Seismic data processing
- 1. Computer Systems and Digital Filtering Concepts in Seismic Data Processing
- 2. The general impact of the digital computer upon seismic prospecting has been in the processing of seismic data. Such processing had its beginnings when field systems were introduced in the early 1950s for reproducible recording on analog magnetic tape. But it was not until modern high-speed digital computers became available that the full potential of processing techniques could begin to be realized for improving the quality and usefulness of seismic field data. The basic objective of all-seismic processing is to convert the information recorded in the field into a form that can be used for geological interpretation. The data initially recorded on magnetic tape (digital or analog) are transformed in the processing center into a record section comparable in some ways to a geological structure section.
- 3. One object of the processing is to eliminate or at least suppress all noise (defined here as signals not associated with primary reflections. Actually we are not particularly interested in reducing the absolute level of noise but in increasing the ratio of signal level to noise level, the signal-to-noise ratio. Another is to present the reflections on the record sections with the greatest possible resolution and clarity and in the proper geometrical relationship to each other. Reproducibly recorded data and playback facilities have made it economically feasible to implement new field recording systems designed to facilitate the suppression of noise relative to reflection signals. Many of these systems require processing of the data as an integral part of the data acquisition.
- 4. Among these are common-depth-point recording and vibroseis, as well as many marine systems. Most non-dynamite sources require summing of signals, which can be carried out either on special digital compositing units in the field or, at a later stage in a playback center. Data processing is a lengthy affair, comprising five major types of corrections or adjustments; time, amplitude, frequency-phase content, data compression, and data repositioning.
- 5. Seismic data are recorded in the field on magnetic tapes About three to six weeks later the information is transmitted to the interpreter as a seismic section All of the intervening steps comprise the data processing phase of seismic exploration Processing of seismic data consists of applying a sequence of computer programs, each designed to achieve one step along the path from field tape to record section
- 6. From ten to twenty programs are usually used in a processing sequence In each company’s library there may be several programs designed to produce the same effect by different approaches
- 7. The selection of the processing sequence for a given set of field data depends upon: Geologic environment Processing philosophy of the company Personal preference of the user Cost
- 8. Processing programs fall into four categories: Data reduction Geometric corrections Data analysis and Parameter optimization Data refinement Data presentation and storage
- 9. Some procedures are common to every processing sequence. Others are used only occasionally or rarely. Some processors always come at the same stage of the sequence. Others can be used in any of several stages. Some may be used more than once There is no standard processing recipe for all types of data
- 10. DATA ASSEMBLY AND PREPROCESSING Field tapes are often delivered to the processing centre in multiplexed format. They are accompanied by hand written notes prepared by members of the field party. These notes contain the information necessary to identify and process the data, including the following:
- 11. Line number & shot numbers Location of source and receiver stations Elevation of source & Receiver stations Station interval Shot hole depths and charge sizes Date and time of recording Uphole times may be read and recorded by hand, or they may be read automatically by the computer. All of this information is recorded on scratch tape.
- 12. When the field tape is demultiplexed, the archival data is merged with the field tape. These data are recorded in the reel header and trace header blocks on an output tape, which later becomes the input to the main processing run stream
- 13. ROUTINE PROCESSES AND CORRECTIONS Making CDP gathers Restoring true amplitudes Applying static corrections Analyzing and applying filters Running velocity analysis and applying NMO corrections Stacking
- 14. PRELIMINARY EVALUATION The brute stack is usually regarded as a trial run for quality control. It is played out directly on paper and is not recorded on film for permanent retention. Processors examine the brute stack for evidence of residual problems that may need special processing.
- 15. At this stage it often helps to examine the preliminary displays. Corrected CDP gathers may be especially useful, since they provide the last chance to see the field traces before they are stacked.
- 16. DEMULTIPLEXING The output of a geophone group during field recording is a continously varying electrical voltage A graph of this output signal showing voltage as a continous function of time is one form of a seismic trace, an analog display Since seismic data are processed in digital computers, the continuous geophone output voltage must be converted to digital format by sampling
- 17. Sampling occurs in the field or in the processing centre in an analog to digital converter (A/D) Consider for geophone arrays, whose outputs are brought to four terminals arranged in a circular pattern The instataneous voltage at each terminal is recorded each time the rotating arm sweeps over it, yielding an array of samples
- 18. If we identify each sample by its geophone group source (A, B, C, D) and by its chronological sequence in that group (1, 2, 3, …) Then output of A/D converter is A1,B1,C1,D1,A2,B2,C2,D2,A3,B3,C3,D3,…… This scrambled sequence is referred to as a “multiplexed” array
- 19. However, it is more convenient to process seismic data in trace sequential array A1,A2,A3,…B1,B2,B3…C1,C2,C3,….,D1,D2,D3,….. Unscrambling a multiplexed array into a trace sequential array is called “Demultiplexing” It is accomplised by a simple computer sorting program and is the first step in any data processing sequence
- 20. DATA REDUCTION
- 21. Data Reduction Introduction Data reduction is the first and most important step in data processing flow. It includes following four categories of software programs: i. Demultiplex ii. Display iii. Edit iv. Amplitude adjustment
- 22. Demultiplex Demultiplexing in the geophysical sense is the unscrambling of multiplexed field data to trace sequential form. To the data processor, demultiplex has come to mean a whole family of processing functions that affect or are affected by the specific process of demultiplexing. Demultiplexing will be explained in the following six major steps. a. Tape Dump and Format Determination b. Preliminary checks c. Binary gain recovery d. Vibroseis correction e. Summing f. Diversity stacking
- 23. Tape Dump and Format Determination A tape dump is a printed listing of the data on tape. A tape dump could be printed in binary but this would generate a large quantity of paper. It is much more convenient to convert binary numbers to “hexadecimal” for listing purposes. Hexadecimal is a base 16 number system using the characters 1,2,3,4,5,6,7,8,9, A,B,C,D,E, and F. Four bits are required to express one hexadecimal number. Hexadecimal numbers are often written preceded by an “X” to identify them e.g. ‘X16B8C1AE.DC’. To convert X64C1B to binary: Hex 6 A C 1 B Binary 0110 1010 1100 0001 1011 i.e. X6AC1B: 01101010110000011011
- 24. Preliminary Checks The initial tests should be as simple as possible. Demultiplexing 5-10 record should be adequate for the purpose. These runs should not contain any option except resample and / or binary gain removal. In this manner one can see basic data undisturbed by gain curve application or correlation. Output of auxilia channels (time break, uphole etc) also are helpful at this stage. At this time, the processor has a tape format sheet, a hex- dump, a rim time printout from computer and a display of demultiplexed data. With the help this data follows field tape errors can be diagnosed and removed.
- 25. Rate Errors. A rate error occurs when the multiplexed data is coming too fast form the magnetic tape to the disk. This normally occurs when the wrong format has been specified or there are too many zeros ahead of the first sync frame. Binary Gain Recovery Modern digital seismic instruments use various gain ranging techniques to increase their dynamic range. This is necessary because of the very rapid decay of seismic signal strength with time.
- 26. Gain ranging instruments apply a gain to this signal before recording it and also keep a record of what gains were applied. Since the gain applied to the signal are recorded (usually on auxiliary channels) along with the signal, the processing centre can “de-gain” the data, that is, remove the gain applied by the field instruments. To correct for the rapid decay of signal energy, a gain function may be applied. A gain function is a smoothly time varying gain applied to the data to bring the overall amplitude up to a fairly constant level. This is referred to as gain recovery.
- 27. Vibroseis Correlation Most seismic methods use an energy source that generates a very short pulse. When these pulses are reflected and recorded they can be used directly to examine subsurface structure. The signal generated in the vibroseis method is not a short pulse, but rather a “sweep” lasting some seven seconds or longer. The sweep is sent into the earth by a pad that is pressed against the ground surface and vibrated. The sweep is transmitted through the earth and reflected as is any seismic signal, but on vibroseis record, each reflection is as long as the input sweep rather than being a pulse or wavelet as with an impulsive source.
- 28. Each reflection is a near duplicate of the sweep itself and so the reflections in the vibroseis record overlap and are generally indistinguishable. To make the vibroseis record useable, we must “compress” the reflections into short wavelets. This is done by cross correlating the data with the original input sweep. If we were to correlate a sweep with itself, the result would be a wavelet at time zero containing all of the frequencies in the sweep. This an autocorrelogram of the sweep and is called a “Klauder wavelet”
- 29. Similarly, when a vibroseis trace is correlated with a sweep, the result is that a wavelet is output every time a reflected sweep is encountered. On the correlated output record. Each reflection has been “compressed” to a wavelet. Thus correlated vibroseis data looks similar to impulsive source data and can be treated in much the same way. There are some important differences, however, which will be discussed later in this course.
- 30. Header Generation After all of the samples from a given field trace are assembled into an array, a large amount of archival information is placed in a reserved block called a trace header, which is located on the tape just ahead of the data samples. Trace header information may include location and elevation of source and receiver, field record number, trace number, etc. A real header block is also placed at the head of each reel, for recording line number, reel number etc.
- 31. Display At the end of the processing sequence, and sometimes at intermediate stages, it is necessary to reconvert the digitized data to analog format so that they can be displayed and then examined visually. There are several methods of plotting digital data in analog form. The most common method is to pass the digital trace through a sample and hold device, which produces an output voltage proportional to each sample and which holds the voltage constant until the next sample arrives. This procedure produces a “Stair step” type of analog output which can then be smoothed into a continuous trace by passage through a low pass filter.
- 32. Editing Raw seismic data inevitably contains some unwanted noise and perhaps some dead traces. If obviously useless information is to be removed from the processing stream, it must first be identified and then blanked or muted by assigning zero values to all samples in the affected time interval. Unwanted data are usually identified by visual examination of raw field traces, although obvious cases (dead traces, strong noise bursts etc.) can be detected and edited automatically. The raw field trace used in this editing procedure can be obtained from field monitor records or from one of the trace gathers previously described.
- 33. Amplitude Adjustment In a seismic section the variations in the amplitudes of reflection can be important factors in the interpretation Lateral amplitude variations from trace to trace, within a reflection event (bright spot) may be direct indications of the presence of hydrocarbons. Vertically amplitude variations, from event to event may be helpful in identifying and correlating reflecting horizons. To preserve these important amplitude variations, the seismic analyst must exercise care in applying gain recovery and scaling (trace equalization). In some cases, however, the amplitude variations are so great that low level events become difficult to follow or even invisible In order to raise the level of these weak events relative to the strong ones so that geologic structure can be made visible, the analyst can apply a digital balance or “ AGC” (automatic gain control).
- 34. Geometric Corrections A seismic trace on a field monitor shows reflected energy bursts form subsurface rock layer interfaces. We will later measure the travel times form source down to reflector and back to geophone and use them together with average velocity information (if we have it) to compute depths to the various reflectors. However, before we use these reflected energy bursts and their travel times, we must apply several corrections to compensate for geometric effects. The corrections include static correction and normal move out corrections applied on the trace gathered data.
- 35. Trace Gathering Traces are routinely gathered into groups having some common elements. The types of gathers usually made are: - Common source point - Common depth point - Common receiver - Common offset The information needed to generate these various gathers may be contained in the trace headers already inserted by some geometry definition programs.
- 36. Static Corrections Static corrections are constant for an entire trace. They consist of weathering corrections and elevation corrections. There is usually a thin layer – from several feet to several hundred feet thick - of low velocity material (1000 to 2500 ft/sec) at the earth’s land surface, underlain by rock of much higher velocity (more than 5000 ft/sec). Geophysicists call this low velocity zone the weathered layer. Geologists use the same term to describe a generally thinner surface zone in which the elements lend to decompose rock into soil. The thickness and velocity of the weathered layer can change along a seismic line, there by causing changes in raw reflection time which are unrelated to subsurface structure.
- 37. This effect must be removed by applying a weathering correction. Since seismic sources and receivers are usually at or near the earth’s surface, raw reflection times are influenced by topographic effects, which are also independent of subsurface structure. These effects are removed by elevation corrections.
- 38. Dynamic Correction (NMO) In general, shot points and their associated geophone stations are separated by distances ranging up to three kms or even more. The slant, or non-vertical component of travel time can be large and must be removed before we can attempt to relate reflection time to reflection depth. For a flat reflector, the difference between travel time to a remote geophone station and travel time to a coincident (zero offset) station is called the normal move out (NMO). The purpose of this process is to derive normal moveout functions from the velocity-time functions and apply normal moveout corrections to each trace in each CDP gather. Normal moveout is a function of reflection-time and trace offset (source to detector distance) and, thus, a normal moveout time function must be derived for each trace offset.
- 39. Data analysis and parameter optimization Filtering A filter is a system, which discriminates against some of its input. Seismic data always contain some signal information, which we want to preserve. Everything else is called noise, and we want to remove or reduce it. In a sense, almost all of the processes we use are filters of a sort, because they are designed to preserve signals and attenuate noise. Those systems, which are generally called filters work either by convolution in the time domain or by spectral shaping in the frequency domain.
- 40. The most common types of filters are the following: - Low pass frequency filters - High pass frequency filters - Band pass frequency filters - Notch frequency filters (attenuate a narrow frequency band) - Deconvolution filters (shorten pulses and/ or reverberate) - Velocity filters (multi-channel filters which attenuate seismic events having specified range of apparent velocities)
- 41. Deconvolution Deconvolution is a filtering process designed to improve resolution and suppress multiple reflections. Deconvolution can be considered either in the time domain or in the frequency domain. In the time domain the object is to convert each wavelet, with its reverberations and multiple, into a single spike. If we know the shape of the wavelet, we can design an operator which, when convolved with the seismic trace, will convert each wavelet into a single spike.
- 42. Seismic Velocities In this process velocity time functions are derived from reflections by a scanning technique velocity “spectra” are produced and plotted for interpretation by the geophysicist. Supplemental data, usually from well velocity surveys, also may be used to increase accuracy of the functions. This analysis is performed continuously such that velocity time functions are derived at intervals of, generally, one kilometer along the seismic line.
- 43. Processing Sequence/ Flow Chart Successful processing requires selecting the appropriate programs and parameters for a given set of data. Several diagnostic programs, e.g. velocity analysis, DCON analysis and spectral analysis can be used to reveal details which help in the choice of data enhancement and refinement techniques. The finalization of this set of programs to process a special project or group of seismic profiles are managed in a flow chart or processing sequence.
- 44. Data Refinement The processes described thus for are used to reformat, correct and diagnose data characteristics. This category includes various procedures intended for data refinement. Stacking After static and NMO corrections have been applied, primary reflections should appear as horizontal events across the gather, and a normalized sum of the traces should then form one composite trace in which reflections are preserved at full strength, while all other energy is reduced. This procedure is known as stacking or composting Stacked data are also referred to as CDP or CRP (common or reflection point) data.
- 45. Some geophysicists prefer the term CMP (common midpoint), in recognition of the fact that the reflection point of the gathered traces do not coincide unless all reflectors are flat. Stacking came into general use in the early 1960s. Its original purpose was to attenuate multiple reflections, which are echoes of marine water bottom reflections or of strong reflections from subsurface layers. It has also proved useful in suppressing other kinds of noises both organized and random.
- 46. Residual Statics Routine static corrections mentioned earlier, are often not adequate to correct satisfactorily for near surface effects. Migration. A trace on a record section can be thought of as an idealized field trace recorded by one geophone station from a shot at the same station. Note that from one trace it is generally not possible to distinguish signal for noise. When a trace is plotted vertically beneath its common shot/ receiver station. A reflected event is identified by its alignment across an array of traces, and the reflection time is measured under each station. If the reflector is flat, the reflection point will be located directly beneath the shot/ receiver station, and the record section displays the event in its true position plotted in time rather than depth.
- 47. The process of re-plotting seismic event in their true spatial positions is called migration, and there are several methods of implementing this process. Data Presentation And Storage It involves the labeling of processing sequence, field information, velocity data and other useful parameters on the final section. Moreover the storage of essential seismic data for future use is also an important subject of data processing concern.