating the Reflectivity function from the trace is known as "Deconvolution". So, when we apply Deconvolution to real seismic data, we are looking to extract the reflectivity function from the seismic trace, and the reflectivity function should be a better representation of the reflectors in the subsurface. It should highlight structures better and increase the temporal resolution. Trace balancing makes the amplitude uniform. Déconvolution et normalisation des traces: 3 CMP sorting Sorts the data into CMP gathers, so it can be corrected for the NMO and stacked after that (see step 7). Triage par CMP: 4 Velocity analysis Gives info about velocities in the subsurface layers. It finds the stacking velocity (very close in value to the RMS velocity) that best fits our data (This step may be delayed after step 5). Analyse de vitesse: 5 Residual statics corrections Counts for near-surface velocity variations that causes some static and dynamic distortion problems. Corrections statiques résiduelles: 6 Velocity analysis (see step 4). If we have residual statics problems, we do velocity analysis after we count for them. Analyse de vitesse: 7 NMO Correction Counts for the increase in travel time with increasing offset distance. This increase makes flat reflectors look dipping, and makes dipping reflectors look even more dipping. The amount of correction needed decreases with depth, so that shallower reflectors get more “stretched” than do the deeper ones. Muting: It is just a fancy word for deleting a part of a trace. Muting a whole trace is called “killing”. Correction NMO: Correction DMO: Correction NMO inverse: Analyse de vitesse: Correction NMO: 8 Stacking After sorting the data into CMP gathers and applying the NMO correction, reflectors line up nicely and hence their stacking gives a stronger signal. Multiples\ random noises do not line up. Stacking increases S/N ratio by decreasing multiples\ random noise from the data which enhances the overall quality. It also reduces the seismic data volume to the plane of the seismic section Sommation: 9 Déconvolution: 10 Blanchiment de spectre à temps variable: 11 Time-variant band-pass filtering Filters unwanted "signals" based on their frequencies. For example, ground roll has a lower frequency (and higher amplitude) compared to the rest of the section, so we can filter it out based on that fact. [noise reduction] [time/frequency domain transforms] Filtre à temps variable: 12 Migration So far, each trace is plotted under its CMP location which puts reflectors in the wrong subsurface location. Migration process moves those reflectors into their true subsurface locations, and that improves lateral resolution. Migration also collapses diffractions into identifiable points on the seismic section. Migration: 13 Gain recovery Seismic energy gets lost in many different ways (e.g. scattering, frication, etc.). Gain is the inverse function of energy loss. It is very hard to predict the attenuation function because primary reflections, multiple reflections, and random noise have different decay functions. Instead, we examine (test) different gain values, and see which value gives a better looking data. Gain:

## Seismic sections

### Migracion pre-apilamiento en tiempo - Kirchhoff

Inline 1450 (Kirchhoff PreSTM).
Acercamiento sobre la inline 1450 (Kirchhoff PreSTM).

### Migracion pre-apilamiento en profundidad - Kirchhoff

PreSDM empleando distintos campos de velocidad (inline 1450). Las dos primeras, de izquierda a derecha, se corresponden con el modelo generado en 2008 y con el actual. La tercera ha sido migrada empleando un modelo similar al actual pero disminuyendo la velocidad de la capa de baja velocidad hasta 2100 m/s, y la cuarta es el mismo caso anterior pero con una velocidad de 1900 m/s.